Virtual product standard portrait storage method and device, equipment and medium

Abstract

The embodiment of the invention discloses a virtual product standard portrait storage method and device, equipment and a medium. A specific embodiment of the method comprises the steps of obtaining a product use data set; for each product use object, generating an object portrait; carrying out attribute and attribute information cutting on the object portrait set to obtain at least one product use attribute and a removed object portrait set; for each product use attribute, executing a generation step: determining a product attribute information set; obtaining an attribute anomaly determination rule set and an attribute anomaly generation model set; respectively screening out a target attribute anomaly determination rule and a target attribute anomaly generation model; generating attribute exception standard information; generating a first standard information portrait; and storing the first standard information portrait and the removed object portrait set. According to the embodiment, under the condition of adapting to the current calculation residual resources, the standard information portrait for carrying out abnormity judgment on the target virtual product can be efficiently and accurately generated.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a method, device, equipment and medium for storing standard portraits of virtual products. Background Art

[0002] At present, in the use scenarios of virtual products, the abnormality judgment of virtual products is crucial and can effectively avoid value loss. For the generation of abnormality judgment standards corresponding to virtual products, the usual method is to directly input the product usage data set corresponding to the target virtual product into the multi-attribute abnormality standard information output model to obtain the abnormality judgment standards under each attribute.

[0003] However, when using the above method, the following technical problems often occur:

[0004] The data sets used by products are often large in size, which results in a large amount of computing time wasted when directly input into the multi-attribute anomaly standard information output model, causing the multi-attribute anomaly standard information output model to learn more redundant features. In addition, if the multi-attribute anomaly standard information output model requires a large amount of computing resources, if the terminal cannot effectively supply computing resources, the output will be interrupted and the generation of anomaly judgment standards will be affected.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0006] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] Some embodiments of the present disclosure propose methods, devices, equipment and media for storing standard portraits of virtual products to solve the technical problems mentioned in the above background technology section.

[0008] In a first aspect, some embodiments of the present disclosure provide a method for storing a standard portrait of a virtual product, comprising: obtaining a product usage data set of a target virtual product within a predetermined time period; for each product usage object in a product usage object set, generating an object portrait corresponding to the product usage object according to the product usage data corresponding to the product usage object; performing attribute and attribute information clipping on a product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and a set of object portraits after the removal; for each product usage attribute in the at least one product usage attribute, using a distributed graphics processor cluster, executing a generation step: determining a product attribute information set corresponding to the product usage attribute in the object portrait set; obtaining the product The method comprises the following steps: a) determining a set of attribute anomaly determination rules and a set of attribute anomaly generation models in advance according to the product usage attributes; b) filtering out target attribute anomaly determination rules and target attribute anomaly generation models from the attribute anomaly determination rule set and the attribute anomaly generation model set according to the computing surplus resources corresponding to the above-mentioned graphics processor cluster; c) generating attribute anomaly standard information for the above-mentioned product usage attributes by using the above-mentioned target attribute anomaly determination rules and the above-mentioned target attribute anomaly generation model according to the above-mentioned product attribute information set; d) generating a first standard information portrait corresponding to at least one attribute anomaly standard information obtained, wherein the above-mentioned first standard information portrait and the object portrait have the same tree structure; and d) storing the above-mentioned first standard information portrait and the above-mentioned removed object portrait set in a target storage terminal.

[0009] In a second aspect, some embodiments of the present disclosure provide a virtual product standard portrait storage device, including: an acquisition unit, configured to acquire a product usage data set of a target virtual product within a predetermined time period; a first generation unit, configured to generate, for each product usage object in a product usage object set, an object portrait corresponding to the product usage object according to the product usage data corresponding to the product usage object; a cropping unit, configured to perform attribute and attribute information cropping on a product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and a post-removal object portrait set; an execution unit, configured to, for each product usage attribute in the at least one product usage attribute, use a distributed graphics processor cluster to execute a generation step: determining a product attribute in the object portrait set corresponding to the product usage attribute information set; obtaining a pre-stored attribute anomaly determination rule set and an attribute anomaly generation model set corresponding to the above-mentioned product usage attributes; according to the computing surplus resources corresponding to the above-mentioned graphics processor cluster, respectively screening out a target attribute anomaly determination rule and a target attribute anomaly generation model from the above-mentioned attribute anomaly determination rule set and the attribute anomaly generation model set; according to the above-mentioned product attribute information set, using the above-mentioned target attribute anomaly determination rule and the above-mentioned target attribute anomaly generation model, generating attribute anomaly standard information for the above-mentioned product usage attributes; a second generating unit, configured to generate a first standard information portrait corresponding to at least one attribute anomaly standard information obtained, wherein the above-mentioned first standard information portrait and the object portrait are in the same tree structure; a storage unit, configured to store the above-mentioned first standard information portrait and the above-mentioned removed object portrait set in a target storage terminal.

[0010] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner in the first aspect.

[0011] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner in the first aspect is implemented.

