Data index management method and device, storage medium and computer program product

Abstract

The invention discloses a data index management method and device, a storage medium and a computer program product, and the method comprises the steps: a processor obtains a first weight of an enterprise organization structure, a second weight of a data index and a third weight of a user behavior, and determines a reward function based on the first weight, the second weight and the third weight; the processor determines a first feature vector of the enterprise organization structure, a second feature vector of the data index and a third feature vector of the user behavior, and tiles and splices the first feature vector, the second feature vector and the third feature vector to obtain a target state vector; the processor inputs the target state vector into a target deep reinforcement learning model and outputs priority information of the obtained data index, and the target deep reinforcement learning model is obtained through reward function training. According to the method and the device, the priority information of the data indexes is automatically generated, and the generation efficiency and flexibility of the priority information are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more specifically, to a data indicator management method and device, a storage medium, and a computer program product. Background Art

[0002] In related technologies, the sorting of task priorities mainly relies on manual operations or pre-set rules, which have the problems of low efficiency and poor flexibility. Manual sorting requires staff to manually adjust task priorities based on experience, which is difficult to respond in a timely manner in the face of complex business environments and is prone to subjectivity and inconsistency. Pre-determined fixed priorities lack dynamic adjustment capabilities and are difficult to adapt to real-time changing needs.

[0003] In addition, when dealing with complex tasks, existing solutions often fail to comprehensively consider the mutual influence of multi-dimensional factors, which can easily lead to problems such as unreasonable resource allocation or insufficient optimization, and therefore cannot meet the efficiency and flexibility requirements of modern enterprises for task scheduling. Summary of the invention

[0004] In view of this, the present invention provides a data indicator management method, a data indicator management device, a computer-readable storage medium and a computer program product, which aim to automatically generate priority information of data indicators and improve the efficiency and flexibility of generating priority information.

[0005] In order to achieve the above object, the present invention adopts the following technical scheme:

[0006] The present invention provides a data indicator management method, the method comprising:

[0007] The processor obtains a first weight of the enterprise organizational structure, a second weight of the data indicator, and a third weight of the user behavior, and determines a reward function based on the first weight, the second weight, and the third weight;

[0008] The processor determines a first eigenvector of the enterprise organizational structure, a second eigenvector of the data indicator, and a third eigenvector of the user behavior, and flattens and splices the first eigenvector, the second eigenvector, and the third eigenvector to obtain a target state vector;

[0009] The processor inputs the target state vector into a target deep reinforcement learning model and outputs priority information of the data indicator, wherein the target deep reinforcement learning model is obtained by training the reward function.

[0010] Optionally, determining a reward function based on the first weight, the second weight, and the third weight includes:

[0011] Determining the product of the third weight and the preset coefficient;

[0012] Add the product to the first weight and the second weight to get the reward function.

[0013] Optionally, the enterprise organizational structure includes at least one of position information, level information, and department information; the data indicator is an indicator corresponding to the enterprise organizational structure, and the processor determines a first feature vector of the enterprise organizational structure, a second feature vector of the data indicator, and a third feature vector of the user behavior, including:

[0014] The processor determines the department information to which the data indicator belongs;

[0015] The processor encodes the position information, the level information, the department information and the affiliated department information respectively to obtain a first feature vector; and,

[0016] The processor determines the output duration and computing power consumption of the data indicator within the first preset time period as a second feature vector; and,

[0017] The processor determines the number of times of user behavior in the second preset time period as a third feature vector, wherein the user behavior includes at least one of a user browsing data indicator, a user clicking data indicator, and a user collection data indicator.

[0018] Optionally, the target deep reinforcement learning model is trained by:

[0019] Get the policy network built by the deep neural network;

[0020] Optimize the output of the policy network based on the reward function until the output of the policy network is the highest;

[0021] The policy network with the highest output result is determined as the target deep reinforcement learning model.

