CN114625784A - Big data analysis method and system applied to digital space interaction - Google Patents

Abstract

The big data analysis method and the system applied to digital space interaction of the embodiment of the disclosure can effectively expand the determination interval of the derived digital service items by determining the derived digital service items of the target digital service items in the undetermined digital service item set bound with the multiple types of item detail characteristics and guarantee the precision and the credibility of the determined derived digital service items to a certain extent, so that the derived digital service items marked for the target digital service items of the digital virtual interaction equipment can be determined in the coarse-screening digital service items with the multiple types of item detail characteristics, and by the design, the derived digital service items associated with the target digital service items can be completely and accurately obtained as much as possible, and corresponding service interaction software and hardware measures are deployed in advance according to the derived digital service items to improve the service interaction intelligence degree of the subsequent digital virtual interaction equipment, and the service interaction waiting time of the digital virtual interaction equipment is reduced.

Description

Big data analysis method and system applied to digital space interaction

Technical Field

The present disclosure relates to the field of digitization and big data technologies, and in particular, to a big data analysis method and system applied to digital space interaction.

Background

The Digital part of Digital transformation involves a shift from existing systems and infrastructures to modern platforms and software delivered as a service through the cloud (rather than traditional desktop applications). Digital transformation is a complex process, but its potential is broad and diverse. At present, the leading-edge technologies such as online payment, remote office, cloud games, block chains, virtual spaces and the like can be understood as the actual landing field of digital transformation. It is clear that digital transformation is constantly changing our lives.

In the digital era, the interaction between a user and a server is the key for the operation of digital services, so the intelligent degree of the interaction between the user and the server restricts the development of digitization. Even though actual digital services such as digital payment, digital office, digital space, etc. have come into use, the improvement of the degree of intelligence for the above digital services is still in the way. .

Disclosure of Invention

An object of the present disclosure is to provide a big data analysis method and system applied to digital space interaction, which can obtain a derivative digital service item associated with a target digital service item as completely and accurately as possible, so as to deploy corresponding service interaction software and hardware measures in advance according to the derivative digital service item, thereby improving the service interaction intelligentization degree of subsequent digital virtual interaction devices, and reducing the service interaction waiting time of the digital virtual interaction devices.

The technical scheme of the disclosure is realized by at least some of the following embodiments.

A big data analysis method applied to digital space interaction is applied to a big data analysis system, and the method at least comprises the following steps: determining a coarse screen type digital service project through a target digital service project; obtaining session activity description content based on the target digital service item and the coarse screening type digital service item, and determining a pending digital service item set according to the session activity description content; and obtaining a derivative digital service item of the target digital service item according to the pending digital service item set.

The method is applied to the embodiment, the derived digital service items of the target digital service items are determined in the undetermined digital service item set bound with the multi-class item detail characteristics, the determination interval of the derived digital service items can be effectively expanded, the accuracy and the credibility of the determined derived digital service items are guaranteed to a certain extent, so that the derived digital service items can be determined for the target digital service items marked by the digital virtual interaction equipment in the coarse-screening digital service items with the multi-class item detail characteristics, the derived digital service items associated with the target digital service items can be completely and accurately obtained as far as possible by the design, and corresponding service interaction software and hardware measures are deployed according to the derived digital service items in advance to improve the service interaction intelligentization degree of the subsequent digital virtual interaction equipment, and the service interaction waiting time of the digital virtual interaction equipment is reduced.

Under an independently implementable design concept, the determining a coarse screening type digital service item through a target digital service item comprises: positioning a target digital service item marked by a digital virtual interactive device, and determining a plurality of coarse-screen digital service items with upstream and downstream constraint conditions existing in the target digital service item; wherein each coarse screening type digital service item is bound with at least one type of the upstream and downstream constraint condition, and the plurality of coarse screening type digital service items are bound with a plurality of types of the upstream and downstream constraint conditions; the obtaining of session activity description content based on the target digital service item and the coarse screening type digital service item, and determining a set of pending digital service items according to the session activity description content, includes: mining the session activity description content of the target digital service project to obtain a first session activity description content, and mining the session activity description content of each coarse screening type digital service project to obtain a second session activity description content; determining a plurality of sets of pending digital service items in conjunction with the first session activity description and the second session activity description; each pending digital service item set covers at least a part of coarse screening digital service items in the plurality of coarse screening digital service items, and each pending digital service item set is bound with at least one type of upstream and downstream constraint conditions; the obtaining of the derivative digital service item of the target digital service item according to the pending digital service item set includes: and screening a target undetermined digital service item set reaching a set index in the plurality of undetermined digital service item sets, and determining coarse-screening digital service items in the target undetermined digital service item set as derivative digital service items of the target digital service items.