[0012] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the virtual product standard portrait storage method of some embodiments of the present disclosure, under the condition of adapting to the current computing surplus resources, the standard information portrait for abnormal determination of the target virtual product can be efficiently and accurately generated. Specifically, the reason why the relevant abnormal determination is not accurate enough and is not generated in a timely manner is that the product usage data set is often large in data volume, resulting in a large amount of computing time wasted by directly inputting it into the multi-attribute abnormal standard information output model, and allowing the multi-attribute abnormal standard information output model to learn more redundant features. And in the case where the multi-attribute abnormal standard information output model requires a large amount of computing resources, if the terminal cannot achieve the effective supply of computing resources, the output is interrupted and the generation of the abnormal determination standard is affected. Based on this, the virtual product standard portrait storage method of some embodiments of the present disclosure, first, obtains the product usage data set of the target virtual product within a predetermined time period, so as to be used as the basic data set for the abnormal determination of each product usage attribute. Based on this, the accurate generation of the virtual product standard portrait can be achieved. Then, for each product usage object in the product usage object set, according to the product usage data corresponding to the above product usage object, the object portrait corresponding to the above product usage object can be accurately generated. Through the object portrait, it is convenient to determine the object portrait of each user object corresponding to the target virtual product, so as to facilitate the query of the subsequent product use object related information. Then, the product use attribute set corresponding to the obtained object portrait set is subjected to attribute and attribute information clipping, and at least one product use attribute and the removed object portrait set are obtained to determine at least one product use attribute for key monitoring and the removed object portrait set including the key product use attribute information. Then, for each product use attribute in the above-mentioned at least one product use attribute, a distributed graphics processor cluster is used to perform a generation step: the first step is to determine the product attribute information set corresponding to the above-mentioned product use attribute in the above-mentioned object portrait set, so as to facilitate the generation of attribute abnormality standard information corresponding to the subsequent product use attribute. In addition, the use of a distributed graphics processor cluster can effectively provide computing resources that can be used for calculation. The second step is to obtain the attribute abnormality determination rule set and the attribute abnormality generation model set corresponding to the above-mentioned product use attribute, so as to jointly determine the abnormality judgment standard from the perspective of rules and the perspective of neural network models, which can greatly improve the accuracy of judgment. The third step is to select target attribute anomaly determination rules and target attribute anomaly generation models from the attribute anomaly determination rule set and attribute anomaly generation model set respectively according to the computing surplus resources corresponding to the above-mentioned graphics processor cluster, so as to effectively select rules and models for subsequent real-time generation of attribute anomaly standard information under the current computing resources. The fourth step is to accurately generate attribute anomaly standard information for the usage attributes of the above-mentioned product using the above-mentioned target attribute anomaly determination rules and the above-mentioned target attribute anomaly generation model according to the above-mentioned product attribute information set.Further, a first standard information portrait corresponding to at least one attribute abnormality standard information is generated, wherein the first standard information portrait and the object portrait are in the same tree structure. Here, in the form of the first standard information portrait, the efficiency of abnormality determination can be improved and the determination time can be saved in the subsequent abnormality determination. Finally, the first standard information portrait and the above-mentioned removed object portrait set are stored in the target storage terminal to facilitate the execution of subsequent abnormality determination and the regular update of the first standard information portrait. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0014] Figure 1 is a flow chart of some embodiments of a method for storing a virtual product standard portrait according to the present disclosure;

[0015] Figure 2 is a schematic diagram of the structure of some embodiments of a virtual product standard portrait storage device according to the present disclosure;

[0016] Figure 3 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0017] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0018] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0019] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0020] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0021] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0022] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0023] refer to Figure 1 , shows a process 100 of some embodiments of the virtual product standard portrait storage method according to the present disclosure. The virtual product standard portrait storage method includes the following steps:

[0024] Step 101, obtain a product usage data set of the target virtual product within a predetermined time period. In some embodiments, the execution subject (for example, an electronic device) of the above-mentioned virtual product standard portrait storage method can obtain the product usage data set of the target virtual product within a predetermined time period through a wired connection method or a wireless connection method. Among them, the target virtual product may be a virtual product to be subjected to abnormal standard judgment. In practice, the virtual product may be a virtual product related to value. For example, in the field of credit investigation, the virtual product may be a credit product. For another example, for the express delivery scenario, the corresponding virtual product may be an express delivery App. The predetermined time period may be a pre-set time period. The predetermined time period may be a historical time period before the current time. The product usage data may be the usage data of the target virtual product used by the product user object within the predetermined time period. Each product usage data corresponds to a product user object. The product user object may be an object that uses the target virtual product.

[0025] Step 102, for each product usage object in the product usage object set, generate an object portrait corresponding to the product usage object based on the product usage data corresponding to the product usage object.

[0026] In some embodiments, the execution subject may generate an object portrait corresponding to each product user object in the product user object set according to the product usage data corresponding to the product user object. The object portrait may be a descriptive portrait corresponding to the product user object. The object portrait may characterize the usage characteristics of various aspects of the target virtual product used by the product user object. In practice, the object portrait may include but is not limited to at least one of the following: product user object name, product usage time, investment amount, profit amount, product usage time, and product natural person.

[0027] As an example, first, the execution subject may extract each product usage attribute information corresponding to each product usage attribute from the product usage data, and then construct an object portrait in a tree structure according to each product usage attribute information.

[0028] Step 103 , performing attribute and attribute information clipping on the product usage attribute set corresponding to the obtained object portrait set, to obtain at least one product usage attribute and the object portrait set after the attribute is removed.

[0029] In some embodiments, the execution subject may perform attribute and attribute information clipping on the product usage attribute set corresponding to the obtained object portrait set, and obtain at least one product usage attribute and a post-removal object portrait set. Among them, at least one product usage attribute may characterize each product usage attribute that is critical in the current abnormality determination process of the target virtual product. That is, the abnormal situation of the product usage object using the target virtual product is determined by the attribute content corresponding to at least one product usage attribute. The post-removal object portrait may be a portrait of the object portrait without each product usage attribute information that is critical in the abnormality determination process of the target virtual product.

[0030] As an example, first, each product usage attribute that is crucial to the abnormality determination process of the target virtual product selected by the product usage object in the target interface is obtained as at least one product usage attribute. Then, for each object portrait in the object portrait set, attribute information other than the at least one product usage attribute is removed from the above object portrait to obtain a removed object portrait.

[0031] In some optional implementations of some embodiments, the execution subject may perform attribute and attribute information clipping on the product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and the object portrait set after the attribute is removed, including the following steps:

[0032] The first step is to generate effective attribute prediction information corresponding to each product usage attribute in the product usage attribute set to obtain an effective attribute prediction information set. The effective attribute prediction information is an effective value that characterizes the importance of the attribute and the strength of the attribute prediction capability. The strength of the attribute prediction capability can characterize the ease with which the attribute content corresponding to the attribute can be predicted. The higher the value of the corresponding attribute prediction capability strength, the easier it is to predict the corresponding attribute content. The effective attribute prediction information can be a value set for each attribute by relevant technical personnel based on historical experience.

[0033] The second step is to remove product usage attributes whose corresponding attribute prediction effective information is less than a target value from the above product usage attribute set to obtain at least one product usage attribute. The target value may be a value pre-set by relevant technical personnel based on experience.

[0034] In the third step, for each object portrait in the object portrait set, attribute information corresponding to the at least one product usage attribute is removed from the object portrait to obtain a post-removal object portrait.

[0035] Step 104, for each product usage attribute in the at least one product usage attribute, using a distributed graphics processor cluster, execute a generation step:

[0036] Step 1041, determining a product attribute information set in the object portrait set corresponding to the product usage attributes.

[0037] In some embodiments, the execution subject may determine a product attribute information set in the object portrait set corresponding to the product usage attribute. The graphics processor cluster may be a cluster based on a graphics processor (GPU). The object portrait in the object portrait set and the product attribute information in the product attribute information set have a usage object correspondence relationship.