[0022] Optionally, after the processor inputs the target state vector into the target deep reinforcement learning model and outputs priority information of the obtained data indicators, it includes:

[0023] The processor applies the priority information to the big data platform through automation technology and adjusts the output order of the priority information so that users can process the corresponding data indicators according to the priority information on the big data platform.

[0024] Optionally, after the processor inputs the target state vector into the target deep reinforcement learning model and outputs priority information of the obtained data indicators, it also includes:

[0025] The processor obtains the updated fourth weight of the user behavior again after a preset time;

[0026] The processor determines an updated reward function based on the first weight, the second weight, and the fourth weight;

[0027] The processor inputs the target state vector into an updated target deep reinforcement learning model, and outputs updated priority information, wherein the updated target deep reinforcement learning model is trained by the updated reward function.

[0028] Optionally, the processor encodes the position information, the level information, the department information and the affiliated department information respectively to obtain a first feature vector, including:

[0029] The position information, level information, department information and affiliated department information are respectively encoded by five-bit binary to obtain the position information code, level information code, department information code and affiliated department information code;

[0030] The position information code, the level information code, the department information code and the belonging department information code are combined to obtain the first eigenvector.

[0031] In a second aspect, the present invention provides a data indicator management device, including an acquisition module, a splicing module and an analysis module, wherein:

[0032] An acquisition module, electrically connected to the splicing module, for acquiring a first weight of the enterprise organizational structure, a second weight of the data indicator, and a third weight of the user behavior, and determining a reward function based on the first weight, the second weight, and the third weight;

[0033] A splicing module, electrically connected to the acquisition module and the analysis module, respectively, for determining a first feature vector of the enterprise organizational structure, a second feature vector of the data indicator, and a third feature vector of the user behavior, and flattening and splicing the first feature vector, the second feature vector, and the third feature vector to obtain a target state vector;

[0034] The analysis module is electrically connected to the splicing module, and is used to input the target state vector into the target deep reinforcement learning model and output the priority information of the data indicators, wherein the target deep reinforcement learning model is obtained by training the reward function.

[0035] The data indicator management device provided by the present invention can ensure that multiple slave controllers can start the upgrade process at the same time, and the delay-based control mechanism effectively avoids signal conflicts or excessive loads caused by simultaneous responses of multiple slave controllers, thereby improving the stability, reliability and upgrade efficiency of the upgrade process.

[0036] In a third aspect, the present invention provides a computer-readable storage medium, comprising: computer instructions, which, when executed on a data indicator management device, enable the data indicator management device to execute a data indicator management method as in the first aspect and any possible implementation thereof.

[0037] The computer-readable storage medium provided by the present invention can comprehensively consider multi-dimensional factors, thereby achieving a more comprehensive and accurate assessment of the status of an enterprise or user, and helping to improve the accuracy and effectiveness of generating priority information. Furthermore, by using deep reinforcement learning to determine the priority information of data indicators, the output results can be adaptively optimized, improving the flexibility and efficiency of generating priority information.

[0038] In a fourth aspect, the present invention provides a computer program product, comprising: computer instructions, which, when executed on a data indicator management device, cause the data indicator management device to execute the data indicator management method of the first aspect and any possible implementation thereof.

[0039] The computer program product provided by the present invention can comprehensively consider multi-dimensional factors, thereby achieving a more comprehensive and accurate assessment of the status of an enterprise or user, and helping to improve the accuracy and effectiveness of generating priority information. Furthermore, by using deep reinforcement learning to determine the priority information of data indicators, the output results can be adaptively optimized, improving the flexibility and efficiency of generating priority information.

