Resource allocation method for machine room

Abstract

The invention provides a resource allocation method for a machine room. The method comprises the following steps: inputting a plurality of pieces of first machine room efficiency influence data corresponding to a first machine room into a machine room efficiency evaluation model to obtain first machine room efficiency evaluation results corresponding to the plurality of pieces of first machine room efficiency influence data; determining target machine room efficiency influence data from the multiple pieces of first machine room efficiency influence data according to first machine room efficiency evaluation results corresponding to the multiple pieces of first machine room efficiency influence data; performing resource configuration on the first machine room according to first machine room influence dimension data included in the target machine room efficiency influence data; the machine room efficiency evaluation model is obtained by training an artificial intelligence model by using the second machine room efficiency influence data and a machine room efficiency real result; the computer room cost related data is obtained by processing a computer room cost image based on a deep learning image processing strategy, and the authenticity of the computer room cost image and the accuracy of the computer room cost related data are audited according to a predetermined auditing rule.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a resource configuration method for a computer room. Background Art

[0002] In order to ensure the normal operation of the computer room, a lot of expenses need to be invested. In order to reduce operating costs, the communications industry has included energy conservation and emission reduction work into its work priorities. As the focus of energy conservation and emission reduction, the computer room needs to strengthen energy conservation measures and reduce energy consumption.

[0003] In the related art, the text that can provide the data of the computer room cost dimension is manually audited. If the manual audit is passed, the computer room cost dimension data obtained from the text is used to evaluate the energy consumption of the computer room, so as to guide the energy saving of the computer room, that is, to achieve energy saving of the computer room from the perspective of how to reduce costs.

[0004] In the process of implementing the present application, it was found that there are at least the following problems in the relevant technology: the accuracy of the energy consumption assessment of the computer room is low and it is difficult to provide a basis for the resource allocation of the computer room. Summary of the invention

[0005] The present application provides a resource configuration method, device, electronic device, storage medium and computer program product for a computer room, which can ensure the accuracy of computer room performance evaluation and achieve reasonable computer room resource configuration.

[0006] In a first aspect, the present application provides a resource configuration method for a computer room, comprising: inputting a plurality of first computer room performance impact data corresponding to a first computer room into a computer room performance evaluation model, obtaining first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data, wherein the first computer room performance impact data includes first computer room cost dimension data, first computer room revenue dimension data and first computer room impact dimension data, and the first computer room impact dimension data represents data that affects the first computer room cost dimension data and the first computer room revenue dimension data; determining target computer room performance impact data from the plurality of first computer room performance impact data according to the first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data; and performing resource configuration on the first computer room according to the first computer room impact dimension data included in the target computer room performance impact data; Wherein, resource allocation includes at least one of the following: computer room site selection and equipment selection; wherein, the computer room efficiency evaluation model is obtained by training an artificial intelligence model using the second computer room efficiency impact data and the real results of the computer room efficiency, the second computer room efficiency impact data includes second computer room cost dimension data, second computer room revenue dimension data and second computer room impact dimension data, the second computer room impact dimension data represents data that affects the second computer room cost dimension data and the second computer room revenue dimension data, and the real results of the computer room efficiency represent the real computer room efficiency results of the second computer room; wherein, the computer room cost-related data includes the second computer room cost dimension data, the computer room cost-related data is obtained by processing the computer room cost image based on the image processing strategy of deep learning, and the authenticity of the computer room cost image and the accuracy of the computer room cost-related data are audited and passed by predetermined audit rules.

[0007] The resource allocation method of the present application, since the second computer room efficiency impact data includes the second computer room cost dimension data, the second computer room revenue dimension data and the second computer room impact dimension data, the second computer room cost dimension data represents the cost data invested in maintaining the operation of the second computer room, the second computer room revenue dimension data represents the business data generated by operating the second computer room, and the second computer room impact dimension data represents the data that affects the second computer room cost dimension data and the second computer room revenue dimension data, therefore, the second computer room efficiency impact data is relatively comprehensive. In addition, the computer room cost-related data includes the second computer room cost dimension data, and the computer room cost-related data is obtained by processing the computer room cost image based on the image processing strategy of deep learning. The authenticity of the computer room cost image and the accuracy of the computer room cost-related data are audited by the predetermined audit rules, therefore, the accuracy of the second computer room cost dimension data is guaranteed. On this basis, since the computer room efficiency evaluation model is obtained by training the artificial intelligence model using the second computer room efficiency impact data and the real results of the computer room efficiency, therefore, the reliability and accuracy of the computer room efficiency evaluation model are guaranteed. Since the first computer room performance evaluation result is obtained by processing the first computer room performance impact data using the computer room performance evaluation model, the accuracy of the first computer room performance evaluation result is guaranteed. Since the first computer room is resource-configured based on the first computer room impact dimension data included in the target computer room performance impact data, the target computer room performance impact data is determined from the multiple first computer room performance impact data based on the first computer room performance evaluation results corresponding to each of the multiple first computer room performance impact data, thus achieving reasonable configuration of computer room resources.

[0008] In a second aspect, the present application provides a resource configuration device for a computer room, comprising: an acquisition module, used to input multiple first computer room performance impact data corresponding to a first computer room into a computer room performance evaluation model, and obtain first computer room performance evaluation results corresponding to each of the multiple first computer room performance impact data, wherein the first computer room performance impact data includes first computer room cost dimension data, first computer room revenue dimension data and first computer room impact dimension data, and the first computer room impact dimension data represents data that affects the first computer room cost dimension data and the first computer room revenue dimension data; a determination module, used to determine target computer room performance impact data from multiple first computer room performance impact data according to the first computer room performance evaluation results corresponding to each of the multiple first computer room performance impact data; and a configuration module, used to configure the first computer room performance impact data according to the first computer room impact dimension data included in the target computer room performance impact data A computer room is configured with resources; wherein the resource configuration includes at least one of the following: computer room site selection and equipment selection; wherein the computer room performance evaluation model is obtained by training an artificial intelligence model using the second computer room performance impact data and the real results of the computer room performance, the second computer room performance impact data includes the second computer room cost dimension data, the second computer room revenue dimension data and the second computer room impact dimension data, the second computer room impact dimension data represents the data that affects the second computer room cost dimension data and the second computer room revenue dimension data, and the real results of the computer room performance represent the real computer room performance results of the second computer room; wherein the computer room cost-related data includes the second computer room cost dimension data, the computer room cost-related data is obtained by processing the computer room cost image based on the image processing strategy of deep learning, and the authenticity of the computer room cost image and the accuracy of the computer room cost-related data are audited and passed by predetermined audit rules.

[0009] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores instructions, and when the instructions are executed by the processor, the method described in the first aspect is implemented.

[0010] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0011] In a fifth aspect, an embodiment of the present application provides a computer program product, which may include a computer program. When the computer program is executed by a processor, the method described in the first aspect is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0013] Figure 1 Schematically shows an example schematic diagram of processing a computer room expense image using an image processing strategy based on deep learning to obtain computer room expense related data;

[0014] Figure 2 The schematic diagram schematically shows the principle of resource allocation based on computer room performance evaluation according to an embodiment of the present application;

[0015] Figure 3 The exemplary system architecture of the training method of the computer room performance evaluation model, the resource configuration method and the device for the computer room according to the embodiment of the present application is schematically shown;

[0016] Figure 4 A schematic diagram schematically shows a principle of processing a computer room expense image according to an image processing strategy based on deep learning according to an embodiment of the present application to obtain computer room expense related data;

[0017] Figure 5 A flowchart of a method for training a computer room performance evaluation model according to an embodiment of the present application is schematically shown;

[0018] Figure 6 A flowchart of another method for training a computer room performance evaluation model according to an embodiment of the present application is exemplified;

[0019] Figure 7 An example schematic diagram schematically shows a training process of a computer room performance evaluation model according to an embodiment of the present application;

[0020] Figure 8 A flow chart of a resource configuration method for a computer room according to an embodiment of the present application is schematically shown;

[0021] Fig. 9 A block diagram schematically shows a resource configuration device for a computer room according to an embodiment of the present application; and

[0022] Fig.10 A block diagram of an electronic device according to an embodiment of the present application is exemplarily shown. DETAILED DESCRIPTION

[0023] The term "and / or" mentioned in this article is a kind of relationship that describes the association of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The symbol " / " in this article indicates that the associated objects are in an or relationship, for example, A / B means A or B.

[0024] The terms "first" and "second" in the specification and claims herein are used to distinguish different objects rather than to describe a specific order of objects. For example, a first feature representation and a second feature representation are used to distinguish different feature representations rather than to describe a specific order of feature representations.

[0025] In the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0026] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more than two. For example, multiple processing units refer to two or more processing units, etc.; multiple elements refer to two or more elements, etc.

[0027] It should be noted that the sequence numbers of the operations in the following method are only used as representations of the operations for the purpose of description, and should not be regarded as representing the execution order of the operations. Unless explicitly stated, the method does not need to be executed completely in the order shown.

[0028] The following will describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] To facilitate understanding, some concepts involved in this application are first explained.

[0030] A base station is one of the devices in a wireless communication network, which can be used to communicate and transmit data with wireless devices, realize wireless signal transmission, data transmission, connection and switching management, and spectrum management. Base stations can be divided into 4G (4th Generation Mobile Communication Technology) base stations and 5G (5th Generation Mobile Communication Technology) base stations. A base station can include a main device and an auxiliary device. The main device can include an antenna, a wireless transmission device, and a signal processing device. The auxiliary device can include a transmission medium, a refrigeration device, a power supply device, and a monitoring device. The antenna can be used to send and receive signals to and from a wireless device. The wireless transmission device can be used to transmit and receive radio frequency signals. For example, the wireless transmission device can include a radio frequency transmitter and a radio frequency receiver. The signal processing device can be used to process and forward signals sent and received by the wireless device. For example, the signal processing device can include a signal processor, a modem, and a codec. The transmission medium can be used to connect the base station to the core network. For example, the transmission medium can include an optical fiber.

[0031] The base station room can be equipped with signal processing equipment, refrigeration equipment, power supply equipment, monitoring equipment, transmission media, etc. In addition, there are also convergence rooms and core rooms. The convergence room can be used to connect the base station room and the core room. The base station room, convergence room and core room can be referred to as communication rooms. It should be noted that the room involved in the embodiment of the present application can refer to a communication room. The communication room can be any one of the base station room, convergence room and core room.

[0032] The data affecting the efficiency of the computer room may represent data affecting the efficiency of the computer room. The data affecting the efficiency of the computer room may include at least one of the following: computer room cost dimension data, computer room income dimension data, and computer room impact dimension data.

[0033] The computer room cost dimension data may represent the cost data invested in maintaining the operation of the computer room. For example, the computer room cost dimension data may include at least one of the following: computer room electricity fee, computer room maintenance fee, service fee, computer room rental fee, etc.

[0034] The data of the revenue dimension of the computer room may represent the business data generated by operating the computer room. For example, the data of the revenue dimension of the computer room may include at least one of the following: the flow of the computer room and the volume of the telephone traffic of the computer room.

[0035] The computer room impact dimension data may represent data that affects the computer room cost dimension data and the computer room revenue dimension data. For example, the computer room impact dimension data may include at least one of the following: environmental impact dimension data and equipment impact dimension data. The environmental impact dimension data may include at least one of the following: computer room location, computer room ambient temperature, the number of computer room associated base stations, the number of computer room associated areas, and the number of computer room associated objects. The number of computer room associated base stations may represent the number of base stations served by the computer room. The number of computer room associated areas may represent the number of areas served by the computer room. The number of computer room associated objects may represent the number of objects served by the computer room. The equipment impact dimension data may include at least one of the following: equipment attribute data and equipment health data. The equipment attribute data may include equipment attribute dimension data corresponding to at least one device respectively. The equipment attribute dimension data may include at least one of the following: equipment model, equipment performance, and equipment manufacturer, etc. The equipment health data may include the equipment health of at least one device respectively. The equipment health may represent the health of the device.

[0036] The computer room expense image may refer to an image that can provide data related to the computer room expense. For example, the computer room expense image may be obtained by photographing a text that can provide data related to the computer room expense using an image acquisition device. For example, the text may include at least one of the following: a payment slip, an invoice, an audit slip, a maintenance work order, and a contract. The image acquisition device may include at least one of the following: a scanner, a high-speed scanner, and a camera. It should be noted that the computer room expense related data may include computer room expense dimension data. In addition, the computer room expense related data may also include computer room expense auxiliary dimension data. The computer room expense auxiliary dimension data may represent data used to audit the accuracy of the computer room expense dimension data. The computer room expense auxiliary dimension data may include at least one of the following: computer room electricity expense auxiliary dimension data, computer room maintenance expense auxiliary dimension data, and computer room rental fee auxiliary dimension data. The computer room electricity expense auxiliary dimension data may represent auxiliary data corresponding to the computer room electricity expense. The computer room maintenance expense auxiliary dimension data may represent auxiliary data corresponding to the computer room maintenance expense. The computer room rental fee auxiliary dimension data may represent auxiliary data corresponding to the computer room rental fee.

[0037] The authenticity of the computer room expense image can be reflected by at least one of the following: whether the computer room expense image is edited, whether there are repeated computer room expense images, and the consistency of the computer room expense image.

[0038] The accuracy of the data related to the computer room fees can be reflected by at least one of the following: whether the payment period of the computer room fees is accurate, the accuracy of the computer room fees, whether the power consumption is accurate, whether there are associated equipment, and whether there are associated electricity charges.

[0039] The computer room performance result can represent the degree of influence of the computer room performance influencing data on the computer room performance. The computer room performance result can be represented by the ratio between the computer room revenue dimension data and the computer room cost dimension data.

[0040] The computer room performance evaluation result can represent the predicted computer room performance result of the computer room.

[0041] The real result of the computer room performance can represent the real result of the computer room performance of the computer room. It should be noted that in the embodiment of the present application, the real result of the computer room performance refers to the real result of the computer room performance of the second computer room.