[0038] Step 1042, obtaining the pre-stored attribute anomaly determination rule set and attribute anomaly generation model set corresponding to the above-mentioned product usage attributes.

[0039] In some embodiments, the above-mentioned execution subject may obtain the attribute anomaly determination rule set and the attribute anomaly generation model set pre-stored corresponding to the above-mentioned product usage attributes. The attribute anomaly determination rule may be a determination rule for determining attribute anomalies pre-set by relevant expert technicians. The attribute anomaly generation model may be a pre-trained neural network model for predicting attribute anomaly information. In practice, the model structure or model training method corresponding to each attribute anomaly generation model in the attribute anomaly generation model set may be different. The rule content corresponding to each attribute anomaly determination rule in the attribute anomaly determination rule set may be different. The model structure corresponding to the attribute anomaly generation model is determined according to the content type of the attribute content corresponding to the product usage attributes. For example, for a content type of label type, each attribute anomaly generation model in the corresponding attribute anomaly generation model set may be a classification model. For a content type of numerical continuous type, each attribute anomaly model in the corresponding attribute anomaly generation model set may be a regression model.

[0040] Step 1043 , based on the computing surplus resources corresponding to the GPU cluster, select a target attribute anomaly determination rule and a target attribute anomaly generation model from the attribute anomaly determination rule set and the attribute anomaly generation model set, respectively.

[0041] In some embodiments, the execution subject may use various methods to filter out target attribute anomaly determination rules and target attribute anomaly generation models from the attribute anomaly determination rule set and attribute anomaly generation model set, respectively, based on the remaining computing resources corresponding to the graphics processor cluster.

[0042] In some optional implementations of some embodiments, the execution subject may respectively filter out a target attribute anomaly determination rule and a target attribute anomaly generation model from the attribute anomaly determination rule set and the attribute anomaly generation model set according to the computing surplus resources corresponding to the graphics processor cluster, including the following steps:

[0043] The first step is to determine the computational remaining resource sequence of the above-mentioned graphics processor cluster in the target future time period. The target future time period may be a time period within a future predetermined time period. The future predetermined time period may be pre-set. The computational remaining resources in the computational remaining resource sequence correspond to future time points in the target future time period in a one-to-one correspondence. The computational remaining resource sequence may characterize the change of resources of the graphics processor cluster in the target future time period.

[0044] The second step is to determine the attribute division resource sequence corresponding to the above-mentioned product usage attributes from the above-mentioned calculated remaining resource sequence, wherein each product usage attribute has a corresponding attribute division resource sequence. The attribute division resources can be the remaining resources that can be used by the product usage attributes at the target future time point. The attribute division resources in the attribute division resource sequence have a one-to-one correspondence with the future time point in the target future time period. That is, the attribute division resource sequence can represent the resource change sequence that the graphics processor can provide for the attribute calculation process corresponding to the product usage attributes.

[0045] As an example, for each computational remaining resource in the computational remaining resource sequence, the number of attributes corresponding to the at least one product usage attribute is determined, and then the computational remaining resource is divided by the number of attributes to obtain the attribute partitioned resource corresponding to the product usage attribute.

[0046] The third step is to determine a collocation combination set between the above-mentioned attribute anomaly determination rule set and the above-mentioned attribute anomaly generation model set. Each collocation combination includes: an attribute anomaly determination rule and an attribute anomaly generation model. Each collocation combination in the collocation combination set is flexibly collocated based on the total computing amount consumed by the attribute anomaly determination rule and the attribute anomaly generation model. The total computing amount can represent the resource consumption.

[0047] It should be noted that each attribute anomaly determination rule has a corresponding rule accuracy and calculation amount, and the attribute anomaly generation model has a corresponding model accuracy and calculation amount. Relevant R&D personnel can also make corresponding combinations of each attribute anomaly determination rule and each attribute anomaly generation model by dragging and dropping on the interface to obtain a matching combination set.

[0048] Step 4: For each combination in the above combination set, perform the following generation steps:

[0049] Sub-step 1, determining the estimated consumption time and estimated consumption resources corresponding to the attribute anomaly determination rule and the attribute anomaly generation model in the above combination. The estimated consumption time can be the total time estimated to be consumed by the attribute anomaly determination rule and the attribute anomaly generation model to complete the attribute anomaly determination. The estimated consumption resources can be the total amount of resources estimated to be consumed by the attribute anomaly determination rule and the attribute anomaly generation model to complete the attribute anomaly determination.

[0050] As an example, first, determine the first consumption duration corresponding to the attribute anomaly determination rule and the second consumption duration corresponding to the attribute anomaly generation model. Then, determine the highest consumption duration between the first consumption duration and the second consumption duration as the estimated consumption duration. Next, determine the first consumption resource corresponding to the attribute anomaly determination rule and the second consumption resource corresponding to the attribute anomaly generation model. Then, determine the sum of the resources between the first consumption resource and the second consumption resource as the estimated consumption resource.

[0051] Sub-step 2, determining whether the above-mentioned attribute division resource sequence satisfies the consumption requirements corresponding to the above-mentioned estimated consumption duration and estimated consumption resources, and obtaining satisfaction information. The satisfaction information may be one of the following: satisfaction information indicating that the above-mentioned estimated consumption duration and the consumption requirements corresponding to the estimated consumption resources are satisfied, and satisfaction information indicating that the above-mentioned estimated consumption duration and the consumption requirements corresponding to the estimated consumption resources are not satisfied. The consumption requirement may be that the attribute division resource corresponding to the attribute division resource subsequence under the estimated consumption duration in the attribute division resource sequence is greater than the above-mentioned estimated consumption resources.

[0052] In the fifth step, the satisfying information that represents the consumption requirements corresponding to the estimated consumption time and the estimated consumption resources are selected from the obtained satisfying information set to obtain at least one satisfying information.

[0053] Step 6: Select the combination with the highest comprehensive accuracy and the shortest estimated consumption time from at least one combination as the target combination. The target combination includes: a target attribute abnormality determination rule and a target attribute abnormality generation model. There is a one-to-one correspondence between the at least one satisfying information and the at least one combination.