[0040] Compared with the prior art, the data indicator management method provided by the present invention achieves at least the following beneficial effects: the processor obtains the first weight of the enterprise organizational structure, the second weight of the data indicator, and the third weight of the user behavior, and determines the reward function based on the first weight, the second weight, and the third weight; the processor determines the first eigenvector of the enterprise organizational structure, the second eigenvector of the data indicator, and the third eigenvector of the user behavior, and tiles and splices the first eigenvector, the second eigenvector, and the third eigenvector to obtain the target state vector; the processor inputs the target state vector into the target deep reinforcement learning model, and outputs the priority information of the data indicator, wherein the target deep reinforcement learning model is obtained by training the reward function. This application can comprehensively consider multi-dimensional factors by respectively obtaining the weights and eigenvectors of the enterprise organizational structure, data indicators, and user behavior, and tiling and splicing them, so as to achieve a more comprehensive and accurate assessment of the status of the enterprise or user, which helps to improve the accuracy and effectiveness of generating priority information. Furthermore, by using deep reinforcement learning to determine the priority information of data indicators, the output results can be adaptively optimized to improve the flexibility and efficiency of generating priority information.

[0041] Of course, any product implementing the present invention does not necessarily need to achieve all of the above technical effects at the same time.

[0042] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0044] Figure 1 It is a flowchart of a data indicator management method provided by an embodiment of the present invention;

[0045] Figure 2 It is a structural diagram of a data indicator management device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0047] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0048] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0049] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0050] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0051] Example 1

[0052] Reference Figure 1 , Figure 1 The present invention provides a flow chart of a data indicator management method, wherein the data indicator management method includes the following steps.

[0053] In step S11, the processor obtains a first weight of the enterprise organizational structure, a second weight of the data indicator, and a third weight of the user behavior, and determines a reward function based on the first weight, the second weight, and the third weight.

[0054] Among them, the first weight is the weight corresponding to the enterprise organizational structure; the second weight is the weight corresponding to the data indicator; the third weight is the weight corresponding to the user behavior. Furthermore, the first weight, the second weight and the third weight can be preset according to user experience, and this application does not limit this.

[0055] For example, the enterprise organizational structure is used to characterize the department or hierarchical structure of the enterprise. For example, the enterprise organizational structure may include the headquarters, branches, and various business departments (such as the company office, publicity department, and finance department). Since different organizations have different weights in corporate decision-making, for example, the weight of the headquarters is relatively larger than that of the branches. Therefore, the corresponding weights can be set in advance based on different organizational structures to reflect their relative importance in the overall enterprise organizational structure.

[0056] For example, data indicators are used to represent key performance indicators of concern in the internal management and operation of an enterprise, such as sales, profit margins, and customer satisfaction. Since different categories of data indicators have different levels of importance, different weights can be set in advance according to different categories of data indicators to reflect the relative importance of different categories of data indicators of the enterprise.

[0057] For example, user behaviors may include browsing, clicking, and collecting, etc., which are not limited in this application. Further, the weight of user behaviors may be determined by time decay, thereby reflecting the relative importance of different user behaviors.

[0058] It should be noted that time decay is a mathematical model or algorithm used to gradually reduce the weight or influence of a value over time. Its purpose is to reflect the decreasing importance of data in the time dimension. For example, in user behavior analysis, recent behaviors are usually more reflective of the user's current interests or needs, while older behaviors may no longer have the same reference value and are therefore assigned lower weights. Therefore, the present invention can capture real-time changes in user behavior through time decay, and then more realistically assign weights to different user behaviors.

[0059] For example, after obtaining the first weight, the second weight, and the third weight, they can also be combined to obtain a reward function. For example, they can be directly added together; or they can be multiplied by corresponding preset parameters to reflect their own influence in the overall reward function, so as to add up to obtain a reward function, which is not limited in this application.

[0060] The above implementation method, by calculating the reward function based on the weight of the company's organizational structure, indicator classification and user behavior, can comprehensively consider multi-dimensional factors, thereby achieving a more comprehensive and accurate assessment of the status of the enterprise or user, which helps to improve the accuracy and effectiveness of generating priority information.

[0061] In step S12, the processor determines a first eigenvector of the enterprise organizational structure, a second eigenvector of the data indicator, and a third eigenvector of the user behavior, and tiles and splices the first eigenvector, the second eigenvector, and the third eigenvector to obtain a target state vector.