[0042] The resource configuration of the computer room may refer to the configuration of relevant resources involved in the computer room. For example, the resource configuration of the computer room may include at least one of the following: computer room site selection and equipment selection.

[0043] The first computer room performance impact data may refer to the computer room performance impact data corresponding to the first computer room. The first computer room performance impact data may include first computer room cost dimension data, first computer room revenue dimension data, and first computer room impact dimension data. The first computer room cost dimension data may refer to the computer room cost dimension data corresponding to the first computer room. The first computer room revenue dimension data may refer to the computer room revenue dimension data corresponding to the first computer room. The first computer room impact dimension data may refer to the computer room impact dimension data corresponding to the first computer room.

[0044] The second computer room performance impact data may refer to the computer room performance impact data corresponding to the second computer room. The second computer room performance impact data may include second computer room cost dimension data, second computer room revenue dimension data, and second computer room impact dimension data. The second computer room cost dimension data may refer to the computer room cost dimension data corresponding to the second computer room. The second computer room revenue dimension data may refer to the computer room revenue dimension data corresponding to the second computer room. The second computer room impact dimension data may refer to the computer room impact dimension data corresponding to the second computer room.

[0045] The first computer room performance evaluation result may represent a predicted computer room performance result of the first computer room.

[0046] The performance evaluation result of the second computer room can guarantee the predicted computer room performance result of the second computer room.

[0047] It should be noted that the first computer room performance impact data may be computer room performance impact data used for model application of the computer room performance evaluation model. The first computer room performance evaluation result may be a computer room performance evaluation result corresponding to the first computer room performance impact data. The second computer room performance impact data may be computer room performance impact data used to participate in model training of the computer room performance evaluation model. The second computer room performance evaluation result may be a computer room performance evaluation result corresponding to the second computer room performance impact data.

[0048] It should also be noted that the first computer room can be a computer room to be built or a computer room that has been built, and the second computer room can be a computer room that has been built.

[0049] The text detection results can include the text location. The object detection results can include the object location and object category.

[0050] The present application will be described in its entirety below.

[0051] In the related art, the text that can provide the data of the computer room cost dimension is manually audited. If the manual audit is passed, the computer room cost dimension data obtained from the text is used to evaluate the energy consumption of the computer room, so as to guide the energy saving of the computer room, that is, to achieve energy saving of the computer room from the perspective of how to reduce costs.

[0052] In the process of implementing this application, it is found that since the relevant technology evaluates the energy consumption of the computer room based on the computer room cost dimension data, and lacks data related to the computer room revenue, the method in the relevant technology cannot achieve the computer room performance evaluation. In addition, since there is a possibility of text tampering, it is difficult to ensure the authenticity of the text, and it is difficult to manually identify the authenticity of the text. Therefore, the method in the relevant technology of manually auditing the text that can provide the computer room cost dimension data cannot effectively ensure the accuracy of the computer room cost dimension data. Therefore, the accuracy of the energy consumption evaluation of the computer room based on the computer room cost dimension data is also relatively low. In the process of implementing this application, it is found that the configuration of resources will affect the performance of the computer room. Therefore, if an accurate computer room performance evaluation can be achieved, it can provide a basis for the reasonable configuration of the resources of the computer room, that is, it can be reasonable. Allocate resources from the perspective of the performance evaluation of the computer room.

[0053] To this end, it is necessary to achieve accurate computer room performance evaluation, that is, it is necessary to ensure the accuracy of the computer room performance evaluation results. The accuracy of the computer room performance evaluation results is affected by the comprehensiveness and accuracy of the data involved in the computer room performance evaluation. Therefore, it can be considered from the comprehensiveness and accuracy of the data involved in the computer room performance evaluation. In addition, since the method of obtaining the data involved in the computer room performance evaluation will affect the accuracy of the data, the method of constructing the computer room performance evaluation will also affect the accuracy of the performance evaluation data. Therefore, it can be considered from the reliability of the method of obtaining the data involved in the computer room performance evaluation and the reliability of the method of constructing the computer room performance evaluation. Accordingly, reasonable resource allocation from the perspective of computer room performance evaluation can be achieved from the data level and the method level. For the sake of simplicity, the data involved in the computer room performance evaluation will be referred to as computer room performance impact data below.

[0054] Regarding the data level, in the process of implementing this application, it is found that the data affecting the efficiency of the computer room can include the computer room cost dimension data, so the computer room cost dimension data can be used as the dimension data of the computer room efficiency impact data. However, the computer room cost dimension data only reflects the funds invested in maintaining the operation of the computer room, and the operation of the computer room, in addition to the need to invest funds, can also obtain income, therefore, the computer room income dimension data can also be used as the dimension data of the computer room efficiency impact data. In addition, it is also found that there are data that can affect the computer room cost dimension data and the computer room income dimension data. For example, the computer room ambient temperature will affect the computer room electricity bill. The computer room location will affect the computer room rental fee. The number of computer room associated objects will affect the computer room traffic and the computer room traffic volume. Equipment performance and equipment health will affect the computer room maintenance fee. Equipment manufacturers will affect equipment performance and equipment health, etc. Therefore, the computer room impact dimension data can also be used as the dimension data of the computer room efficiency impact data. So far, considering the comprehensiveness of the computer room efficiency impact, it is determined that the computer room efficiency impact data can include computer room cost dimension data, computer room income dimension data, and computer room impact dimension data.

[0055] In terms of methods, in order to ensure the reliability of the construction method of the computer room performance evaluation, the computer room performance evaluation model can be constructed by using training samples to train the artificial intelligence model. The training samples can include the second computer room performance impact data corresponding to the second computer room and the actual results of the computer room performance.

[0056] For example, as an implementation method, the performance impact data of the second computer room can be input into the artificial intelligence model to obtain the computer room performance evaluation result. The artificial intelligence model is trained according to the performance evaluation result of the second computer room and the actual result of the computer room performance until the predetermined end condition is met. The artificial intelligence model obtained when the predetermined end condition is met is determined as the computer room performance evaluation model. The artificial intelligence model can be configured according to actual business needs, which is not limited here, and it is sufficient to build a reliable computer room performance evaluation model. For example, optionally, the artificial intelligence model may include a feedforward neural network model. The feedforward neural network model may include a BP (Back Propagation) neural network model. The predetermined end condition may characterize the conditions that need to be met for the model training to be completed. For example, the predetermined end condition may include at least one of the following: the training round reaches the maximum training round and other conditions. Other conditions may include the convergence of the loss function value.

[0057] In order to ensure the reliability of the method of obtaining the computer room cost dimension data, a combination of deep learning-based image processing strategies and predetermined audit rules can be used to achieve this. The predetermined audit rules can be used to audit the authenticity of the computer room cost image and the accuracy of the computer room cost dimension data, so that the computer room cost dimension data and computer room cost image participating in the computer room efficiency evaluation model training are audited by the predetermined audit rules. The predetermined audit rules can be constructed based on expert experience.

[0058] It should be noted that the accuracy of the computer room cost dimension data can be guaranteed while ensuring the reliability of the method for obtaining the computer room cost dimension data. It should also be noted that the use of deep learning-based image processing strategies and predetermined audit rules to ensure the reliability of the computer room cost dimension data is aimed at the model training part of the above-mentioned computer room performance evaluation model. That is, the computer room cost dimension data involved in this part can refer to the second computer room cost dimension data.

[0059] How to process the computer room expense image based on the deep learning image processing strategy to obtain the computer room expense related data can be achieved in the following manner. The computer room expense related data can include the second computer room expense dimension data.

[0060] The image processing strategy based on deep learning may include a target detection strategy based on deep learning and a text recognition strategy based on deep learning. The target detection strategy based on deep learning may be used to determine the computer room expense image area from the computer room expense image. The computer room expense image area represents the image area in the computer room expense image that is related to the computer room expense related data. The text recognition strategy based on deep learning may be used to identify the computer room expense related data from the computer room expense image area. In the case of obtaining the computer room expense image area, the text recognition may be performed on the sample expense image area based on the text recognition strategy of deep learning to obtain the computer room expense related data. This is in consideration of the fact that the computer room expense related data accounts for a relatively small proportion in the computer room expense image. Therefore, if text recognition is directly performed on the computer room expense image, part of the text data may be lost due to the scaling of the image, resulting in a decrease in accuracy. Thus, target detection may be performed on the computer room expense image before text recognition is performed.

[0061] Figure 1 The figure schematically shows an example of processing a computer room expense image using a deep learning-based image processing strategy to obtain computer room expense-related data.

[0062] like Figure 1As shown, the target detection strategy based on deep learning determines the computer room expense image area from the computer room expense image. The computer room expense image area represents the image area related to the "amount 3780". Then, the text recognition strategy based on deep learning identifies the computer room expense related data from the computer room expense image area, thereby obtaining the computer room expense related data "3780".

[0063] Figure 2 The schematic diagram schematically shows the principle of resource allocation based on computer room performance evaluation according to an embodiment of the present application.

[0064] like Figure 2 As shown, it is divided into two parts, namely the model training part of the computer room efficiency evaluation model and the model application part of the computer room efficiency evaluation model (ie, the resource allocation part of the computer room is realized based on the computer room efficiency evaluation model).

[0065] 1) Model training for the computer room performance evaluation model

[0066] Figure 1 The schematic diagram schematically shows the principle of resource allocation based on computer room performance evaluation according to an embodiment of the present application.

[0067] The authenticity of the computer room expense image corresponding to the second computer room can be audited based on predetermined audit rules to obtain first audit data. When it is determined that the authenticity audit of the computer room expense image has passed according to the first audit data, the computer room expense image is processed based on the image processing strategy of deep learning to obtain computer room expense related data corresponding to the second computer room. The computer room expense related data may include second computer room expense dimension data and computer room expense auxiliary dimension data. The accuracy of the second computer room expense dimension data is audited based on predetermined audit rules and computer room expense auxiliary dimension data to obtain second audit data. When it is determined that the accuracy audit of the second computer room expense dimension data has passed according to the second audit data, the second computer room expense dimension data is used as the dimension data of the second computer room performance impact data corresponding to the second computer room to participate in the training of the computer room performance model. In addition, the second computer room performance impact data also includes second computer room impact dimension data and second computer room revenue dimension data.

[0068] When the performance impact data of the second computer room is obtained, the performance impact data of the second computer room is input into the artificial intelligence model to obtain the performance evaluation result of the second computer room. The artificial intelligence model is trained according to the performance evaluation result of the second computer room corresponding to the second computer room and the actual performance result of the computer room until a predetermined end condition is met. The artificial intelligence model obtained when the predetermined end condition is met is determined as the computer room performance evaluation model in the model application stage. Optionally, the actual performance result of the computer room corresponding to the second computer room can be determined based on the cost dimension data of the second computer room and the income dimension data of the second computer room.

[0069] 2) Model application part for computer room performance evaluation model

[0070] A plurality of first computer room performance impact data corresponding to the first computer room can be obtained. For the first computer room performance impact data among the plurality of first computer room performance impact data, the first computer room performance impact data is input into the computer room performance evaluation model to obtain the first computer room performance evaluation result corresponding to the first computer room performance impact data. Thus, the first computer room performance evaluation result corresponding to each of the plurality of first computer room performance impact data can be obtained.

[0071] When first computer room performance evaluation results corresponding to multiple first computer room performance impact data are obtained, target computer room performance impact data can be determined from the multiple first computer room performance impact data according to the first computer room performance evaluation results corresponding to the multiple first computer room performance impact data.

[0072] When the target computer room performance impact data is determined, resources may be configured for the first computer room according to the first computer room impact dimension data included in the target computer room performance impact data.

[0073] It should be noted that, for the second computer room performance impact data involved in the model training part and the first computer room performance impact data involved in the model application part, the methods for obtaining the second computer room performance impact data and the first computer room performance impact data can be similar or different.

[0074] For example, with respect to the second computer room cost dimension data included in the second computer room performance impact data, and the first computer room cost dimension data included in the first computer room performance impact data,

[0075] In the case that the first computer room is a computer room to be built, the first computer room cost dimension data and the second computer room cost dimension data are obtained in different ways. For example, the first computer room cost dimension data may be generated according to a predetermined cost generation strategy.

[0076] In the case where the first computer room is an existing computer room, there are first computer room cost dimension data in the plurality of first computer room cost dimension data that are obtained in a different manner from the second computer room cost dimension data. In addition, there may also be first computer room cost dimension data in the plurality of first computer room cost dimension data that are obtained in a similar manner to the second computer room cost dimension data, that is, the computer room cost image corresponding to the first computer room cost dimension data is processed by the image processing strategy based on deep learning to obtain the first computer room cost dimension data.

[0077] For the second computer room revenue dimension data included in the second computer room performance impact data, and the first computer room cost dimension data included in the first computer room performance impact data,

[0078] Among the multiple first computer room revenue dimension data, there may be first computer room revenue dimension data obtained in a similar manner to the second computer room revenue dimension data, and among the multiple first computer room revenue dimension data, there may also be first computer room revenue dimension data obtained in a different manner from the second computer room revenue dimension data.

[0079] For the second computer room impact dimension data included in the second computer room performance impact data, and the second computer room impact dimension data included in the second computer room performance impact data,

[0080] The plurality of first machine room impact dimension data may include first machine room impact dimension data obtained in a similar manner to the second machine room impact dimension data, or may include first machine room impact dimension data obtained in a different manner from the second machine room impact dimension data.

[0081] It should also be noted that the second computer room performance impact data may be computer room performance impact data that matches the actual situation of the second computer room, that is, the second computer room performance impact data may be the actual computer room performance impact data of the second computer room. Since the first computer room may be an already built computer room or a computer room to be built, if the first computer room is a computer room to be built, then the multiple first computer room performance impact data are not the actual computer room performance impact data of the first computer room. If the first computer room is an already built computer room, then there is at least one first computer room performance impact data among the multiple first computer room performance impact data that is not the actual computer room performance impact data of the first computer room. In addition, the actual computer room performance impact data of the first computer room may also exist among the multiple first computer room performance impact data.