[0054] Optionally, the execution subject may determine the attribute division resource sequence corresponding to the product usage attribute from the calculated remaining resource sequence, comprising the following steps:

[0055] The first step is to determine the attribute weight corresponding to each of the at least one product usage attribute to obtain at least one attribute weight. The attribute weight may represent the attribute importance corresponding to the product usage attribute. The attribute weight may be a value between 0 and 1, and the higher the value, the more important the attribute content corresponding to the product usage attribute is.

[0056] In the second step, each computing remaining resource in the above computing remaining resource sequence is divided accordingly according to the above at least one attribute weight to generate available computing resources at the corresponding time for the product usage attribute, and obtain at least one attribute-divided resource sequence.

[0057] As an example, for each computing surplus resource in the computing surplus resource sequence, the computing surplus resource is divided into at least one portion according to the at least one attribute weight, to obtain at least one attribute-divided resource.

[0058] The third step is to select the attribute division resource sequence corresponding to the product usage attribute from the at least one attribute division resource sequence.

[0059] In some optional implementations of some embodiments, the execution subject may respectively filter out a target attribute anomaly determination rule and a target attribute anomaly generation model from the attribute anomaly determination rule set and the attribute anomaly generation model set according to the computing surplus resources corresponding to the graphics processor cluster, including the following steps:

[0060] In the first step, the remaining computing resources corresponding to the GPU cluster are displayed on the screening interface, wherein the screening interface displays the remaining computing resources and the initial attribute usage computing resources corresponding to each attribute.

[0061] The second step is to obtain the adjustment information of the target user adjusting the attribute usage computing resources for each attribute in the above-mentioned screening interface.

[0062] In the third step, the initial attribute usage computing resources corresponding to each attribute are adjusted according to the adjustment information, and the attribute usage computing resources corresponding to each attribute after the adjustment are displayed on the screening interface.

[0063] In the fourth step, in response to confirming and clicking the attribute control corresponding to the target attribute in the filtering interface, a pop-up window for configuring the rules and models corresponding to the target attribute is popped up. Among them, the left half of the pop-up window for configuring the rules and models is the first configuration area corresponding to the target attribute abnormality determination rule. The right half of the pop-up window for configuring the rules and models is the second configuration area corresponding to the target attribute abnormality generation model. Among them, the first configuration area and the second configuration area support custom setting of rules and models. The first configuration area and the second configuration area also support importing rules and models from external files.

[0064] The fifth step, in response to determining to configure the attribute anomaly determination rule and the target attribute anomaly generation model for the target attribute in the rule and model configuration pop-up window and clicking the configuration save control, set the attribute control corresponding to the above target attribute to the target color and display the identification number corresponding to the attribute anomaly determination rule and the identification number corresponding to the target attribute anomaly generation model in the upper right part of the attribute control corresponding to the above target attribute, so as to realize the configuration of the target attribute anomaly determination rule and the target attribute anomaly generation model corresponding to the target attribute and the resource configuration of the target attribute.

[0065] Step 1044, based on the above product attribute information set, using the above target attribute anomaly determination rule and the above target attribute anomaly generation model, generate attribute anomaly standard information for the above product usage attributes.

[0066] In some embodiments, the execution subject may generate attribute anomaly standard information for the product usage attributes based on the product attribute information set, using the target attribute anomaly determination rule and the target attribute anomaly generation model. The attribute anomaly standard information may represent anomaly judgment criteria for abnormal situations corresponding to the product usage attributes. That is, the attribute anomaly standard information represents when the product usage attributes are abnormal and when they are not abnormal.

[0067] As an example, first, the execution subject may input the product attribute information set into the target attribute anomaly determination rule and the target attribute anomaly generation model to obtain the first anomaly information and the second anomaly information. Then, the average value of the anomaly between the first anomaly information and the second anomaly information is determined as the attribute anomaly standard value. Finally, the attribute anomaly standard judgment information corresponding to the attribute anomaly standard value is generated as the attribute anomaly standard information.

[0068] In some optional implementations of some embodiments, the execution subject may generate attribute anomaly standard information for the product usage attributes according to the product attribute information set, using the target attribute anomaly determination rule and the target attribute anomaly generation model, including the following steps:

[0069] In the first step, in response to determining that the comprehensive accuracy corresponding to the target collocation combination corresponding to the target attribute anomaly determination rule and the target attribute anomaly generation model is not the highest collocation combination in the collocation combination set, wait for the computing surplus resources corresponding to the graphics processor cluster to meet the resource requirements of the collocation combination with the highest comprehensive accuracy. The comprehensive accuracy rate may be the accuracy rate between the target attribute anomaly determination rule and the target attribute anomaly generation model included in the target collocation combination. That is, the comprehensive accuracy rate is the sum of the accuracy rate corresponding to the target attribute anomaly determination rule and the accuracy rate corresponding to the target attribute anomaly generation model.

[0070] In the second step, in response to the determination of satisfaction, based on the product attribute information set, using the remaining computing resources corresponding to the graphics processor cluster, and using the highest matching combination, generate alternative attribute abnormality standard information for the product usage attributes.

[0071] As an example, the execution entity may utilize computing surplus resources to control the input of the product attribute information set into the attribute anomaly determination rules and attribute anomaly generation model included in the highest matching combination to generate candidate attribute anomaly standard information for the product usage attributes.

[0072] And the above method also includes:

[0073] The first step is to generate a second standard information portrait corresponding to the obtained at least one candidate attribute abnormal standard information, wherein the second standard information portrait and the object portrait have the same tree structure.

[0074] The second step is to store the second standard information portrait and the set of the removed object portraits in the target storage terminal to replace the first standard information portrait and the set of the removed object portraits.

[0075] In some optional implementations of some embodiments, the execution subject may generate attribute anomaly standard information for the product usage attributes according to the product attribute information set, using the target attribute anomaly determination rule and the target attribute anomaly generation model, including the following steps:

[0076] The first step is to determine the rule accuracy corresponding to the target attribute anomaly determination rule and the model accuracy corresponding to the target attribute anomaly generation model, wherein both the rule accuracy and the model accuracy are values ​​between 0 and 1.

[0077] The second step is to enhance the information of each product attribute information in the above product attribute information set to obtain an enhanced product attribute information set.

[0078] In the third step, in response to determining that the model accuracy is higher than or equal to the rule accuracy, the target attribute anomaly determination rule is used to determine the first anomaly information corresponding to each product attribute information in the enhanced product attribute information set, and obtain the first anomaly information set. The first anomaly information can represent that the attribute content corresponding to the product attribute information is anomaly content. In practice, the first anomaly information can be information in numerical form or in label form.