[0062] Among them, the first eigenvector is the eigenvector corresponding to the enterprise organizational structure; the second eigenvector is the eigenvector corresponding to the data indicator; and the third eigenvector is the eigenvector corresponding to the user behavior.

[0063] For example, the first eigenvector, the second eigenvector, and the third eigenvector can be determined by a vectorization method. For example, for the first eigenvector, assuming that the enterprise organizational structure includes five main departments, a five-dimensional vector can be used to represent the contribution of each department, and its specific value can be normalized according to the number of employees or the influence of each department, thereby obtaining the first eigenvector.

[0064] For example, after obtaining the first eigenvector, the second eigenvector, and the third eigenvector, the three vectors can be tiled and spliced ​​so that multiple vectors are connected in parallel on the same dimension to form a unified vector, that is, the target state vector. This can realize the integration of the enterprise's organizational structure, data indicators, and user behavior for further analysis and processing.

[0065] It should be noted that tiling and splicing is an operation that combines multiple vectors or matrices in sequence. Its purpose is to represent features of different dimensions or sources in the same space, so that they can be analyzed as a whole or input into the machine learning model for subsequent processing.

[0066] In the above implementation modes, the processor can integrate multi-dimensional information into a unified representation, making complex data easier to process and analyze later.

[0067] In step S13, the processor inputs the target state vector into the target deep reinforcement learning model and outputs priority information of the data indicator, wherein the target deep reinforcement learning model is obtained by training the reward function.

[0068] For example, the target state vector contains comprehensive information about the enterprise's organizational structure, data indicators, and user behavior, and the combination of this information can represent the current overall state or business scenario of the enterprise. Therefore, the processor can pass the target state vector as input to a pre-trained target deep reinforcement learning model. As a result, the target deep reinforcement learning model can process the input information and output corresponding priority information. Among them, the priority information can help enterprises quickly determine the business indicators that should be paid most attention to in a complex business environment in order to make optimization decisions.

[0069] For example, the priority information can be output in the form of a numerical value or probability to indicate the importance of each data indicator in the current state. For example, sales, user retention rate, and profit rate can be assigned different priorities, so that the actions that should be taken first can be selected according to the priority information.

[0070] For example, the training of the target deep reinforcement learning model can be achieved through a reward function. Specifically, the reward function is the core of the target deep reinforcement learning model and can be used to evaluate the quality of the output of each target deep reinforcement learning model. Therefore, the parameters of the target deep reinforcement learning model can be adjusted through the reward function so that the target deep reinforcement learning model can output results with higher accuracy.

[0071] In the above implementation mode, the processor can input the target state vector that integrates the enterprise organizational structure, data indicators and user behavior into the target deep reinforcement learning model, and output the priority information of the data indicators, so as to adaptively optimize the output results and improve the flexibility and efficiency of priority information generation.

[0072] The data indicator management method provided by the present invention achieves at least the following beneficial effects: the processor obtains the first weight of the enterprise organizational structure, the second weight of the data indicator, and the third weight of the user behavior, and determines the reward function based on the first weight, the second weight, and the third weight; the processor determines the first eigenvector of the enterprise organizational structure, the second eigenvector of the data indicator, and the third eigenvector of the user behavior, and tiles and splices the first eigenvector, the second eigenvector, and the third eigenvector to obtain the target state vector; the processor inputs the target state vector into the target deep reinforcement learning model, and outputs the priority information of the data indicator, wherein the target deep reinforcement learning model is obtained by training the reward function. This application can comprehensively consider multi-dimensional factors by respectively obtaining the weights and eigenvectors of the enterprise organizational structure, data indicators, and user behavior, and tiling and splicing them, so as to achieve a more comprehensive and accurate assessment of the status of the enterprise or user, which helps to improve the accuracy and effectiveness of generating priority information. Furthermore, by using deep reinforcement learning to determine the priority information of data indicators, the output results can be adaptively optimized to improve the flexibility and efficiency of generating priority information.