[0082] In order to facilitate understanding of the technical solution of the present application, the embodiments of the present application will be combined with Figure 3 The training method of the computer room performance evaluation model, the resource configuration method and the exemplary system architecture of the device for the computer room to which the present application example can be applied are described. It should be noted that: Figure 3 The figure is only an example of a system architecture to which the embodiments of the present application can be applied, to help those skilled in the art understand the technical content of the present application, and the positional relationship between the devices, components, modules, etc. shown in the figure does not constitute any limitation. Figure 3 In the embodiment, the data storage system 350 is an external memory relative to the execution device 310. In other cases, the data storage system 350 can also be placed in the execution device 210.

[0083] Figure 3 The exemplary system architecture of the training method of the computer room performance evaluation model, the resource configuration method and the device for the computer room according to the embodiments of the present application is schematically shown.

[0084] First, how to Figure 3An exemplary architecture is shown to implement a training method for a computer room effectiveness assessment model.

[0085] like Figure 3 As shown, the data collection device 360 ​​can be used to collect the computer room cost image corresponding to the second computer room, the second computer room impact dimension data and the business data corresponding to at least one area.

[0086] After collecting the computer room expense image and the business data corresponding to at least one area, the data collection device 360 ​​can store the computer room expense image and the business data corresponding to at least one area in the database 330. The database 330 can also store predetermined audit rules, at least one first predetermined association relationship, and at least one second predetermined association relationship. The first predetermined association relationship can represent the association relationship between the area and the base station. The second predetermined association relationship can represent the association relationship between the base station and the computer room. The training device 320 can perform model training based on the data maintained in the database 330 to obtain the computer room performance evaluation model 301. The computer room performance evaluation model 301 can realize the computer room performance evaluation function provided by the present application.

[0087] The training device 320 can obtain second computer room cost dimension data in the second computer room performance impact data corresponding to the second computer room participating in the training of the computer room efficiency evaluation model based on the computer room cost image obtained from the database 330 and the predetermined audit rules, and obtain second computer room revenue dimension data in the second computer room performance impact data participating in the training of the computer room efficiency evaluation model based on the business data corresponding to at least one area respectively obtained from the database 330, at least one first preset association relationship and at least one second predetermined association relationship.

[0088] For the second computer room cost dimension data, the training device 320 can use Figure 2 It can be obtained in a similar way, which will not be described here.

[0089] For the revenue dimension data of the second computer room, the training device 320 may determine the base station corresponding to the second computer room from at least one second predetermined association relationship. When the base station corresponding to the second computer room is determined, the area corresponding to the second computer room may be determined from at least one first predetermined association relationship according to the base station corresponding to the second computer room. When the area corresponding to the second computer room is determined, the business data of the area corresponding to the second computer room may be determined as the revenue dimension data of the second computer room.

[0090] For the model training process, the training device 320 can use Figure 2 This can be achieved in a similar way and will not be described in detail here.

[0091] It should be noted that the data maintained in the database 330 may not all come from the data collected by the data collection device 360, but may also be received from other devices. It should also be noted that the training device 320 may not train the computer room performance evaluation model 301 based entirely on the data maintained by the database 330, but may also obtain training samples from the cloud or other devices for model training. The above description should not be used as a limitation on the embodiments of the present application.

[0092] The computer room performance evaluation model 301 obtained by training the training device 320 can be applied to different systems or devices, for example, Figure 2 In the execution device 310 in . The execution device 310 can be a terminal. For example, the terminal can include at least one of the following: a mobile phone, a tablet computer, a laptop computer, augmented reality (AR), virtual reality (VR), mixed reality (MR), extended reality (XR), a wearable device, an intelligent robot, and a vehicle-mounted terminal. In addition, the execution device 310 can also be a server. The server can be a server that provides various services. For example, the server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, and solves the defects of difficult management and weak business scalability in traditional physical hosts and VPS (Virtual Private Server) services.

[0093] Next, how to Figure 3 The exemplary architecture shown is used to implement a resource configuration method for a computer room.

[0094] exist Figure 3 In the embodiment, the execution device 310 configures an (Input / Output) I / O interface 312 for exchanging data with an external device. The user can input data to the I / O interface 312 through the client device 340, and the data in the embodiment of the present application may include a plurality of first computer room performance impact data corresponding to the first computer room. Optionally, the data in the embodiment of the present application may also include a plurality of initial first computer room performance impact data corresponding to the first computer room. It should be noted that the first computer room performance impact data may be input in any manner.

[0095] The preprocessing module 313 and the preprocessing module 314 can be used to preprocess the input data received by the I / O interface 312. For example, in the embodiment of the present application, the preprocessing module 313 processes the initial first computer room performance impact data to obtain the first computer room performance impact data.

[0096] The calculation module 311 is used to perform resource allocation processing operations based on the computer room performance evaluation model on the data input from the preprocessing module 313 and the preprocessing module 314 according to the computer room performance evaluation model 301 .

[0097] For the model application process, the calculation module 311 can use Figure 2 This can be achieved in a similar way and will not be described in detail here.

[0098] When the execution device 310 preprocesses the input data, or when the computing module 311 of the execution device 310 performs resource configuration processing operations based on the computer room performance evaluation model, the execution device 310 can call the data, code, etc. of the database storage system 350 for corresponding processing, and can also store the data, instructions, etc. obtained by the corresponding processing into the data storage system 350.

[0099] Finally, the I / O interface 312 returns the resource configuration result to the client device 340 and provides it to the user.

[0100] exist Figure 3 In the case shown in , the user can manually specify the data to be input into the execution device 310. For example, the first computer room performance impact data can be manually input in the interface provided by the I / O interface 312. In another case, the client device 340 can automatically input data into the I / O interface 312 and obtain the results. If the automatic data input of the client device 340 requires the user's authorization, the user can set the corresponding permissions in the client device 340. The user can view the results output by the execution device 310 on the client device 340, and the specific presentation form can be display, sound, action, and other specific methods. The client device 340 can also be used as a data acquisition terminal to collect data such as Figure 3 The input data of the input I / O interface 312 and the second computer room performance evaluation result output by the computer room performance evaluation model are used as new training samples and stored in the database 330 .

[0101] It should be noted that the execution subject of the training method of the computer room efficiency evaluation model and the execution subject of the resource configuration method for the computer room can be the same or different. For example, for different execution subjects, please refer to the above Figure 3 , the execution subject of the training method of the computer room effectiveness evaluation model is the training device 320. The execution subject of the resource configuration method for the computer room is the execution device 310. In the case where the execution subjects are the same, the execution subject of the training method of the computer room effectiveness evaluation model and the execution subject of the resource configuration method for the computer room can be, for example, Figure 3 The training device 320, or both Figure 3The execution device 330 in the embodiment. In addition, the execution subject may also be other devices, which are not limited here. Accordingly, in the case where the training device 320 executes the training method of the computer room efficiency evaluation model, the training device of the computer room efficiency evaluation model may be arranged in the training device 320. In the case where the training method of the computer room efficiency evaluation model is executed by other devices, the training device of the computer room efficiency evaluation model may be arranged in other devices. In the case where the resource configuration method for the computer room is executed by the execution device 310, the resource configuration device for the computer room may be arranged in the execution device 310. In the case where the resource configuration method is executed by other devices, the resource configuration device for the computer room may be arranged in other devices.

[0102] The following will further explain the technical solution of the model training part in conjunction with the accompanying drawings. Prior to this, the process of model training is first explained.

[0103] The process of model training may include data collection, data processing, data partitioning, hyperparameter configuration, model training, model evaluation, model tuning, and model testing and deployment. It should be noted that the embodiments of the present application mainly involve parts other than model testing and deployment.

[0104] 1) For data collection

[0105] For example, an initial computer room cost image corresponding to the second computer room, initial second computer room impact dimension data, and initial business data corresponding to at least one area may be collected. The initial second computer room impact dimension data may include at least one of the following: initial second environment impact dimension data and initial second equipment impact dimension data. The initial second environment impact dimension data may include at least one of the following: initial second computer room location, initial second computer room ambient temperature, initial second computer room associated base station number, initial second computer room associated area number, and initial second computer room associated object number. The initial second equipment impact dimension data may include at least one of the following: initial second equipment attribute data and initial second equipment health. The initial second equipment attribute data may include at least one of the following: initial second equipment model, initial second equipment performance, and initial second equipment manufacturer.

[0106] 2) For data processing

[0107] The training samples based on the participation model training include the second computer room performance impact data and the computer room performance real results. The second computer room performance impact data may include the second computer room cost dimension data, the second computer room revenue dimension data, and the second computer room impact dimension data. The computer room performance real results may be determined based on the second computer room revenue dimension data and the second computer room cost dimension data. Therefore, the following will explain how to obtain the second computer room cost dimension data, the second computer room revenue dimension data, and the second computer room impact dimension data.

[0108] 2.1) Data preprocessing

[0109] For the initial second computer room impact dimension data, the initial second computer room impact dimension data is processed based on a preprocessing method to obtain the second computer room impact dimension data. The preprocessing method may include at least one of the following: a missing value processing method, a data filtering method, and a data standardization processing method. The missing value processing method may include at least one of the following: a data deletion method and a data filling method. The data filling method may include at least one of the following: a statistical value filling method and a K-nearest neighbor filling method. In the case where the initial second computer room impact dimension data is continuous data, the statistical value may be at least one of the average value and the median of the initial second computer room impact dimension data. In the case where the initial second computer room impact dimension data is discrete data, the statistical value may be the mode of the initial second computer room impact dimension data. The K-nearest neighbor filling method may be to determine a neighbor feature whose similarity with the feature of the missing value (i.e., the missing feature) is greater than or equal to a predetermined similarity threshold, and then fill the missing feature by weighting the neighbor feature according to the similarity between the neighbor feature and the missing feature.

[0110] 2.2) Data processing

[0111] 2.2.1) Data on the cost dimension of the second computer room

[0112] First, how to obtain the second computer room cost dimension data is explained in general.

[0113] The computer room expense image can be processed based on the image processing strategy of deep learning to obtain computer room expense related data. The computer room expense image can include at least one of the following: a computer room electricity expense image, a computer room maintenance expense image, and a computer room rental image. The computer room expense related data can include second computer room expense dimension data and computer room expense auxiliary dimension data. The second computer room expense dimension data can include at least one of the following: a second computer room electricity expense, a second computer room maintenance expense, and a second computer room rental. Accordingly, the computer room expense auxiliary dimension data can include at least one of the following: computer room electricity expense auxiliary dimension data, computer room maintenance expense auxiliary dimension data, and computer room rental auxiliary dimension data.

[0114] The authenticity of the computer room expense image and the accuracy of the computer room expense related data may be audited through predetermined audit rules.

[0115] Next, it is explained that the authenticity of the computer room expense image and the accuracy of the computer room expense related data have been audited and passed according to the predetermined audit rules.

[0116] For the computer room electricity bill image, the second computer room electricity bill, and the computer room electricity bill auxiliary dimension data

[0117] The auxiliary dimension data of the computer room electricity bill may include at least one of the following: payment slip code, payment period start, payment application date, electricity purchase method, reimbursement type, tax bearing method, actual electricity bill rate, actual electricity bill amount excluding tax, actual electricity bill tax, actual total power consumption, system calculated electricity bill amount excluding tax, system calculated reimbursement amount, actual price, actual tax, actual reimbursement amount, payment period end, payment days, whether the site is held on behalf of others, bill type, other fee tax rate, other fees (excluding tax), other fee taxes, other installations (non-deductible), system calculated electricity fee taxes and computer room electricity meter numbers, etc.

[0118] When the predetermined audit rules include predetermined electricity charge audit rules, the predetermined electricity charge audit rules include predetermined electricity charge image audit rules and predetermined electricity charge data audit rules, the predetermined electricity charge image audit rules include at least one of predetermined electricity charge image editing rules and predetermined electricity charge image repetition rules, and the predetermined electricity charge data audit rules include at least one of predetermined electricity charge payment cycle accuracy rules, predetermined power consumption accuracy rules and predetermined electricity charge accuracy rules, the authenticity of the computer room electricity charge image, the accuracy of the second computer room electricity charge and the accuracy of the computer room charge auxiliary dimension data can be determined in the following manner.

[0119] As an implementation method, the attribute data of the power bill image of the computer room can be determined based on the auxiliary dimension data of the power bill of the computer room, and when it is determined that the attribute data matches the actual attribute data, it is determined that the authenticity of the power bill image of the computer room meets the predetermined power bill image editing rules. The attribute data can be used to evaluate whether the power bill image of the computer room has been edited.

[0120] As another implementation, it can be determined whether there is an image matching the computer room electricity bill image in the historical computer room electricity bill image. In the case where it is determined that there is no image matching the computer room electricity bill image in the historical computer room electricity bill image, it can be determined that the authenticity of the computer room electricity bill image meets the predetermined electricity bill image repetition rule. In the case where it is determined that there is an image matching the computer room electricity bill image in the historical computer room electricity bill image, it can be determined that the authenticity of the computer room electricity bill image does not meet the predetermined electricity bill image repetition rule, in which case, a failure information can be generated. The failure information may include that the authenticity of the computer room electricity bill image does not meet the predetermined electricity bill image repetition rule.

[0121] As another implementation, the computer room electricity fee payment cycle data corresponding to the second computer room electricity fee can be determined based on the computer room electricity fee auxiliary dimension data, and when it is determined that the computer room electricity fee payment cycle data matches the actual electricity fee payment cycle data, it is determined that the second computer room electricity fee meets the predetermined electricity fee payment cycle accuracy rule. The computer room electricity fee payment cycle data may include at least one of the following: payment application date, payment cycle, payment period start and payment period end.

[0122] As another implementation method, the power consumption of the computer room corresponding to the electricity charge of the second computer room can be determined based on the auxiliary dimension data of the electricity charge of the computer room, and when it is determined that the power consumption of the computer room is less than or equal to the predetermined power consumption threshold, it is determined that the electricity charge of the second computer room meets the predetermined power consumption accuracy rule.