[0079] The fourth step is to bind the first exception information in the first exception information set and the product attribute information in the product attribute information set accordingly to generate binding information and obtain a binding information set.

[0080] In the fifth step, the binding information set and the target attribute abnormal information generation model are sent to the target node in the GPU cluster, so that the target node performs the following steps:

[0081] The first step is to obtain the target attribute abnormal information generation model and the above binding information set.

[0082] In the second step, according to the remaining computing resources corresponding to the target node, at least one similar model corresponding to the target attribute abnormal information generation model is obtained from the model repository. The sum of application resources corresponding to the at least one similar model is less than or equal to the remaining computing resources. The similar model can be a neural network model with similar model accuracy and network structure to the target attribute abnormal information generation model.

[0083] The third sub-step is, for each similarity model in the at least one similarity model, inputting each binding information in the binding information set into the similarity model to generate second abnormal information and obtain a second abnormal information set.

[0084] The fourth sub-step is to send the obtained at least one second abnormal information set to the generation node corresponding to the attribute abnormality standard information.

[0085] In the sixth step, each binding information in the binding information set is input into the target attribute anomaly generation model to generate third anomaly information and obtain the third anomaly information set.

[0086] Step 7: For each binding information in the above binding information set, perform the following determination steps:

[0087] Sub-step 1: determine the third exception information and at least one second exception information corresponding to the above binding information.

[0088] Sub-step 2, determining the comprehensive abnormality information for the third abnormality information and the at least one second abnormality information according to at least one model weight corresponding to at least one similar model and the model weight corresponding to the target attribute abnormality information generation model. Wherein, the model weight in at least one model weight has a one-to-one correspondence with the similar model in at least one similar model. The model weight can characterize the importance of the model output content.

[0089] As an example, the above-mentioned execution entity can perform weighted processing on the above-mentioned third abnormality information and at least one second abnormality information according to at least one model weight corresponding to at least one similarity model and the model weight corresponding to the target attribute abnormality information generation model to generate comprehensive abnormality information.

[0090] In the eighth step, comprehensive abnormality information representing the existence of abnormalities is screened out from the obtained comprehensive abnormality information set as target comprehensive abnormality information, and at least one target comprehensive abnormality information is obtained.

[0091] The ninth step is to determine the attribute information range corresponding to the at least one target comprehensive abnormal information as the target attribute information range.

[0092] As an example, the execution subject may determine at least one product attribute information corresponding to at least one target comprehensive abnormality information as at least one target product attribute information, and then determine an attribute information range corresponding to at least one target product attribute information as the target attribute information range.

[0093] The tenth step is to set attribute abnormality standard information that characterizes attribute information that is within the above target attribute information range as abnormal information and that is not within the above target attribute information range as non-abnormal information.

[0094] In the eleventh step, in response to determining that the model accuracy is less than the rule accuracy, the target attribute anomaly generation model is used to determine the fourth anomaly information corresponding to each product attribute information in the enhanced product attribute information set to obtain a fourth anomaly information set.

[0095] Step 12: For each product attribute information in the product attribute information set, perform the following generation steps:

[0096] Sub-step 1, determining the first abnormal information and the fourth abnormal information corresponding to the above-mentioned product attribute information as the first target abnormal information and the fourth target abnormal information respectively.

[0097] Sub-step 2, in response to determining that both the first target abnormality information and the fourth target abnormality information indicate the presence of an abnormality, generating judgment information indicating that the above-mentioned product attribute information indicates the presence of an abnormality.

[0098] Sub-step 3, in response to determining that the first target abnormal information and the fourth target abnormal information have abnormal information indicating that there is no abnormality, generating judgment information indicating that there is no abnormality in the above-mentioned product attribute information.

[0099] In the thirteenth step, at least one product attribute information indicating the existence of an abnormality is determined based on the obtained judgment information as at least one abnormal product attribute information.

[0100] The fourteenth step is to determine the attribute information range corresponding to the at least one abnormal product attribute information as the abnormal attribute information range.

[0101] Step 15: Generate attribute abnormality standard information based on the above-mentioned abnormal attribute information range.

[0102] The above "in some optional implementations of some embodiments", as one of the inventive points, solves another technical problem of the present disclosure, "the abnormal information generation resources corresponding to the target attribute abnormality determination rule and the above target attribute abnormality generation model cannot be fully utilized, resulting in the inaccurate generation of attribute abnormality standard information". Based on this, the present disclosure first determines the generation method of the corresponding attribute abnormality standard information by determining the respective accuracies corresponding to the target attribute abnormality determination rule and the target attribute abnormality generation model. Then, on the basis that the corresponding accuracy of the target attribute abnormality determination rule is higher than or equal to the target attribute abnormality generation model, the abnormal information set (i.e., the first abnormal information set) corresponding to the enhanced product attribute information set after the attribute information is enhanced is determined by first using the target attribute abnormality determination rule. Then, considering that the resources of the node currently generating the attribute abnormality standard information are limited, at least one second abnormal information set corresponding to the binding information set generated based on the first abnormal information set is determined at the target node through at least one similar node to further supplement the information corresponding to the abnormal information. Based on this, the comprehensive abnormal information corresponding to each binding information can be accurately generated. On the basis of determining to generate at least one comprehensive abnormal information, the corresponding target attribute information range is determined by screening the comprehensive abnormal information that is abnormal. On the basis that the corresponding accuracy of the target attribute anomaly determination rule is less than that of the target attribute anomaly generation model, the judgment information of whether the product attribute information has anomalies can be determined by comprehensively considering the abnormal information corresponding to the output of the target attribute anomaly determination rule and the abnormal information corresponding to the output of the target attribute anomaly generation model. Finally, based on the obtained judgment information set, the range of abnormal attribute information can be accurately determined to further determine the attribute abnormality standard information.

[0103] Step 105: Generate a first standard information portrait corresponding to at least one attribute abnormality standard information obtained.

[0104] In some embodiments, the execution subject may generate a first standard information portrait corresponding to at least one attribute abnormality standard information obtained. The first standard information portrait and the object portrait are in the same tree structure. The first standard portrait may represent the attribute abnormality judgment standard corresponding to each product usage attribute under at least one product usage attribute.

[0105] Step 106: storing the first standard information portrait and the removed object portrait set in a target storage terminal.

[0106] In some embodiments, the execution subject may store the first standard information portrait and the removed object portrait set in a target storage terminal, wherein the target storage terminal may be a predetermined storage terminal.