[0073] Optionally, determining the reward function based on the first weight, the second weight and the third weight includes: determining the product of the third weight and a preset coefficient; and adding the product to the first weight and the second weight to obtain the reward function.

[0074] The preset coefficient is an adjustment parameter used to control the influence of the third weight and can be obtained through experience or experimental setting.

[0075] For example, a reward function is obtained by determining the product of the third weight and a preset coefficient and adding the product to the first weight and the second weight. The reward function can be used to comprehensively consider the influence of information in multiple dimensions and to train the target deep reinforcement learning model.

[0076] In the above implementation manner, the influence of the third weight can be flexibly controlled by presetting the coefficient, and the importance of different weights in the reward function can be ensured.

[0077] Optionally, the enterprise organizational structure includes at least one of position information, level information and department information; the data indicator is an indicator corresponding to the enterprise organizational structure, and the processor determines the first feature vector of the enterprise organizational structure, the second feature vector of the data indicator and the third feature vector of the user behavior, including: the processor determines the department information to which the data indicator belongs; the processor encodes the position information, level information, department information and the department information respectively to obtain a first feature vector; and, the processor determines the output time and computing power consumption of the data indicator within a first preset time period as a second feature vector; and, the processor determines the number of user behaviors within a second preset time period as a third feature vector, wherein the user behavior includes at least one of a user browsing data indicator, a user clicking data indicator and a user collection data indicator.

[0078] For example, the department information of the data indicator is used to characterize the department to which the data indicator belongs, such as the company office, publicity department, and finance department, etc., which is not limited in this application. Furthermore, the position information, level information, department information, and department information can be encoded separately to obtain a first feature vector. For example, the first feature vector includes a position information vector, a level information vector, a department information vector, and a department information vector, and its corresponding encoding can include (0,1,0,1,1), (1,0,0,1,1), (1,1,0,1,1), and (0,0,0,1,1), etc.

[0079] Furthermore, the output duration of the data indicator is used to indicate the time characteristics required to generate or update the data indicator; the computing power consumption is used to indicate the characteristics of the computing resources consumed during the processing and analysis of the data indicator, such as CPU utilization, memory usage, execution time, IO read and write speed, and number of concurrent tasks.

[0080] Furthermore, the first preset time period may include 7 days or 30 days; the second preset time period may include 15 days, 30 days, and 60 days, etc., which is not limited in the present application.

[0081] For example, the output duration and computing power consumption of the data indicator within the first preset time period can be determined as the second feature vector to evaluate the efficiency and resource usage of the data indicator. For example, the second feature vector includes (50, 20, 15, 20, 6).

[0082] For example, the number of user browsing data indicators, user clicking data indicators, and user favorite data indicators within the second preset time period may be counted and determined as a third feature vector, for example, the third feature vector includes (10, 9, 15).

[0083] The above implementation method can use these three types of feature vectors to enable the processor to comprehensively evaluate the organizational structure, business performance and user interest preferences of the enterprise, thereby providing data support for the subsequent determination of priority information.

[0084] Optionally, the target deep reinforcement learning model is trained in the following manner: obtaining a policy network constructed by a deep neural network; optimizing the output of the policy network based on a reward function until the output result of the policy network is the highest; and determining the policy network with the highest output result as the target deep reinforcement learning model.

[0085] It should be noted that the policy network is the core component in deep reinforcement learning, which can output the optimal action strategy in the current state based on the input feature data.

[0086] For example, in order to enable the policy network to output effective decisions, deep reinforcement learning can guide the optimization process of the policy network through a reward function. That is, the reward function can define an evaluation criterion to determine whether the decision output by the policy network is superior. Furthermore, if the output result of the policy network is evaluated as the highest by the reward function, it means that the policy network can make the best decision in the current environment, so the policy network with the highest output result can be determined as the target deep reinforcement learning model.