Applied to this embodiment, since each set of pending digital service items binds at least one type of upstream and downstream constraint, different sets of pending digital service items correspond to different item detail characteristics (session activity description content). By determining the derivative digital service items of the target digital service items in the pending digital service item set bound with the detail characteristics of the multiple types of items, the determination interval of the derivative digital service items can be effectively expanded, the precision and the credibility of the determined derivative digital service items are guaranteed to a certain extent, so that the derivative digital service items can be determined for the target digital service items marked by the digital virtual interaction equipment in the coarse screening type digital service items with various item detail characteristics, the derivative digital service items associated with the target digital service items can be obtained as completely and accurately as possible, corresponding service interaction software and hardware measures are deployed in advance according to the derivative digital service projects, so that the service interaction intelligentization degree of subsequent digital virtual interaction equipment is improved, and the service interaction waiting time of the digital virtual interaction equipment is reduced.

Under an independently implementable design concept, said determining a number of sets of pending digital service items in conjunction with said first session activity description content and said second session activity description content, comprising: determining a basic digital service item set which covers a plurality of coarse screening digital service items and corresponds to each upstream and downstream constraint condition; determining a quantitative hit rating for each of the coarse-screening type digital service items in each of the basic digital service item sets in conjunction with the first session activity description content and the second session activity description content; and combining the quantitative hit evaluation to clean coarse-screening digital service items which do not reach the hit evaluation index in each basic digital service item set, and determining the plurality of undetermined digital service item sets according to each basic digital service item set which is cleaned.

The method and the device are applied to the embodiment, the thought of determining the plurality of undetermined digital service item sets according to the quantitative hit evaluation of each coarse-screen digital service item in each basic digital service item set can accurately and credibly determine the undetermined digital service item sets corresponding to different detailed information, and therefore derived digital service items can be determined for the digital virtual interaction device more efficiently.

Under an independently implementable design concept, said determining a quantitative hit rating for each of said coarse-screening type digital services in each of said set of base digital services in combination with said first session activity description content and said second session activity description content, comprising: determining a binary operation result between the first session activity description content and the second session activity description content to obtain a target session activity description content; and determining the quantitative hit evaluation of each coarse-screening type digital service item in each basic digital service item set in combination with the target session activity description content.

The method is applied to the embodiment, and by combining the ideas, the quantitative hit evaluation of the coarse-screen type digital service items in each basic digital service item set can be determined quickly, accurately and credibly, and errors generated in the determination of the quantitative hit evaluation are avoided.

Under an independently implementable design concept, the cleaning, in combination with the quantitative hit evaluation, of the coarse-screening-type digitalized service items in each of the basic digitalized service item sets that do not reach the hit evaluation index includes: determining a target coarse screening type digital service item which is not greater than a first judgment value in the quantitative hit evaluation in each basic digital service item set; and cleaning the target coarse screening type digital service item in a plurality of coarse screening type digital service items of the basic digital service item set.

The method is applied to the embodiment, and the thought of cleaning the coarse-screening type digital service items in the basic digital service item set according to the comparison condition is realized by comparing the quantitative hit evaluation with the first judgment value, so that the coarse-screening type digital service items which do not have a pairing relationship with the upstream and downstream constraint conditions of the basic digital service item set can be cleaned accurately in time.

Under an independently implementable design concept, the step of screening the plurality of pending digital service item sets for a target pending digital service item set that meets a set index includes: respectively determining the overall description content of each to-be-determined digital service item set by combining the session activity description content of each coarse-screen digital service item in each to-be-determined digital service item set; obtaining interactive behavior description content of the digital virtual interactive device; and determining the confidence coefficient of each pending digital service item set by combining the overall description content and the interactive behavior description content, and determining the pending digital service item set reaching a set index in the plurality of determined confidence coefficients as the target pending digital service item set.