[0107] In some optional implementations of some embodiments, after step 107, the steps further include:

[0108] In the first step, in response to determining that the object data representing the request for the target virtual product is received, the object data is preprocessed to obtain preprocessed object data. The data preprocessing may include but is not limited to one of the following: data fuzzy removal and data missing removal.

[0109] The second step is to generate a target object portrait with the same structural format as the removed object portrait based on the pre-processed object data. The third step is to match the content corresponding to the target object portrait with the standard content corresponding to the first standard information portrait to generate a matching result portrait.

[0110] The fourth step is to obtain the current attribute focus direction selected in the attribute focus direction set. Different attribute focus directions correspond to different attribute scoring portraits. The tree structure corresponding to the above attribute scoring portrait is the same as the tree structure corresponding to the first standard information portrait. The current attribute focus direction can represent the various attributes that are focused on monitoring. That is, the current attribute focus direction can include: the key attributes that are focused on monitoring. The current attribute focus direction can represent the various attributes that are focused on.

[0111] The fifth step is to determine the attribute scoring portrait corresponding to the current attribute emphasis direction as the first target attribute scoring portrait.

[0112] In the sixth step, the matching result portrait is scored according to the first target attribute scoring portrait to generate a first anomaly score.

[0113] As an example, first, the execution subject may score each attribute information in the matching result portrait according to the attribute anomaly standard information corresponding to each attribute in the first target attribute scoring portrait to generate each anomaly score. Then, each anomaly score is weighted to generate a first anomaly score.

[0114] In step 7, based on the first anomaly score and the first target attribute scoring portrait, a first product usage anomaly report corresponding to the target object is generated. The target object corresponds to the object data. The first product usage anomaly report may be an anomaly report of the target product usage.

[0115] In the eighth step, the first product usage abnormality report is sent in an encrypted form to the abnormality review terminal for further product usage abnormality review.

[0116] In some optional implementations of some embodiments, the steps further include:

[0117] In the first step, in response to determining that the selected attribute emphasis direction does not exist in the above-mentioned attribute emphasis direction set, the attribute portrait corresponding to the above-mentioned first standard information portrait is displayed on the target attribute screening interface, so that the target personnel can manually allocate the combination of each attribute on the above-mentioned target attribute screening interface. Among them, the attribute portrait is a portrait of each attribute corresponding to the tree structure. Among them, the allocation method of the combination of each attribute is the component dragging method in the interface. The target attribute screening interface can be an interface for displaying attribute portraits for attribute screening. Unimportant attributes in the attribute portrait can be removed by dragging components.

[0118] The second step is to obtain the attribute combination and combination naming information selected by the target person in the target attribute screening interface, wherein the combination naming information may be a naming identifier corresponding to the attribute combination.

[0119] The third step is to determine the attribute scoring portrait corresponding to the above attribute combination as the second target attribute scoring portrait.

[0120] The fourth step is to score the matching result portrait according to the second target attribute scoring portrait to generate a second anomaly score. The details will not be repeated here, please refer to the determination of the first anomaly score.

[0121] The fifth step is to generate a second product usage anomaly report corresponding to the target object based on the second anomaly score and the second target attribute scoring portrait.

[0122] The sixth step is to send the above-mentioned second product usage abnormality report in an encrypted form to the abnormality review terminal for further product usage abnormality review.

[0123] In the seventh step, the attribute combination and the combination naming information are added to the attribute emphasis direction set in the form of a key-value pair for storage.

[0124] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the virtual product standard portrait storage method of some embodiments of the present disclosure, under the condition of adapting to the current computing surplus resources, the standard information portrait for abnormal determination of the target virtual product can be efficiently and accurately generated. Specifically, the reason why the relevant abnormal determination is not accurate enough and is not generated in a timely manner is that the product usage data set is often large in data volume, resulting in a large amount of computing time wasted by directly inputting it into the multi-attribute abnormal standard information output model, and allowing the multi-attribute abnormal standard information output model to learn more redundant features. And in the case where the multi-attribute abnormal standard information output model requires a large amount of computing resources, if the terminal cannot achieve the effective supply of computing resources, the output is interrupted and the generation of the abnormal determination standard is affected. Based on this, the virtual product standard portrait storage method of some embodiments of the present disclosure, first, obtains the product usage data set of the target virtual product within a predetermined time period, so as to be used as the basic data set for the abnormal determination of each product usage attribute. Based on this, the accurate generation of the virtual product standard portrait can be achieved. Then, for each product usage object in the product usage object set, according to the product usage data corresponding to the above product usage object, the object portrait corresponding to the above product usage object can be accurately generated. Through the object portrait, it is convenient to determine the object portrait of each user object corresponding to the target virtual product, so as to facilitate the query of the subsequent product use object related information. Then, the product use attribute set corresponding to the obtained object portrait set is subjected to attribute and attribute information clipping, and at least one product use attribute and the removed object portrait set are obtained to determine at least one product use attribute for key monitoring and the removed object portrait set including the key product use attribute information. Then, for each product use attribute in the above-mentioned at least one product use attribute, a distributed graphics processor cluster is used to perform a generation step: the first step is to determine the product attribute information set corresponding to the above-mentioned product use attribute in the above-mentioned object portrait set, so as to facilitate the generation of attribute abnormality standard information corresponding to the subsequent product use attribute. In addition, the use of a distributed graphics processor cluster can effectively provide computing resources that can be used for calculation. The second step is to obtain the attribute abnormality determination rule set and the attribute abnormality generation model set corresponding to the above-mentioned product use attribute, so as to jointly determine the abnormality judgment standard from the perspective of rules and the perspective of neural network models, which can greatly improve the accuracy of judgment. The third step is to select target attribute anomaly determination rules and target attribute anomaly generation models from the attribute anomaly determination rule set and attribute anomaly generation model set respectively according to the computing surplus resources corresponding to the above-mentioned graphics processor cluster, so as to effectively select rules and models for subsequent real-time generation of attribute anomaly standard information under the current computing resources. The fourth step is to accurately generate attribute anomaly standard information for the usage attributes of the above-mentioned product using the above-mentioned target attribute anomaly determination rules and the above-mentioned target attribute anomaly generation model according to the above-mentioned product attribute information set.Further, a first standard information portrait corresponding to at least one attribute abnormality standard information is generated, wherein the first standard information portrait and the object portrait are in the same tree structure. Here, in the form of the first standard information portrait, the efficiency of abnormality determination can be improved and the determination time can be saved in the subsequent abnormality determination. Finally, the first standard information portrait and the above-mentioned removed object portrait set are stored in the target storage terminal to facilitate the execution of subsequent abnormality determination and the regular update of the first standard information portrait.