[0087] The above implementation methods can obtain the target deep reinforcement learning model based on reward function training, so that priority information can be generated in complex decision-making scenarios through the target deep reinforcement learning model, reducing human intervention and achieving smarter enterprise management and optimization processes.

[0088] Optionally, after the processor inputs the target state vector into the target deep reinforcement learning model and outputs the priority information of the data indicator, it includes: the processor applies the priority information to the big data platform through automation technology, and adjusts the output order of the priority information so that the user can process the corresponding data indicators on the big data platform according to the priority information.

[0089] For example, after the target deep reinforcement learning model calculates and outputs the priority information of the data indicator, the processor can directly apply the priority information to the enterprise's big data platform, so that the big data platform can perform corresponding processing and optimization based on the priority information.

[0090] For example, in the big data platform, each data indicator has its processing order. The processor arranges the priority information output by the target deep reinforcement learning model in order from high to low. As a result, the data indicators with high priority will be processed first, and the data indicators with low priority will be processed later. Furthermore, users can process the corresponding data indicators according to the priority information on the big data platform.

[0091] In the above implementation methods, the processor can not only apply the priority information output by the deep reinforcement learning model to the big data platform, but also dynamically adjust the output order of data indicators, ensuring that high-priority data indicators are processed first, helping users to process the most important data tasks on the big data platform more efficiently, optimizing the overall workflow and improving the system's response speed.

[0092] Optionally, after the processor inputs the target state vector into the target deep reinforcement learning model and outputs the priority information of the data indicator, it also includes: the processor obtains the updated fourth weight of the user behavior again after a preset time; the processor determines the updated reward function based on the first weight, the second weight and the fourth weight; the processor inputs the target state vector into the updated target deep reinforcement learning model, and outputs the updated priority information, wherein the updated target deep reinforcement learning model is trained by the updated reward function.

[0093] For example, after the target deep reinforcement learning model outputs preliminary priority information, the processor can also monitor changes in user behavior, and after a preset time, obtain the latest user behavior again, and then obtain the fourth weight of the updated user behavior. Furthermore, after obtaining the fourth weight, the processor can also determine the updated reward function in combination with the previous first weight and the second weight, and retrain the updated target deep reinforcement learning model. Through the training method of deep reinforcement learning, the model will continuously adjust its output according to the updated reward function until the highest output result is reached, that is, the updated target deep reinforcement learning model is obtained.

[0094] It should be understood that, compared with the previous priority information, the new priority information is more accurate and can better reflect the user's latest behavior and data demand changes.

[0095] In the above implementation mode, by re-obtaining the fourth weight of the user behavior after a preset time and updating the reward function based on the first weight, the second weight and the fourth weight, the processor can dynamically adjust the output priority of the target deep reinforcement learning model, ensure that the priority information of data processing is consistent with actual business needs and changes in user behavior, and continuously optimize the system's response speed and decision-making accuracy.

[0096] Optionally, the processor encodes the position information, level information, department information and belonging department information respectively to obtain a first feature vector, including: encoding the position information, level information, department information and belonging department information respectively through five-bit binary to obtain a position information code, a level information code, a department information code and belonging department information code; combining the position information code, the level information code, the department information code and the belonging department information code to obtain the first feature vector.

[0097] It should be noted that five-bit binary is a coding method that uses five binary bits to represent data. Binary is a number system with only two states, usually represented by 0 and 1. In five-bit binary, each position of the five binary bits can be 0 or 1, thus representing all binary combinations from 00000 to 11111.

[0098] For example, the processor can independently encode the position information, level information, department information, and affiliated department information through five binary digits, and combine them to obtain the first feature vector. For example, the processor can sequentially combine the position information encoding, level information encoding, department information encoding, and affiliated department information encoding. Since each encoding is a five-bit binary number, the four encodings combined together will form a binary number with a length of 20 bits.

[0099] In the above embodiments, the processor quantifies the position information, level information, department information and affiliated department information through five-bit binary coding, and the obtained first eigenvector provides a basis for subsequent data processing and deep learning.