[0123] The power consumption of the computer room may represent the average daily power consumption in the payment period of the computer room power bill corresponding to the second computer room power bill. The average daily power consumption may be determined based on the actual total power consumption and the payment period. Optionally, the average daily power consumption may be obtained from a database query.

[0124] The predetermined power consumption threshold may be determined based on the power consumption of the reimbursement point corresponding to the second computer room. For example, the predetermined power consumption threshold may be obtained based on the reimbursement point benchmark value and a predetermined coefficient. The reimbursement point benchmark value may be determined based on at least one computer room electricity fee benchmark value corresponding to the target reimbursement point. The computer room electricity fee benchmark value may be determined based on the site load current. The site load current may be determined based on user input data. The target reimbursement point may refer to the reimbursement point corresponding to the second computer room. The predetermined coefficient may be configured based on actual business needs and is not limited here. As an implementation method, the predetermined coefficient may be 0.1. The predetermined power consumption threshold may be obtained by multiplying the reimbursement point benchmark value and the predetermined coefficient. The reimbursement point benchmark value may be obtained by adding at least one computer room electricity fee benchmark value corresponding to the target reimbursement point. The computer room electricity fee benchmark value may be obtained by multiplying the site load current, 54V, 24, and 1.8 and dividing by 1000.

[0125] As another implementation, the computer room electricity display number and the computer room electricity meter record data can be determined based on the computer room electricity fee auxiliary dimension data, and when it is determined that the computer room electricity display number matches the computer room electricity meter record data, it is determined that the second computer room electricity fee meets the predetermined electricity fee accuracy rule. The computer room electricity display number can be obtained based on the computer room electricity meter image. For example, the computer room electricity display number can be obtained by performing text recognition on the computer room electricity meter image. The computer room electricity meter record data can be the actual total power consumption.

[0126] For the computer room maintenance fee image, the second computer room maintenance fee, and the computer room maintenance fee auxiliary dimension data

[0127] The maintenance fee of the second computer room may include at least one of the following: non-line inspection and audit fees and network maintenance fees. The auxiliary dimension data of the computer room maintenance fee may include at least one of the following: work order serial number, work order subject, urgency, professional type, professional subdivision, acceptance time limit, completion time limit, order creation time, dispatcher, dispatched team, work order status, resource identity, resource name, city, district, county, business identification, billing type, unit price, dispatch workload, reply workload, reply review workload, filing time, billing status and second computer room location, etc.

[0128] When the predetermined audit rules include predetermined maintenance fee audit rules, the predetermined maintenance fee audit rules include predetermined maintenance fee image audit rules, and the predetermined maintenance fee image audit rules include at least one of predetermined maintenance fee image repeatability rules and predetermined maintenance fee image consistency rules, the authenticity of the computer room maintenance fee image can be determined in the following manner.

[0129] As an implementation manner, when it is determined that there is no image matching the computer room maintenance fee image in the historical computer room maintenance fee images, it can be determined that the authenticity of the computer room maintenance fee image meets the predetermined maintenance fee image repetition rule.

[0130] As another implementation method, when it is determined that there are no images in the historical computer room maintenance fee image that match the start power generation image, the end power generation image, the alarm record image and the dispatch record image respectively, it can be determined that the start power generation image, the end power generation image, the alarm record image and the dispatch record image satisfy the predetermined maintenance fee image repetition rule.

[0131] As another implementation, it is possible to determine that the computer room maintenance fee image satisfies the predetermined maintenance fee image consistency rule when it is determined that the second computer room location matches the actual computer room location. The computer room maintenance fee auxiliary dimension data may include the second computer room location. The second computer room location may include the longitude of the second computer room and the latitude of the second computer room. The actual computer room location may include the actual longitude of the second computer room and the actual latitude of the second computer room. The actual computer room location may be obtained from a database.

[0132] For room rental images, second room rental, and room rental auxiliary dimension data

[0133] The auxiliary dimension data of computer room rental may include at least one of the following: reimbursement point code, reimbursement point name, associated computer room name, associated computer room, associated computer room type, ownership of associated computer room, average monthly rent for a single station, average annual rent for a single station, contract code, contract name, contract type, contract start time, contract end time, rental payment cycle, whether the rent is tax-inclusive, rent price, rent tax, contract / fixed total amount, annual rent, rental area, house type, property address, scenario division, supplier code, supplier name, supplier location (service category) and computer room equipment association data, etc.

[0134] When the predetermined audit rules include predetermined rental fee audit rules, the predetermined rental fee audit rules include predetermined rental fee data audit rules, and the predetermined rental fee data audit rules include at least one of predetermined equipment association rules, predetermined electricity fee association rules and predetermined rental fee accuracy rules, the accuracy of the second computer room rental fee can be determined in the following manner.

[0135] As an implementation method, the computer room equipment association data can be determined based on the computer room rental auxiliary dimension data, and when it is determined that there is equipment associated with the second computer room based on the computer room equipment association data, it is determined that the second computer room rental meets the predetermined equipment association rule. The computer room equipment association data can characterize the equipment associated with the second computer room. If the second computer room is associated with equipment, the computer room equipment association data can include the equipment associated with the second computer room. If the second computer room is not associated with equipment, the computer room equipment association data can be empty.

[0136] As another implementation, the computer room electricity charge correlation data can be determined based on the computer room rental auxiliary dimension data, and when it is determined that there is a second computer room electricity charge associated with the second computer room rental based on the computer room electricity charge correlation data, it is determined that the second computer room rental satisfies the predetermined electricity charge correlation rule. The computer room electricity charge correlation data can characterize the second computer room electricity charge associated with the second computer room. For example, the computer room electricity charge correlation data can be determined based on the associated computer room name included in the computer room rental auxiliary dimension data. If the second computer room is associated with the second computer room electricity charge, the computer room electricity charge correlation data can include the second computer room electricity charge associated with the second computer room. If the second computer room is not associated with the second computer room electricity charge, the computer room electricity charge correlation data can be empty.

[0137] As another implementation, the electricity bill reimbursement point association data may be determined according to the reimbursement point code included in the auxiliary dimension data of the room rental fee. When it is determined according to the electricity bill reimbursement point association data that there is an electricity bill reimbursement point associated with the rental fee reimbursement point of the second room, it is determined that the rental fee of the second room meets the predetermined electricity fee association rule.

[0138] As another implementation, the actual room rental fee may be determined based on the room rental fee auxiliary dimension data, and when it is determined that the second room rental fee matches the actual room rental fee, it is determined that the second room rental fee meets the predetermined rental fee accuracy rule. For example, the actual room rental fee may be determined based on the annual rent of the room rental fee included in the contract data in the room rental fee auxiliary dimension data.

[0139] Since the authenticity of the computer room cost image and the accuracy of the computer room cost-related data are audited and approved by the predetermined audit rules, the accuracy of the second computer room cost dimension data is guaranteed.

[0140] Next, it explains how to process the computer room cost image based on deep learning image processing strategy to obtain computer room cost related data.

[0141] The following will be combined Figure 4 Let's start with an overall explanation. Figure 4 The schematic diagram shows the principle of processing a computer room expense image according to a deep learning-based image processing strategy according to an embodiment of the present application to obtain computer room expense-related data.

[0142] like Figure 4 As shown, the deep learning-based image processing strategy may include (1) a deep learning-based object detection strategy and (2) a deep learning-based text recognition strategy.

[0143] Thus, the backbone may include an object detection branch and a text recognition branch.

[0144] For the target detection branch, target detection is performed on the computer room fee image based on the deep learning target detection strategy to obtain the computer room fee image area.

[0145] For the text recognition branch, a deep learning-based text recognition strategy is used to perform text recognition on the computer room fee image area to obtain data related to the computer room fee.

[0146] For the target detection branch, (1) the deep learning-based target detection strategy may include one of the following: (1.1) a two-stage based target detection strategy, i.e., the first target detection branch, and (1.2) a single-stage based target detection strategy, i.e., the second target detection branch.

[0147] For the text recognition branch, (2) the deep learning-based text recognition strategy may include one of the following: (2.1) a divide-and-conquer based text recognition strategy, i.e., the first text recognition branch, and (2.2) an end-to-end based text recognition strategy, i.e., the second text recognition branch.

[0148] For the first text recognition branch, (2.1) the divide-and-conquer based text recognition strategy may include (2.1.1) a text detection based strategy, ie, a text detection sub-branch, and (2.1.2) a text recognition based strategy, ie, a text recognition sub-branch.

[0149] For the text detection sub-branch, (2.1.1) the text detection strategy may include (2.1.1.1) a regression-based text detection strategy, i.e., the first text detection sub-branch, and (2.1.1.2) a segmentation-based text detection strategy, i.e., the second text detection sub-branch.

[0150] For the text recognition sub-branch, (2.1.2) the text recognition strategy may include (2.1.2.1) a text recognition strategy based on connection temporal classification, i.e., the first text recognition sub-branch, and (2.1.2.2) a text recognition strategy based on the attention mechanism, i.e., the second text recognition sub-branch.

[0151] (1) Target detection branch

[0152] The target detection strategy based on deep learning is used to perform target detection on the computer room fee image to obtain the computer room fee image area.

[0153] (1.1) For the first target detection branch, that is, a two-stage target detection strategy

[0154] As an implementation method, a fourth feature extraction is performed on the computer room expense image to obtain a third computer room expense feature map. According to at least one candidate area in the computer room expense image, a fourth computer room expense feature map corresponding to each of the at least one candidate area is determined from the third computer room expense feature map. Optionally, at least one candidate area in the computer room expense image can be obtained according to the third computer room expense feature map. For example, the third computer room expense feature map can be subjected to regional candidate processing to obtain at least one candidate area in the computer room expense image. On the basis of obtaining at least one candidate area, a fourth expense feature map corresponding to each of the at least one candidate area can be determined from the third computer room expense feature map based on a pooling strategy of an area of ​​interest. On this basis, a target detection result corresponding to each of the at least one candidate area can be obtained according to the fourth computer room expense feature map corresponding to each of the at least one candidate area. The target detection result may include an object location and an object category. Finally, according to the target detection result corresponding to each of the at least one candidate area, a computer room expense image area is determined from at least one candidate area.

[0155] As an implementation method, at least one second candidate area is determined from the computer room expense image. For example, based on the area selection strategy, at least one second candidate area can be determined from the computer room expense image. Then, the seventh feature is extracted for the at least one second candidate area respectively to obtain the seventh computer room expense feature map corresponding to the at least one second candidate area. Based on the seventh computer room expense feature map corresponding to the at least one second candidate area, the target detection result corresponding to the at least one second candidate area is obtained. Based on the target detection result corresponding to the at least one second candidate area, the computer room expense image area is determined from the at least one second candidate area.

[0156] (1.2) For the second target detection branch, that is, the single-stage target detection strategy

[0157] (1.2.1) As an implementation method, a fifth feature is extracted from the computer room fee image to obtain a fifth computer room fee feature map corresponding to at least one second scale. According to the fifth computer room fee feature map corresponding to at least one second scale, a computer room fee image region is obtained.

[0158] (1.2.2) As another implementation, the sixth feature extraction is performed on the computer room expense image to obtain the sixth computer room expense feature map. For example, the computer room expense image is subjected to residual processing of at least one first cascade level to obtain the sixth computer room expense feature map. Optionally, at least one first cascade level may include M. In the case of 1<m≤M, the m-1th eighth computer room expense feature map is obtained according to the m-1th eighth computer room expense feature map and the m-9th computer room expense feature map, the m-8th computer room expense feature map is obtained. The sixth computer room expense feature map is obtained according to the m-8th computer room expense feature map. The 1st eighth computer room expense feature map may be obtained according to the computer room expense image and the 1st ninth computer room expense feature map. The 1st ninth computer room expense feature map may be obtained according to the computer room expense image. The m-8th computer room expense feature map may represent the eighth computer room expense feature map corresponding to the m-1st first cascade level. The m-9th computer room expense feature map may represent the ninth computer room expense feature map corresponding to the m-1st first cascade level. M may be an integer greater than or equal to 1. m may be an integer greater than or equal to 1 and less than or equal to M.

[0159] Then, according to the sixth computer room cost feature map and the position vector corresponding to the computer room cost image, a fused computer room cost feature vector is obtained. Encode the fused computer room cost feature vector to obtain a coding vector. For example, based on the global feature extraction strategy, the fused computer room cost feature vector is processed at least one second cascade level to obtain a coding vector. The global feature extraction strategy may include a self-attention strategy. Optionally, at least one second cascade level may include N. In the case of 1<n≤N, based on the self-attention strategy, according to the n-1th first intermediate coding vector, the nth second intermediate coding vector is obtained. According to the nth second intermediate coding vector and the n-1th first intermediate coding vector, the nth third intermediate coding vector is obtained. According to the nth third intermediate coding vector, the nth fourth intermediate coding vector is obtained. According to the nth fourth intermediate coding vector and the nth third intermediate coding vector, the nth first intermediate coding vector is obtained. According to the Nth first intermediate coding vector, a coding vector is obtained. The 1st first intermediate coding vector may be obtained based on the 1st fourth intermediate coding vector and the 1st third intermediate coding vector. The 1st third intermediate coding vector may be obtained based on the 1st second intermediate coding vector and the fused computer room cost feature vector. The first second intermediate coding vector can be obtained based on the self-attention strategy and the fusion room fee feature vector. The first fourth intermediate coding vector can be obtained based on the first third intermediate coding vector. The nth first intermediate decoding vector can represent the first intermediate decoding vector corresponding to the nth second cascade level. The nth second intermediate decoding vector can represent the second intermediate decoding vector corresponding to the nth second cascade level. The nth third intermediate decoding vector can represent the third intermediate decoding vector corresponding to the nth second cascade level. The nth fourth intermediate decoding vector can represent the fourth intermediate decoding vector corresponding to the nth second cascade level. N can be an integer greater than or equal to 1. n can be an integer greater than or equal to 1 and less than or equal to N.