[0125] Further references Figure 2 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a virtual product standard image storage device. These device embodiments are similar to Figure 1 Corresponding to the method embodiments shown, the virtual product standard portrait storage device can be specifically applied to various electronic devices.

[0126] like Figure 2 As shown, a virtual product standard portrait storage device 200 includes: an acquisition unit 201, a first generation unit 202, a cropping unit 203, an execution unit 204, a second generation unit 205 and a storage unit 206. The acquisition unit 201 is configured to acquire a product usage data set of a target virtual product within a predetermined time period; the first generation unit 202 is configured to generate, for each product usage object in a product usage object set, an object portrait corresponding to the product usage object according to the product usage data corresponding to the product usage object; the cropping unit 203 is configured to perform attribute and attribute information cropping on the product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and a removed object portrait set; the execution unit 204 is configured to execute the generation steps for each product usage attribute in the at least one product usage attribute using a distributed graphics processor cluster: determining a product attribute information set corresponding to the product usage attribute in the object portrait set; acquiring the above The product usage attributes correspond to the pre-stored attribute anomaly determination rule set and attribute anomaly generation model set; according to the computing surplus resources corresponding to the above-mentioned graphics processor cluster, the target attribute anomaly determination rule and the target attribute anomaly generation model are respectively screened out from the above-mentioned attribute anomaly determination rule set and the attribute anomaly generation model set; the attribute anomaly standard information for the above-mentioned product usage attributes is generated by using the above-mentioned target attribute anomaly determination rule and the above-mentioned target attribute anomaly generation model; the second generation unit 205 is configured to generate a first standard information portrait corresponding to at least one attribute anomaly standard information obtained, wherein the above-mentioned first standard information portrait and the object portrait are in the same tree structure; the storage unit 206 is configured to store the above-mentioned first standard information portrait and the above-mentioned removed object portrait set in the target storage terminal.

[0127] It can be understood that the units recorded in the virtual product standard image storage device 200 and the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the virtual product standard portrait storage device 200 and the units contained therein, and will not be repeated here.

[0128] Reference below Figure 3 , which shows a structural schematic diagram of an electronic device (eg, an electronic device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0129] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0130] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0131] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the ROM 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0132] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0133] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0134] The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist independently without being installed in the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains a product usage data set of the target virtual product within a predetermined time period; for each product usage object in the product usage object set, generates an object portrait corresponding to the above-mentioned product usage object according to the product usage data corresponding to the above-mentioned product usage object; performs attribute and attribute information clipping on the product usage attribute set corresponding to the obtained object portrait set, and obtains at least one product usage attribute and a post-removal object portrait set. For each product usage attribute in the above-mentioned at least one product usage attribute, use a distributed graphics processor cluster to execute the generation step: determine the product attribute information corresponding to the above-mentioned product usage attribute in the above-mentioned object portrait set set; obtain the pre-stored attribute anomaly determination rule set and attribute anomaly generation model set corresponding to the above-mentioned product usage attributes; according to the computing surplus resources corresponding to the above-mentioned graphics processor cluster, respectively filter out the target attribute anomaly determination rule and the target attribute anomaly generation model from the above-mentioned attribute anomaly determination rule set and the attribute anomaly generation model set; according to the above-mentioned product attribute information set, use the above-mentioned target attribute anomaly determination rule and the above-mentioned target attribute anomaly generation model to generate attribute anomaly standard information for the above-mentioned product usage attributes; generate a first standard information portrait corresponding to at least one attribute anomaly standard information obtained, wherein the above-mentioned first standard information portrait and the object portrait have the same tree structure; store the above-mentioned first standard information portrait and the above-mentioned removed object portrait set in the target storage terminal.

[0135] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0136] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0137] The units described in some embodiments of the present disclosure may be implemented by software or by hardware. The described units may also be set in a processor, for example, it may be described as: a processor includes an acquisition unit, a first generation unit, a clipping unit, an execution unit, a second generation unit and a storage unit. The names of these units do not constitute a limitation on the units themselves in some cases. For example, the acquisition unit may also be described as "a unit for acquiring a product usage data set of a target virtual product within a predetermined time period".

[0138] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0139] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A method for storing a virtual product standard image, comprising: Acquire a product usage dataset of a target virtual product within a predetermined time period; For each product usage object in the product usage object set, generating an object portrait corresponding to the product usage object according to the product usage data corresponding to the product usage object; Performing attribute and attribute information clipping on the product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and the object portrait set after the attribute is removed; For each product usage attribute of the at least one product usage attribute, using a distributed graphics processor cluster, a generating step is performed: Determining a product attribute information set corresponding to the product usage attribute in the object portrait set; Obtaining a pre-stored attribute anomaly determination rule set and an attribute anomaly generation model set corresponding to the product usage attributes; According to the remaining computing resources corresponding to the graphics processor cluster, respectively selecting a target attribute anomaly determination rule and a target attribute anomaly generation model from the attribute anomaly determination rule set and the attribute anomaly generation model set; According to the product attribute information set, using the target attribute anomaly determination rule and the target attribute anomaly generation model, attribute anomaly standard information for the product usage attributes is generated; Generate a first standard information portrait corresponding to the obtained at least one attribute abnormality standard information, wherein the first standard information portrait and the object portrait have the same tree structure; The first standard information portrait and the removed object portrait set are stored in a target storage terminal.

2. The method according to claim 1, wherein: The method of performing attribute and attribute information clipping on the product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and the object portrait set after the attribute is removed includes: Generate attribute prediction valid information corresponding to each product usage attribute in the product usage attribute set to obtain an attribute prediction valid information set, wherein the attribute prediction valid information is a valid value representing the importance of the attribute and the strength of the attribute prediction capability; Remove product usage attributes whose corresponding attribute prediction effective information is less than a target value from the product usage attribute set to obtain at least one product usage attribute; For each object portrait in the object portrait set, attribute information corresponding to the at least one product usage attribute is removed from the object portrait to obtain a post-removal object portrait.