[0100] It should be easily appreciated by those skilled in the art that, in combination with the algorithm steps of each example described in the embodiments disclosed herein, the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can 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 the present invention.

[0101] Example 2

[0102] Figure 2 is a structural diagram of a data indicator management device provided by an embodiment of the present invention, such as Figure 2 As shown, the data indicator management device 200 includes an acquisition module 21, a splicing module 22 and an analysis module 23.

[0103] The acquisition module 21 is electrically connected to the splicing module 22, and is used to obtain a first weight of the enterprise organizational structure, a second weight of the data indicator, and a third weight of the user behavior, and determine a reward function based on the first weight, the second weight, and the third weight;

[0104] The splicing module 22 is electrically connected to the acquisition module 21 and the analysis module 23, respectively, and is used to determine the first feature vector of the enterprise organizational structure, the second feature vector of the data indicator, and the third feature vector of the user behavior, and to flatten and splice the first feature vector, the second feature vector, and the third feature vector to obtain a target state vector;

[0105] The analysis module 23 is electrically connected to the splicing module 22, and is used to input the target state vector into the target deep reinforcement learning model and output the priority information of the data indicator, wherein the target deep reinforcement learning model is obtained by training the reward function.

[0106] The acquisition module 21 is further used to determine the product of the third weight and the preset coefficient; and add the product to the first weight and the second weight to obtain the reward function.

[0107] The splicing module 22 is also used to determine the department information to which the data indicator belongs; encode the position information, level information, department information and department information respectively to obtain a first feature vector; and determine the output duration and computing power consumption of the data indicator within the first preset time period as a second feature vector; and determine the number of user behaviors within the second preset time period as a third feature vector, wherein the user behavior includes at least one of a user browsing data indicator, a user clicking data indicator and a user collection data indicator.

[0108] The acquisition module 21 is also used to obtain a policy network built by a deep neural network; optimize the output of the policy network based on the reward function until the output result of the policy network is the highest; and determine the policy network with the highest output result as the target deep reinforcement learning model.

[0109] The analysis module 23 is also used to apply the priority information to the big data platform through automation technology, and adjust the output order of the priority information so that users can process the corresponding data indicators according to the priority information on the big data platform.

[0110] The analysis module 23 is also used to obtain the updated fourth weight of the user behavior again after a preset time; determine the updated reward function based on the first weight, the second weight and the fourth weight; input the target state vector into the updated target deep reinforcement learning model, and output the updated priority information, wherein the updated target deep reinforcement learning model is obtained by training the updated reward function.

[0111] The splicing module 22 is also used to encode the position information, level information, department information and belonging department information respectively through five-bit binary to obtain the position information code, level information code, department information code and belonging department information code; combine the position information code, level information code, department information code and belonging department information code to obtain the first feature vector.

[0112] Of course, the data indicator management device provided by the embodiment of the present invention includes but is not limited to the above modules.

[0113] The data indicator management device provided by the present invention aims to realize automatic generation of priority information of data indicators and improve the generation efficiency and flexibility of priority information.

[0114] Example 3

[0115] An embodiment of the present invention also provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed on a data indicator management device, the data indicator management device executes each step executed by the data indicator management device in the method flow shown in the above method embodiment.

[0116] Example 4

[0117] An embodiment of the present invention also provides a computer program product, which includes computer instructions. When the computer instructions are executed on a data indicator management device, the data indicator management device executes each step executed by the data indicator management device in the method flow shown in the above method embodiment.

[0118] Although some specific embodiments of the present invention have been described in detail by way of example, it will be appreciated by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It will be appreciated by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present invention. The scope of the present invention is defined by the appended claims.

Claims

1. A data indicator management method, characterized in that: The method comprises: The processor obtains a first weight of the enterprise organizational structure, a second weight of the data indicator, and a third weight of the user behavior, and determines a reward function based on the first weight, the second weight, and the third weight; The processor determines a first feature vector of the enterprise organizational structure, a second feature vector of the data indicator, and a third feature vector of the user behavior, and flattens and splices the first feature vector, the second feature vector, and the third feature vector to obtain a target state vector; The processor inputs the target state vector into a target deep reinforcement learning model, and outputs priority information of the data indicator, wherein the target deep reinforcement learning model is trained by the reward function.