[0160] Then, according to the encoding vector, a decoding vector is obtained. For example, according to the encoding vector and the object query vector, a decoding vector is obtained. The encoding vector and the object query vector are processed by at least one third cascade level to obtain a decoding vector. Optionally, at least one third cascade level may include P. In the case of 1<p≤P, based on the self-attention strategy, according to the p-1th first intermediate decoding vector, a pth second intermediate decoding vector is obtained. According to the pth second intermediate decoding vector and the p-1th first intermediate decoding vector, a pth third intermediate decoding vector is obtained. Based on the first cross attention strategy, according to the pth third intermediate decoding vector and the encoding vector, a pth fourth intermediate decoding vector is obtained. According to the pth fourth intermediate decoding vector and the pth third intermediate decoding vector, a pth fifth intermediate decoding vector is obtained. According to the pth fifth intermediate decoding vector, a pth sixth intermediate decoding vector is obtained. According to the pth sixth intermediate decoding vector and the pth fifth intermediate decoding vector, a pth first intermediate decoding vector is obtained. According to the pth first intermediate decoding vector, a decoding vector is obtained. The 1st second intermediate decoding vector may be obtained according to the object query vector based on the self-attention strategy. The first third intermediate decoding vector may be obtained according to the first second intermediate decoding vector and the object query vector. The first fourth intermediate decoding vector may be obtained according to the first third intermediate decoding vector and the encoding vector based on the first cross attention strategy. The first fifth intermediate decoding vector may be obtained according to the pth fourth intermediate decoding vector and the first third intermediate decoding vector. The first sixth intermediate decoding vector may be obtained according to the pth fifth intermediate decoding vector. The first first intermediate decoding vector may be obtained according to the first sixth intermediate decoding vector and the first fifth intermediate decoding vector. The pth first intermediate decoding vector may represent the first intermediate decoding vector corresponding to the pth third cascade level. The pth second intermediate decoding vector may represent the second intermediate decoding vector corresponding to the pth third cascade level. The pth third intermediate decoding vector may represent the third intermediate decoding vector corresponding to the pth third cascade level. The pth fourth intermediate decoding vector may represent the fourth intermediate decoding vector corresponding to the pth third cascade level. The pth fifth intermediate decoding vector may represent the fifth intermediate decoding vector corresponding to the pth third cascade level. The pth sixth intermediate decoding vector may represent the sixth intermediate decoding vector corresponding to the pth third cascade level. P may be an integer greater than or equal to 1. p may be an integer greater than or equal to 1 and less than or equal to P.

[0161] Finally, according to the decoded vector, the room fee image area is obtained.

[0162] Regarding how to obtain the fifth machine room cost characteristic graph corresponding to at least one second scale in (1.2.1)

[0163] (1.2.1.1) As an implementation method, according to the computer room cost image, the tenth computer room cost feature map and the eleventh computer room cost feature map are obtained. According to the tenth computer room cost feature map, the twelfth computer room cost feature map is obtained. According to the tenth computer room cost feature map and the twelfth computer room cost feature map, the thirteenth computer room cost feature map is obtained. According to the eleventh computer room cost feature map and the thirteenth computer room cost feature map, the fourteenth computer room cost feature map is obtained. The fourteenth computer room cost feature map is processed based on the visual attention strategy to obtain the fifteenth computer room cost feature map. Then, according to the fifteenth computer room cost feature map, a fifth computer room cost feature map corresponding to at least one second scale is obtained. The visual attention strategy may include at least one of the following: a first channel attention strategy, a spatial attention strategy, a branch attention strategy, and a mixed attention strategy. The mixed attention strategy can be obtained based on the channel attention strategy and the spatial attention strategy.

[0164] (1.2.1.2) As another implementation, according to the machine room cost feature map, a sixteenth machine room cost feature map corresponding to at least one second scale is obtained. For example, the machine room cost image is subjected to residual processing of at least one fourth cascade level to obtain a sixteenth machine room cost feature map corresponding to at least one fourth cascade level. Then, according to the sixteenth machine room cost feature map corresponding to at least one fourth cascade level, a fifth machine room cost feature map corresponding to at least one second scale is obtained. For example, at least one second scale includes Q. In the case of 1<q<Q, according to the sixteenth machine room cost feature map corresponding to the qth second scale and the seventeenth machine room cost feature map, a fifth machine room cost feature map corresponding to the qth second scale is obtained. The seventeenth machine room cost feature map corresponding to the qth second scale can be obtained according to the sixteenth machine room cost feature map corresponding to the q+1th second scale. In the case of q=Q, according to the sixteenth machine room cost feature map corresponding to the Qth second scale, a fifth machine room cost feature map corresponding to the Qth second scale is obtained. Q is an integer greater than 1. q is an integer greater than or equal to 1 and less than or equal to Q.

[0165] How to process the 14th computer room cost feature map based on the visual attention strategy in (1.2.1.1) to obtain the 15th computer room cost feature map

[0166] As an implementation method, the fourteenth computer room cost feature map is subjected to a first compression process to obtain the eighteenth computer room cost feature map. The eighteenth computer room cost feature map is subjected to a first activation process to obtain the first channel attention feature map. The fourteenth computer room cost feature map and the first channel attention feature map are subjected to a first fusion process to obtain the fifteenth computer room cost feature map.

[0167] As another implementation method, the fourteenth computer room cost feature map is subjected to a first convolution process of multiple parallel levels to obtain a nineteenth computer room cost feature map corresponding to each of the multiple parallel levels. The nineteenth computer room cost feature map corresponding to each of the multiple parallel levels is subjected to a second fusion process to obtain a twentieth computer room cost feature map. The twentieth computer room cost feature map is subjected to a second compression process to obtain a twenty-first computer room cost feature map. The twenty-first computer room cost feature map is subjected to a second activation process to obtain a second channel attention feature map corresponding to each of the multiple parallel levels. The nineteenth computer room cost feature map corresponding to each of the multiple parallel levels and the second channel attention feature map are subjected to a third fusion process to obtain a twenty-second computer room cost feature map corresponding to each of the multiple parallel levels. The plurality of twenty-second computer room cost feature maps are subjected to a fourth fusion process to obtain a fifteenth computer room cost feature map.

[0168] As another implementation method, the cost characteristic map of the fourteenth computer room is subjected to maximum pooling processing to obtain the cost characteristic map of the twenty-third computer room. The cost characteristic map of the twenty-third computer room is subjected to average pooling processing to obtain the cost characteristic map of the twenty-fourth computer room. The cost characteristic map of the twenty-third computer room and the cost characteristic map of the twenty-fourth computer room are subjected to channel fusion processing to obtain the cost characteristic map of the twenty-fifth computer room. The cost characteristic map of the fifteenth computer room is obtained based on the cost characteristic map of the twenty-fifth computer room.

[0169] As another implementation method, the 14th computer room cost characteristic map is subjected to a second convolution process to obtain the 26th computer room cost characteristic map, the 27th computer room cost characteristic map, and the 28th computer room cost characteristic map. The 26th computer room cost characteristic map, the 27th computer room cost characteristic map, and the 28th computer room cost characteristic map are subjected to a deformation process to obtain the 29th computer room cost characteristic map corresponding to the 26th computer room cost characteristic map, the 30th computer room cost characteristic map corresponding to the 27th computer room cost characteristic map, and the 31st computer room cost characteristic map corresponding to the 28th computer room cost characteristic map. The 29th computer room cost characteristic map and the 30th computer room cost characteristic map are subjected to a fifth fusion process to obtain the 32nd computer room cost characteristic map. According to the 32nd computer room cost characteristic map, the 33rd computer room cost characteristic map is obtained. The 31st computer room cost characteristic map and the 33rd computer room cost characteristic map are subjected to a fifth fusion process to obtain the 34th computer room cost characteristic map. The 14th computer room cost characteristic map and the 34th computer room cost characteristic map are subjected to a sixth fusion process to obtain the 15th computer room cost characteristic map.

[0170] As another implementation method, the first channel attention processing is performed on the fourteenth computer room cost feature map to obtain the second channel attention feature map. The first spatial attention processing is performed on the fourteenth computer room cost feature map to obtain the first spatial attention feature map. The second channel attention feature map and the first spatial attention feature map are subjected to the seventh fusion processing to obtain the fused attention feature map. The fourteenth computer room cost feature map and the fused attention feature map are subjected to the eighth fusion processing to obtain the fifteenth computer room cost feature map.

[0171] As another implementation method, the second channel attention processing is performed on the fourteenth computer room cost feature map to obtain the third channel attention feature map. The fourteenth computer room cost feature map and the third channel attention feature map are subjected to the ninth fusion processing to obtain the thirty-fifth computer room cost feature map. The second spatial attention processing is performed on the thirty-fifth computer room cost feature map to obtain the second spatial attention feature map. The thirty-fifth computer room cost feature map and the second spatial attention feature map are subjected to the tenth fusion processing to obtain the fifteenth computer room cost feature map.

[0172] Regarding how to obtain the machine room fee image area according to the fifth machine room fee feature map corresponding to at least one second scale in (1.2.1)

[0173] As an implementation method, for the second scale in at least one second scale, according to the fifth machine room cost feature map corresponding to the second scale, the position and category corresponding to the second scale are obtained. For example, based on the second channel attention strategy, according to the fifth machine room cost feature map corresponding to the second scale, the position corresponding to the second scale is obtained. For example, the fifth machine room cost feature map corresponding to the second scale is subjected to a third compression process to obtain the thirty-sixth machine room cost feature map corresponding to the second scale. The thirty-sixth machine room cost feature map corresponding to the second scale is subjected to a third activation process to obtain the fourth channel attention feature map corresponding to the second scale. The fifth machine room cost feature map corresponding to the second scale and the fourth channel attention feature map are subjected to an eleventh fusion process to obtain the position corresponding to the second scale. Based on the third attention strategy, according to the fifth machine room cost feature map corresponding to the second scale, the category corresponding to the second scale is obtained. Then, according to the position and category corresponding to each of the at least one second scale, the machine room cost image area is obtained.

[0174] (2) For the text recognition branch

[0175] The text recognition strategy based on deep learning is used to perform text recognition on the computer room fee image area to obtain data related to the computer room fee.

[0176] (2.1) For the first text recognition branch, that is, the divide-and-conquer text recognition strategy

[0177] Perform text detection on the computer room fee image area to obtain a first computer room fee text detection result. The first computer room fee text detection result may include a first text position. Perform text recognition on the computer room fee image area according to the first computer room fee text detection result to obtain computer room fee related data.

[0178] How to perform text detection on the image area of ​​the computer room fee and obtain the first computer room fee text detection result

[0179] As an implementation manner, a third feature extraction is performed on the computer room fee image area to obtain a second computer room fee feature map corresponding to at least one first scale.

[0180] Then, according to the second computer room cost characteristic graphs respectively corresponding to at least one first scale, a second fused computer room cost characteristic graph is obtained.

[0181] As an implementation, at least one first scale may include R. In the case of r=R, the second machine room cost characteristic graph corresponding to the Rth first scale is determined as the thirty-seventh machine room cost characteristic graph corresponding to the Rth first scale. In the case of 1<r<R, the thirty-seventh machine room cost characteristic graph corresponding to the rth first scale is obtained according to the second machine room cost characteristic graph corresponding to the rth first scale and the thirty-seventh machine room cost characteristic graph corresponding to the r+1th first scale. For example, the thirty-seventh machine room cost characteristic graph corresponding to the r+1th first scale may be subjected to a first upsampling process to obtain a thirty-eighth machine room cost characteristic graph corresponding to the r+1th first scale. The second machine room cost characteristic graph corresponding to the rth first scale and the thirty-eighth machine room cost characteristic graph corresponding to the r+1th first scale are subjected to a twelfth fusion process to obtain the thirty-seventh machine room cost characteristic graph corresponding to the rth first scale. The twelfth fusion process may include at least one of connection and addition. The size of the second machine room cost characteristic graph corresponding to the rth first scale and the thirty-eighth machine room cost characteristic graph corresponding to the r+1th first scale may be the same. Then, according to the thirty-seventh machine room cost characteristic graph corresponding to each of the R-1 first scales, a second fused machine room cost characteristic graph is obtained. For example, in the case of 2<r<R, the thirty-seventh machine room cost characteristic graph corresponding to the rth first scale can be subjected to a third convolution process to obtain a thirty-ninth machine room cost characteristic graph corresponding to the rth first scale. The thirty-ninth machine room cost characteristic graph corresponding to the rth first scale is subjected to a second upsampling process to obtain a fortieth machine room cost characteristic graph corresponding to the rth first scale. In the case of r=2, the thirty-seventh machine room cost characteristic graph corresponding to the 2nd first scale can be subjected to a third convolution process to obtain a fortieth machine room cost characteristic graph corresponding to the 2nd first scale. The fortieth machine room cost characteristic graph corresponding to the 2nd first scale to the fortieth machine room cost characteristic graph corresponding to the R-1th first scale are subjected to a thirteenth fusion process to obtain a second fused machine room cost characteristic graph. The thirteenth fusion process may include at least one of connection and addition. R may be an integer greater than 1. r may be an integer greater than 1 and less than or equal to R.

[0182] It should be noted that the size of the 40th machine room cost characteristic graph corresponding to the rth first scale is the same as the size of the 40th machine room cost characteristic graph corresponding to the r+1th first scale. The size of the 37th machine room cost characteristic graph corresponding to the rth first scale may be T times the size of the 37th machine room cost characteristic graph corresponding to the r+1th first scale. The size of the second machine room cost characteristic graph corresponding to the rth first scale may be T times the size of the second machine room cost characteristic graph corresponding to the r+1th first scale. T may be an integer greater than 1. For example, T=2.

[0183] As another implementation, at least one first scale may include S. When s=S, the second computer room cost characteristic graph corresponding to the Sth first scale is determined as the forty-first computer room cost characteristic graph corresponding to the Sth first scale. When 1≤s<S, the forty-first computer room cost characteristic graph corresponding to the sth first scale is obtained based on the second computer room cost characteristic graph corresponding to the sth first scale and the forty-first computer room cost characteristic graph corresponding to the s+1th first scale. The second fused computer room cost characteristic graph is obtained based on the forty-first computer room cost characteristic graph corresponding to the 1st first scale. S may be an integer greater than 1. s may be an integer greater than 1 or equal to 1 and less than or equal to S.