3. The method according to claim 2, wherein: The method further comprises: In response to determining that the object data representing the request for the target virtual product is received, preprocessing the object data to obtain preprocessed object data; Generating a target object portrait having the same structural format as the corresponding post-removal object portrait according to the pre-processed object data; Matching the content corresponding to the target object portrait with the standard content corresponding to the first standard information portrait to generate a matching result portrait; Obtaining a current attribute emphasis direction selected from the attribute emphasis direction set, wherein different attribute emphasis directions correspond to different attribute scoring portraits, and a tree structure corresponding to the attribute scoring portrait is the same as a tree structure corresponding to the first standard information portrait; Determine the attribute scoring portrait corresponding to the current attribute emphasis direction as the first target attribute scoring portrait; According to the first target attribute scoring portrait, scoring the matching result portrait to generate a first anomaly score; generating a first product usage anomaly report corresponding to a target object according to the first anomaly score and the first target attribute scoring portrait, wherein the target object corresponds to the object data; The first product usage exception report is sent in an encrypted form to the exception review terminal for further product usage exception review.

4. The method according to claim 3, wherein: The method further comprises: In response to determining that the selected attribute emphasis direction does not exist in the attribute emphasis direction set, an attribute portrait corresponding to the first standard information portrait is displayed on a target attribute screening interface, so that the target personnel can manually allocate the combination of each attribute on the target attribute screening interface, wherein the attribute portrait is a portrait corresponding to each attribute in a tree structure, wherein the allocation method of the combination of each attribute is a component dragging method in the interface; Acquire the attribute combination and combination naming information selected by the target person in the target attribute screening interface; Determine an attribute scoring portrait corresponding to the attribute combination as a second target attribute scoring portrait; According to the second target attribute scoring portrait, scoring the matching result portrait to generate a second anomaly score; Generate a second product usage anomaly report corresponding to the target object according to the second anomaly score and the second target attribute scoring portrait; Sending the second product use abnormality report in encrypted form to the abnormality review terminal for further product use abnormality review; The attribute combination and the combination naming information are added to the attribute emphasis direction set in the form of a key-value pair for storage.

5. The method according to claim 1, wherein: The step of selecting a target attribute anomaly determination rule and a target attribute anomaly generation model from the attribute anomaly determination rule set and the attribute anomaly generation model set respectively according to the computing surplus resources corresponding to the graphics processor cluster includes: Determining a computational remaining resource sequence of the graphics processor cluster within a target future time period; Determine the attribute division resource sequence corresponding to the product usage attribute from the calculated remaining resource sequence, wherein each product usage attribute has a corresponding attribute division resource sequence; Determine a collocation combination set between the attribute anomaly determination rule set and the attribute anomaly generation model set, wherein each collocation combination includes: an attribute anomaly determination rule and an attribute anomaly generation model; For each collocation combination in the collocation combination set, the following generation steps are performed: Determine the estimated consumption time and estimated consumption resources corresponding to the attribute anomaly determination rule and the attribute anomaly generation model in the collocation combination; Determine whether the attribute-divided resource sequence meets the consumption requirements corresponding to the estimated consumption duration and the estimated consumption resources, and obtain satisfaction information; Filtering out satisfying information representing consumption requirements corresponding to the estimated consumption duration and the estimated consumption resources from the obtained satisfying information set, and obtaining at least one satisfying information; A combination with the highest comprehensive accuracy and the shortest estimated consumption time is selected from at least one combination as a target combination, wherein the target combination includes: a target attribute anomaly determination rule and a target attribute anomaly generation model, and the at least one satisfying information has a one-to-one correspondence with the at least one combination.

6. The method according to claim 5, wherein: The determining of the attribute-divided resource sequence corresponding to the product usage attribute from the calculated remaining resource sequence includes: Determine an attribute weight corresponding to each product usage attribute of the at least one product usage attribute to obtain at least one attribute weight; According to the at least one attribute weight, each computing remaining resource in the computing remaining resource sequence is divided accordingly to generate usable computing resources at a corresponding time according to the product usage attribute, and obtain at least one attribute-divided resource sequence; The attribute division resource sequence corresponding to the product usage attribute is screened out from the at least one attribute division resource sequence.

7. The method according to claim 5, wherein: The generating of attribute abnormality standard information for the product usage attributes by using the target attribute abnormality determination rule and the target attribute abnormality generation model according to the product attribute information set includes: In response to determining that the comprehensive accuracy corresponding to the target collocation combination corresponding to the target attribute anomaly determination rule and the target attribute anomaly generation model is not the highest collocation combination in the collocation combination set, waiting for the computing remaining resources corresponding to the graphics processor cluster to meet the resource demand of the collocation combination with the highest comprehensive accuracy; In response to determining that the condition is satisfied, generating candidate attribute abnormality standard information for the product usage attribute based on the product attribute information set, using the computing surplus resources corresponding to the graphics processor cluster, and using the highest matching combination; And the method further comprises: Generate a second standard information portrait corresponding to the obtained at least one candidate attribute abnormal standard information, wherein the second standard information portrait and the object portrait have the same tree structure; The second standard information portrait and the set of removed object portraits are stored in the target storage terminal to replace the first standard information portrait and the set of removed object portraits.

8. A virtual product standard image storage device, comprising: an acquisition unit, configured to acquire a product usage data set of a target virtual product within a predetermined time period; A first generating unit is configured to generate, for each product usage object in the product usage object set, an object portrait corresponding to the product usage object according to the product usage data corresponding to the product usage object; A clipping unit configured to clip attributes and attribute information of a product usage attribute set corresponding to the obtained object portrait set to obtain at least one product usage attribute and a post-removal object portrait set; The execution unit is configured to, for each product usage attribute of the at least one product usage attribute, use a distributed graphics processor cluster to perform a generation step: determining a product attribute information set corresponding to the product usage attribute in the object portrait set; Acquire a pre-stored attribute anomaly determination rule set and an attribute anomaly generation model set corresponding to the product usage attributes; select a target attribute anomaly determination rule and a target attribute anomaly generation model from the attribute anomaly determination rule set and the attribute anomaly generation model set according to the remaining computing resources corresponding to the graphics processor cluster; generate attribute anomaly standard information for the product usage attributes using the target attribute anomaly determination rule and the target attribute anomaly generation model according to the product attribute information set; A second generating unit is configured to generate a first standard information portrait corresponding to the obtained at least one attribute abnormality standard information, wherein the first standard information portrait and the object portrait have the same tree structure; The storage unit is configured to store the first standard information portrait and the removed object portrait set in a target storage terminal.

9. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images