2. The method according to claim 1, characterized in that The determining of a reward function based on the first weight, the second weight, and the third weight comprises: Determining the product of the third weight and a preset coefficient; The reward function is obtained by adding the product to the first weight and the second weight.

3. The method according to claim 1, characterized in that The enterprise organizational structure includes at least one of position information, level information, and department information; the data indicator is an indicator corresponding to the enterprise organizational structure, and the processor determines a first feature vector of the enterprise organizational structure, a second feature vector of the data indicator, and a third feature vector of the user behavior, including: The processor determines the department information to which the data indicator belongs; The processor encodes the position information, the level information, the department information and the belonging department information respectively to obtain the first feature vector; and The processor determines the output duration and computing power consumption of the data indicator within the first preset time period as the second feature vector; and The processor determines the number of times the user behavior occurs within a second preset time period as the third feature vector, wherein the user behavior includes at least one of a user browsing the data indicator, a user clicking on the data indicator, and a user collecting the data indicator.

4. The method according to claim 1, characterized in that The target deep reinforcement learning model is trained in the following way: Get the policy network built by the deep neural network; Optimizing the output of the policy network based on the reward function until the output result of the policy network is the highest; The policy network with the highest output result is determined as the target deep reinforcement learning model.

5. The method according to claim 1, characterized in that After the processor inputs the target state vector into the target deep reinforcement learning model and outputs the priority information of the data indicator, the processor includes: The processor applies the priority information to the big data platform through automation technology and adjusts the output order of the priority information so that the user can process the corresponding data indicators on the big data platform according to the priority information.

6. The method according to claim 1, characterized in that After the processor inputs the target state vector into the target deep reinforcement learning model and outputs the priority information of the data indicator, it also includes: The processor obtains the updated fourth weight of the user behavior again after a preset time; The processor determines an updated reward function based on the first weight, the second weight, and the fourth weight; The processor inputs the target state vector into an updated target deep reinforcement learning model, and outputs updated priority information, wherein the updated target deep reinforcement learning model is trained by the updated reward function.

7. The method according to claim 3, characterized in that The processor encodes the position information, the level information, the department information, and the belonging department information respectively to obtain the first feature vector, including: The position information, the level information, the department information and the affiliated department information are respectively encoded by five-bit binary to obtain a position information code, a level information code, a department information code and an affiliated department information code; The position information code, the level information code, the department information code and the belonging department information code are combined to obtain the first feature vector.

8. A data indicator management device, characterized in that: It includes acquisition module, splicing module and analysis module, among which, The acquisition module is electrically connected to the splicing module, and is used to acquire a first weight of the enterprise organizational structure, a second weight of the data indicator, and a third weight of the user behavior, and determine a reward function based on the first weight, the second weight, and the third weight; The splicing module is electrically connected to the acquisition module and the analysis module respectively, and is used to determine the first feature vector of the enterprise organizational structure, the second feature vector of the data indicator, and the third feature vector of the user behavior, and to flatten and splice the first feature vector, the second feature vector, and the third feature vector to obtain a target state vector; The analysis module is electrically connected to the splicing module, and is used to input the target state vector into a target deep reinforcement learning model, and output priority information of the data indicator, wherein the target deep reinforcement learning model is obtained by training the reward function.

9. A computer-readable storage medium, characterized in that: include: Computer instructions, when the computer instructions are executed on a data indicator management device, enable the data indicator management device to execute the data indicator management method according to any one of claims 1 to 7.

10. A computer program product, the computer program product comprising: Computer instructions, when the computer instructions are executed on the data indicator management device, make the data indicator management device execute the data indicator management method as described in any one of claims 1-7.

Images