[0184] It should be noted that the size of the 41st machine room cost characteristic graph corresponding to the sth first scale may be U times the size of the 41st machine room cost characteristic graph corresponding to the s+1th first scale. The size of the second machine room cost characteristic graph corresponding to the sth first scale may be U times the size of the second machine room cost characteristic graph corresponding to the s+1th first scale. U may be an integer greater than 1. For example, U=2.

[0185] Finally, based on the second fusion computer room fee feature map, the first computer room fee text detection result is obtained.

[0186] How to obtain the first computer room fee text detection results

[0187] As an implementation method, text detection is performed on the second fusion computer room fee feature map to obtain the first computer room fee text detection result.

[0188] As another implementation, a probability map is obtained according to the second fused machine room fee feature map. The probability map may include the probability of pixels of the machine room fee image. A binarization map is obtained according to the probability map and the predetermined threshold map. The predetermined threshold map may include a predetermined probability threshold of pixels of the machine room fee image. For example, for a pixel in the machine room fee image, the pixel value of the pixel may be set to a first predetermined pixel value when it is determined that the probability of the pixel in the probability map is greater than or equal to the predetermined probability threshold of the pixel in the predetermined threshold map. When it is determined that the probability of the pixel in the probability map is less than the predetermined probability threshold of the pixel in the predetermined threshold map, the pixel value of the pixel is set to a second predetermined pixel value. For example, the first predetermined pixel value may be 255. The second predetermined pixel value may be 0. The binarization map is then processed to obtain a first machine room fee text detection result. For example, a text contraction area is obtained from the binarization map. The text contraction area is expanded based on a cropping strategy to obtain a first machine room fee text detection result. The cropping strategy may include a Vatti cropping strategy. For example, the text contraction area is expanded based on an offset in the Vatti cropping strategy to obtain a first machine room fee text detection result.

[0189] As another implementation method, the second fused machine room fee feature map is processed using a text detection model to obtain the first machine room fee text detection result. The text detection model can be obtained by training a deep learning model using the second machine room fee image, the probability label map, and the threshold label map. For example, the second machine room fee image can be input into the backbone module of the deep learning model to obtain a 42nd machine room fee feature map corresponding to at least one fourth scale. The 42nd machine room fee feature map corresponding to at least one fourth scale is input into the fusion module of the deep learning model to obtain the 43rd machine room fee feature map. According to the 43rd machine room fee feature map, a second probability map and an adaptive threshold map are obtained. Based on a differentiable binarization function, a second binarization map is obtained according to the second probability map and the adaptive threshold map. The deep learning model is trained using the second probability map, the second binarization map, the probability label map, the adaptive threshold map, and the threshold label map to obtain a text detection model.

[0190] How to perform text recognition on the computer room fee image area based on the first computer room fee text detection result to obtain computer room fee related data

[0191] As an implementation method, a second feature extraction is performed on the computer room fee image area according to the first computer room fee text detection result to obtain a first computer room fee feature map. For example, according to the first computer room fee text detection result, the first text image area is determined from the computer room fee image area. A second feature extraction is performed on the first text image area to obtain a first computer room fee feature map. For example, a fourth convolution process can be performed on the first text image area to obtain a first computer room fee feature map.

[0192] Then, according to the first machine room cost feature graph, a machine room cost feature vector sequence is obtained. The machine room cost feature vector sequence may include multiple machine room cost feature vectors. According to the machine room cost feature vector sequence, a first machine room cost text sequence feature is obtained. The machine room cost text sequence feature may characterize the probability distribution corresponding to each of the multiple machine room cost feature vectors. For example, the machine room cost feature vector sequence is processed using a temporal network layer to obtain the first machine room cost text sequence feature. The temporal network layer may include at least one of a recurrent neural network (RNN) layer and a bidirectional recurrent neural network (Bi-directional Recurrent Neural Network, Bi-RNN). The recurrent neural network layer may include at least one of a long short-term memory network (LSTM) layer and a gated recurrent unit (GRU) layer. The bidirectional recurrent network layer may include a bidirectional long short-term memory network (Bi-directional Long Short-Term Memory, Bi-LSTM). In addition, the long short-term memory network may include at least one of a basic long short-term memory network and a deformation network for the basic long short-term memory network. The bidirectional long short-term memory network may include at least one of a basic bidirectional long short-term memory network and a deformation network for the basic bidirectional long short-term memory network. Optionally,

[0193] Finally, according to the first computer room fee text sequence feature, the computer room fee related data is obtained. For example, the computer room fee text sequence feature can be processed based on the connectionist temporal classification (CTC) strategy to obtain the computer room fee related data. Continuous temporal classification. Optionally, the computer room fee related data can be obtained based on the second cross-attention strategy processing according to the computer room fee text sequence feature and the computer room fee feature vector sequence.

[0194] (2.2) For the second text recognition branch, that is, based on the end-to-end text recognition strategy

[0195] As an implementation method, a first feature extraction is performed on the computer room expense image area to obtain a first computer room expense area feature map. For example, according to the computer room expense image area, a second computer room expense area feature map corresponding to at least one third scale is obtained. According to the second computer room expense area feature map corresponding to at least one third scale, a first computer room expense area feature map is obtained. For example, at least one third scale may include V. In the case of 1≤v<V, according to the third computer room expense area feature map corresponding to the v+1 third scale and the second computer room expense area feature map corresponding to the v third scale, a third computer room expense area feature map corresponding to the v third scale is obtained. The third computer room expense area feature map corresponding to the V third scale may be obtained based on the second computer room expense area feature map corresponding to the V third scale. Then, according to the third computer room expense area feature map corresponding to the 1 third scale, a first computer room expense area feature map is obtained. V may be an integer greater than 1. v may be an integer greater than or equal to 1 and less than or equal to V.

[0196] Then, text detection is performed on the room charge area feature map according to the vth step to obtain a second room charge text detection result. The second room charge text detection result may include a second text position. Then, according to the second room charge text detection result, a first room text candidate feature map is determined from the room charge area feature map.

[0197] Finally, text recognition is performed on the first machine room text candidate feature map to obtain machine room fee-related data. For example, the eighth feature extraction is performed on the first machine room text candidate feature map to obtain the second machine room text candidate feature map. According to the second machine room text candidate feature map, a machine room text feature vector sequence is obtained. According to the machine room text feature vector sequence, a second machine room fee text sequence feature is obtained. According to the second machine room fee text sequence feature, data related to machine room fees are obtained.

[0198] As another implementation method, the computer room fee image area is positioned and processed to obtain spatial transformation parameters. The predetermined grid matrix is ​​transformed based on the spatial transformation parameters to obtain a grid transformation matrix. The predetermined grid matrix includes a plurality of grid elements. The grid elements correspond to pixels of the second text image area. The grid transformation matrix indicates the transformation from the computer room fee image area to the second text image area. Based on the sampler, the second text image area is obtained according to the grid transformation matrix and the computer room fee image area. Text recognition is performed on the second text image area to obtain computer room fee related data.

[0199] Since the data related to computer room expenses is obtained by processing the computer room expense image based on the image processing strategy of deep learning, the accuracy of the data of the second computer room expense dimension is guaranteed.

[0200] How to obtain the revenue dimension data of the second computer room

[0201] According to the second computer room, a base station corresponding to the second computer room is determined from at least one second predetermined association relationship. The second predetermined association relationship may represent an association relationship between a base station and a computer room. For example, a target second predetermined association relationship corresponding to the second computer room is determined from at least one second predetermined association relationship. Then, according to the target second predetermined association relationship, a base station corresponding to the second computer room is determined.

[0202] Then, according to the base station corresponding to the second computer room, the area corresponding to the second computer room is determined from at least one first predetermined association relationship. The first predetermined association relationship can represent the association relationship between the area and the base station. For example, the target first predetermined association relationship corresponding to the base station corresponding to the second computer room is determined from at least one first predetermined association relationship. Then, according to the target first predetermined association relationship, the area corresponding to the second computer room is determined.

[0203] Finally, the revenue dimension data of the second computer room is determined according to the business data of the area corresponding to the second computer room. For example, the business data of the area corresponding to the second computer room can be determined as the revenue dimension data of the second computer room.

[0204] 3) Data division

[0205] The samples can be divided into training samples and validation samples. The training samples are used to train the artificial intelligence model. The validation samples are used to evaluate the model effect of the trained artificial intelligence model. The method for dividing the training samples and the validation samples can be selected according to actual business needs and is not limited here. For example, the division method may include at least one of the following: a holdout method, a cross-validation method, and a self-service sampling method.

[0206] 4) For hyperparameter configuration

[0207] The hyperparameters may include at least one of the following: learning rate, batch sample number, maximum training round, model structure, regularization parameter, etc. The regularization parameter may refer to a parameter corresponding to the regularization term in the loss function.

[0208] Regarding the model structure, the artificial intelligence model can be configured according to actual business needs. There is no limitation here. It only needs to be able to build a reliable computer room performance evaluation model.

[0209] For example, the artificial intelligence model may be constructed based on at least one model structure and the connection relationship between different model structures. The model structure may include at least one model substructure and the connection relationship between different model substructures. The model structure may be obtained by connecting at least one model substructure based on the connection relationship between different model substructures. The model substructure may be a structure from at least one operation layer. Thus, the model structure may be a structure obtained by connecting at least one model substructure from each of multiple operation layers based on the connection relationship between different model substructures. For example, at least one operation layer may include an input layer, a hidden layer, an output layer, etc.

[0210] As an implementation method, the artificial intelligence model may include a feedforward neural network model. Optionally, the feedforward neural network model may include a BP neural network model. The number of hidden layers of the BP neural network model may be configured according to actual business needs and is not limited here. For example, in order to ensure the generalization performance of the model, the number of hidden layers of the BP neural network model may be set to 2.

[0211] 5) Model training

[0212] Please refer to the following description of the training method for the computer room performance evaluation model.

[0213] 6) Model evaluation

[0214] The model effect of the computer room performance evaluation model is evaluated using the verification sample to obtain the effect evaluation result.

[0215] 7) Tuning the model

[0216] The training process is optimized according to the effect evaluation results. For example, at least one of the hyperparameters and the model structure may be adjusted according to the effect evaluation results.

[0217] 8) Model deployment

[0218] The computer room performance evaluation model that meets the expected requirements can be deployed as a model so that the computer room performance evaluation model can be applied for resource allocation.

[0219] The model training part will be further explained below with reference to the accompanying drawings.

[0220] Figure 5 A flowchart of a method for training a computer room performance evaluation model according to an embodiment of the present application is schematically shown.

[0221] like Figure 5 As shown, the method 500 may include operations S510 to S520.

[0222] In operation S510, second computer room performance impact data and computer room performance actual results corresponding to the second computer room are obtained.

[0223] It should be noted that, for the method of obtaining the second computer room performance impact data and the actual results of the computer room performance, please refer to the above description of the data processing part, which will not be repeated here.

[0224] In operation S520, the second computer room performance impact data and the actual results of the computer room performance are used to train an artificial intelligence model to obtain a computer room performance evaluation model.

[0225] It should be noted that the second computer room performance impact data may include multiple items and the computer room performance actual results may include multiple items.

[0226] Regarding how to train the computer room performance evaluation model, it can be achieved in the following ways.

[0227] As an implementation method, the artificial intelligence model is trained multiple times using the second computer room performance impact data and the actual computer room performance results until a predetermined end condition is met. The trained artificial intelligence model is determined as the computer room performance evaluation model.

[0228] Regarding how to perform multiple rounds of training on the artificial intelligence model to obtain the computer room performance evaluation model, it can be achieved in the following ways.

[0229] As an implementation method, each round has an artificial intelligence model corresponding to the round. The artificial intelligence model corresponding to the round can be obtained by training the artificial intelligence model of the previous round corresponding to the round. The round in progress can be called the current round. In the case where the current round is the first round, the model parameters of the artificial intelligence model can be initialized to obtain the artificial intelligence model corresponding to the first round. Thus, by executing Figure 6 The method shown is used to obtain the computer room effectiveness evaluation model.

[0230] For example, see Figure 6 , Figure 6 A flowchart of another method for training a computer room performance evaluation model according to an embodiment of the present application is exemplified.

[0231] like Figure 6 As shown, the method 600 may include operations S621 to S625.

[0232] In operation S621, the second computer room performance impact data is input into the artificial intelligence model corresponding to the current round to obtain the second computer room performance evaluation result corresponding to the current round.

[0233] In operation S622, according to the actual results of the computer room performance and the second computer room performance evaluation results corresponding to the current round, the artificial intelligence model corresponding to the current round is trained to obtain the artificial intelligence model corresponding to the next round.

[0234] In operation S623, does the artificial intelligence model corresponding to the next round meet the predetermined end condition? If yes, then execute operation S624; if no, then execute operation S625.

[0235] In operation S624, the trained artificial intelligence model is determined as a computer room performance evaluation model.

[0236] In operation S625, the artificial intelligence model corresponding to the next round is determined as the artificial intelligence model corresponding to the new current round, and the process returns to execute operation S621.

[0237] According to an embodiment of the present application, the predetermined end condition may characterize the conditions that need to be met to complete the model training. For example, the predetermined end condition may include at least one of the following: the training round reaches the maximum training round and other conditions. Other conditions may include the convergence of the loss function value. The convergence of the loss function value may include at least one of the following: the loss function value is less than or equal to the first predetermined threshold and the change value of the model parameter is less than or equal to the second predetermined threshold.

[0238] Regarding how to determine whether the artificial intelligence model corresponding to the next round meets the predetermined end condition, it can be achieved in the following way.

[0239] As an implementation method, it can be determined whether the current round is the maximum training round. If it is determined that the current round is the maximum training round, it can be determined that the artificial intelligence model corresponding to the next round meets the predetermined end condition.

[0240] As another implementation, it can be determined whether the loss function value corresponding to the artificial intelligence model corresponding to the next round converges. If it is determined that the loss function value converges, it can be determined that the artificial intelligence model corresponding to the next round meets the predetermined end condition.

[0241] In the case where it is determined that the artificial intelligence model corresponding to the next round meets the predetermined end condition, the artificial intelligence model that meets the predetermined end condition can be determined as the computer room performance evaluation model. In the case where it is determined that the artificial intelligence model corresponding to the next round does not meet the predetermined end condition, the artificial intelligence model corresponding to the next round can be determined as the artificial intelligence model corresponding to the new current round, that is, the artificial intelligence model corresponding to the next round becomes the artificial intelligence model corresponding to the current round, and then operations S621 to 625 are performed until the artificial intelligence model corresponding to the next round meets the predetermined end condition, thereby determining the artificial intelligence model that meets the predetermined end condition as the computer room performance evaluation model.

[0242] Regarding how to train the artificial intelligence model corresponding to the current round based on the actual results of the computer room performance and the second computer room performance evaluation results corresponding to the current round, and obtain the artificial intelligence model corresponding to the next round, it can be achieved in the following way.

[0243] As an implementation method, the real results of the computer room efficiency and the second computer room efficiency evaluation results corresponding to the current round can be input into the loss function to obtain the loss function value corresponding to the current round. According to the loss function value corresponding to the current round, the model parameters of the artificial intelligence model corresponding to the current round are adjusted to obtain the model parameters of the artificial intelligence model corresponding to the next round. The loss function can be configured according to actual business needs and is not limited here. For example, the loss function may include one of the following: mean square error function, mean absolute error function, and cross entropy loss function, etc. For example, a gradient function can be obtained according to the loss function. The real results of the computer room efficiency and the second computer room efficiency evaluation results corresponding to the current round are input into the gradient function to obtain the gradient. Based on the gradient descent method, the model parameters of the artificial intelligence model corresponding to the current round are adjusted according to the gradient to obtain the model parameters of the artificial intelligence model corresponding to the next round. For example, the gradient descent method may include at least one of the following: stochastic gradient descent method, batch gradient descent method, and mini-batch gradient descent method. In the process of adjusting the model parameters of the artificial intelligence model corresponding to the current round according to the gradient, the model parameters of the artificial intelligence model corresponding to the current round can be adjusted based on the gradient using the back propagation method.

[0244] In order to better understand the training method of the computer room performance evaluation model involved in the embodiment of the present application, a specific example is given below to illustrate.

[0245] Figure 7 An example schematic diagram schematically shows the training process of the computer room performance evaluation model according to an embodiment of the present application.

[0246] like Figure 7 As shown, the artificial intelligence model is a BP neural network model. The BP neural network model may include an input layer, a hidden layer, and an output layer. The hidden layer may include a first hidden layer and a second hidden layer. The input layer may include I neurons, namely, neurons x1, ..., neurons x i , ..., neuron x I The first hidden layer may include J neurons, namely, neurons z 11 , neuron z 12 , ..., neuron z 1j , ..., neuron z 1J The second hidden layer may include J neurons, namely, neurons z 21 , neuron z 22 , ..., neuron z 2j , ..., neuron z2J The output layer may include G neurons, namely, neurons y1, ..., neurons y g , ..., neuron y G .

[0247] Neuronx i With neuron z 1j The model weight between is w i-1j . Neuron z 1j With neuron z 2j The model weight between is w 1j-2j . Neuron z 2j With neuron y g The model weight between is w 2j-g . The activation function is Sigmoid.

[0248] I may be an integer greater than 1. i may be an integer greater than or equal to 1 and less than or equal to I. i∈{1, ..., I}. J may be an integer greater than 1. j may be an integer greater than or equal to 1 and less than or equal to J. j∈{1, ..., J}. G may be an integer greater than or equal to 1. g∈{1, ..., G}.

[0249] The training process of the computer room performance evaluation model may include a forward propagation process and a back propagation process.

[0250] For the forward propagation process, the data affecting the performance of the second computer room is input into the BP neural network model to obtain the performance evaluation result of the second computer room corresponding to the current round. For example, the performance impact data of the second computer room is transmitted to the first hidden layer via the input layer, and each neuron in the first hidden layer processes the performance impact data of the second computer room to obtain the first feature data corresponding to each neuron in the first hidden layer. Each neuron in the first hidden layer transmits its first feature data to the second hidden layer. Each neuron in the second hidden layer processes each first feature data to obtain the second feature data corresponding to each neuron in the second hidden layer. Each neuron in the second hidden layer transmits its second feature data to the output layer. Each neuron in the output layer processes each second feature data to obtain output data corresponding to each neuron in the output layer. According to each output data, the performance evaluation result of the second computer room corresponding to the current round is obtained. Specifically:

[0251] For the input layer → first hidden layer, the neuron z in the first hidden layer 1j The input is

[0252] For the first hidden layer → the second hidden layer, the neuron z in the second hidden layer 2j The input is

[0253] For the second hidden layer → output layer, the neuron y in the output layer g The input is

[0254] For the back propagation process, the actual results of the computer room efficiency and the second computer room efficiency evaluation results corresponding to the current round can be input into the gradient function to obtain the gradient. Based on the gradient, the back propagation method is used to adjust the model parameters of the artificial intelligence model corresponding to the current round to obtain the model parameters of the artificial intelligence model corresponding to the next round. When it is determined that the artificial intelligence model corresponding to the next round meets the predetermined end conditions, the artificial intelligence model that meets the predetermined end conditions can be determined as the computer room efficiency evaluation model. When it is determined that the artificial intelligence model corresponding to the next round does not meet the predetermined end conditions, the artificial intelligence model corresponding to the next round can be determined as the new artificial intelligence model corresponding to the current round, that is, the artificial intelligence model corresponding to the next round becomes the artificial intelligence model corresponding to the current round, and then the normal propagation process is entered again, and the cycle is repeated until the predetermined end conditions are met.

[0255] The model effect of the computer room performance evaluation model can be evaluated using the verification sample to obtain the effect evaluation result. The evaluation index can include the mean square error (MSE). The performance evaluation results are given in Table 1 below. The third computer room performance evaluation result can refer to the computer room performance evaluation result corresponding to the verification sample. Since the mean square error is small, the characterization computer room performance evaluation model can meet the actual requirements.

[0256]

[0257]

[0258] Table 1

[0259] Figure 8 The flowchart of the resource configuration method for a computer room according to an embodiment of the present application is schematically shown.

[0260] like Figure 8 As described, the method 800 may include operations S810 to S830.

[0261] In operation S810, a plurality of first computer room performance impact data corresponding to the first computer room is input into an computer room performance evaluation model to obtain first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data.

[0262] According to an embodiment of the present application, the first computer room performance impact data may include first computer room cost dimension data, first computer room revenue dimension data, and first computer room impact dimension data. The first computer room impact dimension data may include at least one of the following: first environment impact dimension data and first equipment impact dimension data. The first environment impact dimension data may include at least one of the following: the first computer room location, the number of base stations associated with the first computer room, the number of areas associated with the first computer room, and the number of objects associated with the first computer room. The first equipment impact dimension data may include at least one of the following: first equipment attribute data and first equipment health. The first equipment attribute data may include at least one of the following: first equipment model, first equipment performance, and first equipment producer.

[0263] According to an embodiment of the present application, for the first computer room performance impact data among the plurality of first computer room performance impact data, the first computer room performance impact data can be input into the computer room performance evaluation model to obtain the first computer room performance evaluation result corresponding to the first computer room performance impact data. Thus, the first computer room performance evaluation result corresponding to each of the plurality of first computer room performance impact data can be obtained.

[0264] In operation S820, target computer room performance impact data is determined from the plurality of first computer room performance impact data according to the first computer room performance evaluation results respectively corresponding to the plurality of first computer room performance impact data.

[0265] According to an embodiment of the present application, the target computer room performance impact data may represent the first computer room performance impact data corresponding to the target computer room performance evaluation result. The target computer room performance evaluation result may refer to the first computer room performance evaluation result with the largest median value among multiple first computer room performance evaluation results.

[0266] According to the embodiment of the present application, since the first computer room performance evaluation result corresponding to the first computer room performance impact data can reflect the impact degree of the first computer room performance impact data on the computer room performance of the first computer room, and the first computer room performance impact data includes data related to resource configuration, the optimal computer room performance impact data can be determined from the multiple first computer room performance impact data according to the first computer room performance evaluation results corresponding to each of the multiple first computer room performance impact data. The optimal computer room performance impact data can refer to

[0267] The target computer room performance impact data can reflect the computer room performance impact data required to obtain the optimal computer room performance. Therefore, in order to achieve the optimal computer room performance, resources can be configured for the first computer room according to the target computer room performance impact data.

[0268] Regarding how to determine the target computer room performance impact data, it can be achieved in the following ways.

[0269] As an implementation method, the target computer room performance evaluation result can be determined from the first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data. The first computer room performance impact data corresponding to the target computer room performance evaluation result among the plurality of first computer room performance impact data is then determined as the target computer room performance impact data. The target computer room performance evaluation result can represent the optimal performance evaluation result.

[0270] In operation S830, resources are configured for the first computer room according to the first computer room impact dimension data included in the target computer room performance impact data.

[0271] According to the embodiments of the present application, resource configuration may be implemented in the following manner.

[0272] As an implementation method, the first computer room impact dimension data may include first environment impact dimension data and first device impact dimension data. The first environment impact dimension data may include at least one of the following: the location of the first computer room, the number of base stations associated with the first computer room, the number of areas associated with the first computer room, and the number of objects associated with the first computer room. The first device impact dimension data may include first device attribute data and first device health data. The first device attribute data may include first device attribute dimension data corresponding to each of the at least one first device. The first device attribute dimension data may include at least one of the following: the first device model, the first device performance, and the first device manufacturer. The first device health data may include the first device health corresponding to each of the at least one first device.

[0273] Therefore, resource allocation can be achieved in the following ways.

[0274] As an implementation method, the recommended location of the first computer room can be determined based on the first environmental impact dimension data. For example, the recommended location of the first computer room can be determined based on the location of the first computer room in the first environmental impact dimension data, that is, the location of the first computer room can be determined as the recommended location of the first computer room. Optionally, the location area corresponding to the location of the first computer room can be determined based on the number of base stations associated with the first computer room, the number of areas associated with the first computer room, and the number of objects associated with the first computer room in the first environmental impact dimension data. The recommended location is then determined from at least one candidate location included in the location area.

[0275] As an implementation method, attribute data of at least one recommended device of the first computer room may be determined based on the first device impact dimension data. For example, for a first device among at least one first device, when it is determined that the health of the first device corresponding to the first device meets a predetermined condition, attribute data of the recommended device of the first computer room may be determined based on the first device attribute dimension data corresponding to the first device. For example, the first device attribute dimension data corresponding to the first device may be determined as the attribute data of the recommended device of the first computer room. The predetermined condition may indicate that the health of the first device corresponding to the first device is greater than or equal to a predetermined health threshold.

[0276] Since resources for the first computer room are configured based on the first computer room impact dimension data included in the target computer room performance impact data, the target computer room performance impact data are determined from multiple first computer room performance impact data based on the first computer room performance evaluation results corresponding to each of the multiple first computer room performance impact data. The target computer room performance impact data can reflect the computer room performance impact data required to obtain the optimal computer room performance, and therefore, reasonable configuration of computer room resources can be achieved.

[0277] In addition, compared with the related art that uses an experimental matching method to perform resource allocation, the resource allocation method for a computer room provided in the present application improves the efficiency of resource allocation.

[0278] Based on the same concept as the above-mentioned embodiment of a resource configuration method for a computer room, the embodiment of the present application also provides a resource configuration device 900 for a computer room, which can be deployed on any device, equipment, platform or equipment cluster with computing capabilities to implement the resource configuration method for a computer room provided in the embodiment of the present application. The resource configuration device 900 for a computer room includes units or modules for implementing various operations in the resource configuration method for a computer room.

[0279] Fig. 9 A block diagram of a resource configuration device for a computer room according to an embodiment of the present application is schematically shown.

[0280] like Fig. 9 As shown, the resource configuration device 900 for a computer room may include an acquisition mode 910 , a determination module 920 and a configuration module 930 .

[0281] The acquisition module 910 is used to input a plurality of first computer room performance impact data corresponding to the first computer room into the computer room performance evaluation model, and obtain the first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data. The first computer room performance impact data includes first computer room cost dimension data, first computer room revenue dimension data and first computer room impact dimension data. The first computer room impact dimension data represents data that affects the first computer room cost dimension data and the first computer room revenue dimension data.

[0282] The determination module 920 is used to determine the target computer room performance impact data from the plurality of first computer room performance impact data according to the first computer room performance evaluation results respectively corresponding to the plurality of first computer room performance impact data.

[0283] The configuration module 930 is used to perform resource configuration on the first computer room according to the first computer room impact dimension data included in the target computer room performance impact data. Resource configuration may include at least one of the following: computer room site selection and equipment selection.

[0284] The computer room efficiency evaluation model is obtained by training an artificial intelligence model using the second computer room efficiency impact data and the real computer room efficiency results. The second computer room efficiency impact data includes the second computer room cost dimension data, the second computer room revenue dimension data, and the second computer room impact dimension data. The second computer room impact dimension data represents the data that affects the second computer room cost dimension data and the second computer room revenue dimension data. The real computer room efficiency results represent the real computer room efficiency results of the second computer room.

[0285] The data related to the computer room expenses include the second computer room expense dimension data. The data related to the computer room expenses is obtained by processing the computer room expense image based on the image processing strategy of deep learning. The authenticity of the computer room expense image and the accuracy of the data related to the computer room expenses are audited and approved by the predetermined audit rules.

[0286] According to the embodiment of the present application, the resource configuration device 900 for the computer room can correspond to the resource configuration method for the computer room described in the embodiment of the present application, and the above-mentioned and other operations and / or functions of each module in the resource configuration device 900 for the computer room are the corresponding processes of the resource configuration method for the computer room in the embodiment of the present application, which will not be repeated here for the sake of brevity.

[0287] An embodiment of the present application also provides an electronic device, including at least one processor, a memory, and a communication interface, wherein the processor is used to execute the resource configuration method for a computer room of the embodiment of the present application.

[0288] Fig.10 A block diagram of an electronic device according to an embodiment of the present application is exemplarily shown.

[0289] like Fig.10As shown, the electronic device 1000 includes at least one processor 1001, a memory 1002 and a communication interface 1003. The processor 1001, the memory 1002 and the communication interface 1003 are connected in communication, and the communication connection can be achieved by wired means (such as a bus) or by wireless means. The communication interface 1003 is used to send and / or receive data sent by other devices; the memory 1002 stores computer instructions, and the processor 1001 executes the computer instructions to execute the resource configuration method for the computer room in the aforementioned method embodiment.

[0290] It should be understood that in the embodiment of the present application, the processor 1001 may be a central processing unit CPU, and the processor 1001 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0291] The memory 1002 may include a read-only memory and a random access memory, and provides instructions and data to the processor 1001. The memory 1002 may also include a nonvolatile random access memory.

[0292] The memory 1002 may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (Direct Rambus RAM, DR RAM).

[0293] It should be understood that the electronic device 1000 according to the embodiment of the present application can execute the resource configuration method for the computer room according to the embodiment of the present application. The detailed description of the implementation of the above method is given above, and it will not be repeated here for the sake of brevity.

[0294] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned method is implemented.

[0295] An embodiment of the present application provides a chip, which includes at least one processor and an interface, wherein the at least one processor determines program instructions or data through the interface; the at least one processor is used to execute the program instructions to implement the method mentioned above.

[0296] An embodiment of the present application provides a computer program or a computer program product, wherein the computer program or the computer program product comprises instructions, and when the instructions are executed, the computer is caused to execute the above-mentioned method.

[0297] Those of ordinary skill in the art should further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0298] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0299] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only the specific implementation method of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A resource configuration method for a computer room, comprising: Inputting a plurality of first computer room performance impact data corresponding to the first computer room into the computer room performance evaluation model, and obtaining first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data, wherein the first computer room performance impact data includes first computer room cost dimension data, first computer room revenue dimension data, and first computer room impact dimension data, and the first computer room impact dimension data represents data that affects the first computer room cost dimension data and the first computer room revenue dimension data; Determining target computer room performance impact data from the plurality of first computer room performance impact data according to the first computer room performance evaluation results respectively corresponding to the plurality of first computer room performance impact data; and Perform resource allocation on the first computer room according to the first computer room impact dimension data included in the target computer room performance impact data, wherein the resource allocation includes at least one of the following: computer room site selection and equipment selection; The computer room performance evaluation model is obtained by training an artificial intelligence model using the second computer room performance impact data and the real results of the computer room performance, the second computer room performance impact data includes the second computer room cost dimension data, the second computer room revenue dimension data and the second computer room impact dimension data, the second computer room impact dimension data represents the data that affects the second computer room cost dimension data and the second computer room revenue dimension data, and the real results of the computer room performance represent the real computer room performance data of the second computer room; Among them, the data related to computer room expenses include the second computer room expense dimension data, and the data related to computer room expenses are obtained by processing the computer room expense image based on the image processing strategy of deep learning. The authenticity of the computer room expense image and the accuracy of the data related to computer room expenses are audited and approved by predetermined audit rules.

2. The method according to claim 1, wherein: The first computer room impact dimension data includes first environment impact dimension data and first equipment impact dimension data, the first environment impact dimension data includes at least one of the following: first computer room location, the number of base stations associated with the first computer room, the number of areas associated with the first computer room, and the number of objects associated with the first computer room, the first equipment impact dimension data includes first equipment attribute data and first equipment health data, the first equipment attribute data includes first equipment attribute dimension data corresponding to at least one first equipment respectively, the first equipment attribute dimension data includes at least one of the following: first equipment model, first equipment performance, and first equipment manufacturer, the first equipment health data includes first equipment health corresponding to at least one first equipment respectively; The performing resource allocation on the first computer room according to the first computer room impact dimension data included in the target computer room performance impact data includes at least one of the following: determining a recommended location of the first computer room according to the first environmental impact dimension data; and For a first device among the at least one first device, when it is determined that the health of a first device corresponding to the first device meets a predetermined condition, attribute data of a recommended device of the first computer room is determined according to the first device attribute dimension data corresponding to the first device.

3. The method according to claim 1 or 2, wherein: The determining target computer room performance impact data from the plurality of first computer room performance impact data according to the first computer room performance evaluation results respectively corresponding to the plurality of first computer room performance impact data comprises: Determine a target computer room performance evaluation result from the first computer room performance evaluation results corresponding to each of the plurality of first computer room performance impact data; and The first computer room performance impact data corresponding to the target computer room performance evaluation result among the plurality of first computer room performance impact data is determined as the target computer room performance impact data.

4. The method according to any one of claims 1 to 3, wherein: The data related to the computer room expenses is obtained by processing the computer room expense image based on the image processing strategy of deep learning, and includes: The data related to the computer room fee is obtained by performing text recognition on the computer room fee image area based on a text recognition strategy based on deep learning; The computer room fee image area is obtained by performing target detection on the computer room fee image based on a deep learning target detection strategy.

5. The method according to claim 4, wherein: The data related to the computer room fee is obtained by performing text recognition on the computer room fee image area based on a text recognition strategy based on deep learning, and includes one of the following: The data related to the computer room fee is obtained by performing text recognition on the computer room fee image area according to the first computer room fee text detection result; and the first computer room fee text detection result is obtained by performing text detection on the computer room fee image area; The data related to the computer room fee is obtained by performing text recognition on a candidate feature map of the computer room text; the candidate feature map of the computer room text is determined from the computer room fee regional feature map according to the second computer room fee text detection result; the second computer room fee text detection result is obtained by performing text detection on the computer room fee regional feature map; And the computer room fee area feature map is obtained by performing a first feature extraction on the computer room fee image area.

6. The method according to claim 5, wherein: The data related to the computer room fee is obtained by performing text recognition on the computer room fee image area according to the first computer room fee text detection result, and includes: The data related to the computer room fee is obtained according to the text sequence characteristics of the computer room fee; The computer room fee text sequence feature is obtained according to the computer room fee feature vector sequence; The computer room cost feature vector sequence is obtained according to the first computer room cost feature graph; The first computer room fee feature map is obtained by performing a second feature extraction on the computer room fee image area according to the first computer room fee text detection result.

7. The method according to claim 5 or 6, wherein: The first machine room fee text detection result is obtained by performing text detection on the machine room fee image area, and includes: The first computer room fee text detection result is obtained based on a fused computer room fee feature map; the fused computer room fee feature map is obtained based on a second computer room fee feature map corresponding to at least one first scale; and the second computer room fee feature map corresponding to at least one first scale is obtained by performing a third feature extraction on the computer room fee image area.

8. The method according to claim 7, wherein: The first computer room fee text detection result is obtained based on the integrated computer room fee feature map, including one of the following: The first computer room fee text detection result is obtained by processing a binarized image; the binarized image is obtained based on a probability image and a predetermined threshold image; and the probability image is obtained based on the fused computer room fee feature image; The first computer room fee text detection result is obtained by performing text detection on the fused computer room fee feature graph.

9. The method according to any one of claims 5 to 8, wherein: The computer room fee image region is obtained by performing target detection on the computer room fee image based on a target detection strategy based on deep learning, and includes one of the following: The computer room fee image area is determined from the at least one candidate area according to the target detection result corresponding to each of the at least one candidate area; the target detection result corresponding to each of the at least one candidate area is obtained according to the fourth computer room fee feature map corresponding to each of the at least one candidate area; The fourth computer room cost feature map corresponding to each of the at least one candidate area is determined from the third computer room cost feature map according to the at least one candidate area in the computer room cost image, and the third computer room cost feature map is obtained by extracting the fourth feature from the computer room cost image; The room charge image region is obtained according to the fifth room charge feature map corresponding to at least one second scale; The fifth machine room fee feature map corresponding to each of the at least one second scale is obtained by extracting the fifth feature of the machine room fee image; The computer room fee image area is obtained according to the decoding vector; the decoding vector is obtained according to the encoding vector; the encoding vector is obtained by encoding the fused computer room fee feature vector; the fused computer room fee feature vector is obtained according to the sixth computer room fee feature map and the position feature corresponding to the computer room fee image; And the sixth computer room fee feature map is obtained by extracting the sixth feature of the computer room fee image.

10. The method according to any one of claims 1 to 9, wherein: The computer room expense image includes at least one of the following: a computer room electricity expense image, a computer room maintenance expense image, and a computer room rental expense image; The second computer room expense dimension data includes at least one of the following: the second computer room electricity fee, the second computer room maintenance fee, and the second computer room rental fee; the computer room expense-related data also includes computer room expense auxiliary dimension data, and the computer room expense auxiliary dimension data represents data used to audit the accuracy of the second computer room expense dimension data; the computer room expense auxiliary dimension data includes at least one of the following: computer room electricity fee auxiliary dimension data, computer room maintenance fee auxiliary dimension data, and computer room rental fee auxiliary dimension data; Wherein, the predetermined audit rules include at least one of the following: predetermined electricity fee audit rules, predetermined maintenance fee audit rules and predetermined rental fee audit rules; The predetermined electricity fee audit rules include predetermined electricity fee image audit rules and predetermined electricity fee data audit rules, the predetermined electricity fee image audit rules include at least one of the following: predetermined electricity fee image editing rules and predetermined electricity fee image duplication rules, the predetermined electricity fee data audit rules include at least one of the following: predetermined electricity fee payment cycle accuracy rules, predetermined power consumption accuracy rules and predetermined electricity fee accuracy rules; Wherein, the predetermined maintenance fee audit rule includes a predetermined maintenance fee image audit rule, and the predetermined maintenance fee image audit rule includes at least one of the following: a predetermined maintenance fee image repeatability rule and a predetermined maintenance fee image consistency rule; The predetermined rental fee auditing rules include predetermined rental fee data auditing rules, and the predetermined rental fee data auditing rules include at least one of the following: predetermined equipment association rules, predetermined electricity fee association rules, and predetermined rental fee accuracy rules; Wherein, when the computer room expense image includes the computer room electricity expense image, the second computer room expense dimension data includes the second computer room electricity expense, and the computer room expense auxiliary dimension data includes the computer room electricity expense auxiliary dimension data, the method further includes at least one of the following: Determine attribute data of the electricity bill image of the computer room according to the auxiliary dimension data of the electricity bill of the computer room, wherein the attribute data is used to evaluate whether the electricity bill image of the computer room is edited; and determine that the authenticity of the electricity bill image of the computer room satisfies the predetermined electricity bill image editing rule when it is determined that the attribute data matches the actual attribute data; In the case of determining that there is no image matching the electricity bill image of the computer room in the historical electricity bill images of the computer room, determining that the authenticity of the electricity bill image of the computer room satisfies the predetermined electricity bill image repetition rule; Determine, based on the auxiliary dimension data of the electricity fee of the computer room, the electricity fee payment cycle data of the computer room corresponding to the electricity fee of the second computer room; and determine, when it is determined that the electricity fee payment cycle data of the computer room matches the actual electricity fee payment cycle data, that the electricity fee of the second computer room meets the accuracy rule of the predetermined electricity fee payment cycle; Determine, according to the auxiliary dimension data of the electricity fee of the computer room, the power consumption of the computer room corresponding to the electricity fee of the second computer room, wherein the power consumption of the computer room represents the average daily power consumption in the electricity fee payment period of the computer room corresponding to the electricity fee of the second computer room; and when it is determined that the power consumption of the computer room is less than or equal to a predetermined power consumption threshold, determine that the electricity fee of the second computer room meets the predetermined power consumption accuracy rule, wherein the predetermined power consumption threshold is determined according to the power consumption of the reimbursement point corresponding to the second computer room; Determine the power meter number and the power meter record data of the computer room according to the auxiliary dimension data of the power meter number; and determine that the power meter number of the computer room matches the power meter record data of the computer room, and determine that the power meter of the second computer room meets the predetermined power meter accuracy rule; Wherein, when the computer room expense image includes the computer room maintenance expense image, the second computer room expense dimension data includes the second computer room maintenance expense, and the computer room expense auxiliary dimension data includes the computer room maintenance expense auxiliary dimension data, the method further includes at least one of the following: In the case of determining that there is no image matching the computer room maintenance fee image in the historical computer room maintenance fee images, determining that the authenticity of the computer room maintenance fee image satisfies the predetermined maintenance fee image duplication rule; In the case where it is determined that the second computer room location matches the actual computer room location, determining that the computer room maintenance fee image satisfies the predetermined maintenance fee image consistency rule, wherein the computer room maintenance fee auxiliary dimension data includes the second computer room location; Wherein, in the case where the second computer room cost dimension data includes the second computer room rental fee, and the computer room cost auxiliary dimension data includes the computer room rental fee auxiliary dimension data, the method further includes at least one of the following: Determining the equipment association data of the computer room according to the auxiliary dimension data of the computer room rental; and determining that the rental of the second computer room satisfies the predetermined equipment association rule when it is determined that there is equipment associated with the second computer room according to the equipment association data of the computer room; Determining the data related to the electricity charges of the computer room according to the auxiliary dimension data of the computer room rental; and determining that the second computer room rental satisfies the predetermined electricity charge association rule when it is determined that there is a second computer room electricity charge associated with the second computer room rental according to the data related to the electricity charges of the computer room; Determine the actual computer room rental fee according to the computer room rental fee auxiliary dimension data; and when it is determined that the second computer room rental fee matches the actual computer room rental fee, determine that the second computer room rental fee satisfies the predetermined rental fee accuracy rule.

Images