The method is applied to the embodiment, by combining the interactive behavior description content and the session activity description content of each pending digital service item set and determining the thought of the last pending digital service item set in a plurality of pending digital service item sets, the derived digital service items which are adapted to the interactive preference of the digital virtual interactive device as much as possible can be positioned, so that the differentiated marking preference of different digital virtual interactive devices can be achieved.

Under an independently implementable design concept, the mining of the session activity description content of the target digital service item to obtain a first session activity description content, and the mining of the session activity description content of each coarse-screening type digital service item to obtain a second session activity description content includes: obtaining service item operation data of the target digital service item and service item operation data of each coarse screening type digital service item; determining first service item operation data matched with each upstream and downstream constraint condition in the service item operation data of the target digital service item, and determining second service item operation data matched with the upstream and downstream constraint condition in the service item operation data of each coarse screening type digital service item; mining the session activity description content of the first service project operation data to obtain the first session activity description content; and mining the session activity description content of the second service item operation data to obtain the second session activity description content.

When the number of the upstream and downstream constraints is several, a plurality of first session activity description contents and a plurality of second session activity description contents can be mined according to the upstream and downstream constraints, wherein each first session activity description content and each second session activity description content correspond to different upstream and downstream constraints. Further, the session activity description contents mined by the above mining ideas can be combined to determine a set of pending digital service projects corresponding to a plurality of upstream and downstream constraints (or project detail characteristics). By combining the ideas, the session activity description contents bound with the various different project detail characteristics can be obtained, and the pending digital service project set corresponding to the different project detail characteristics can be accurately and reliably determined.

Under an independently implementable design concept, said determining a number of sets of pending digital service items in conjunction with said first session activity description content and said second session activity description content, comprising: calling a first AI network model to process the first session activity description content and the second session activity description content to obtain a plurality of pending digital service item sets; the screening of the target undetermined digital service item sets which reach the set index in the plurality of undetermined digital service item sets comprises the following steps: and calling a second AI network model to screen a target undetermined digital service item set which reaches a set index in the plurality of undetermined digital service item sets.

The method is applied to the embodiment, by setting two AI network models (in other words, a first AI network model and a second AI network model), calling the first AI network model to determine a plurality of undetermined digital service item sets, and calling the second AI network model to screen the thought of the target undetermined digital service item set reaching the set index in the plurality of undetermined digital service item sets, the determination interval of the derived digital service item can be effectively expanded, so that the derived digital service item which is adapted to the interaction preference of the digital virtual interaction device as much as possible is determined.

Under the design idea that can be independently implemented, the method further comprises the following steps: with the help of a target debugging strategy, debugging a first basic AI network model to be debugged to obtain the first AI network model, and debugging a second basic AI network model to be debugged to obtain the second AI network model, wherein the target debugging strategy comprises: a debug policy with authentication examples and a reactivation debug policy.

The method is applied to the embodiment, the first basic AI network model and the second basic AI network model are debugged through the debugging strategy and the re-excitation debugging strategy with the authentication examples to respectively obtain the first AI network model and the second AI network model, so that the model operation quality of the first AI network model and the second AI network model can be improved, and the derived digital service items which are as accurate and credible as possible are obtained.

Under an independently implementable design concept, the target debugging strategy comprises a re-excitation debugging strategy; the debugging of the first basic AI network model to be debugged by means of the target debugging strategy comprises the following steps: determining a first debugging paradigm; the first debugging example covers a plurality of digital service items to be debugged and a plurality of target derivative digital service items of each digital service item to be debugged; loading the first debugging example into a first basic AI network model to be debugged to obtain quantitative hit evaluation of each target derivative digital service item; determining a to-be-debugged digital service item set of the to-be-debugged digital service item by combining the quantitative hit evaluation of the target derivative digital service item; and determining a first model quality characteristic by combining the quantitative hit evaluation of the target derivative digital service item to be debugged in the to-be-debugged to the contribution index of the to-be-debugged first basic AI network model, and improving the model variable of the to-be-debugged first basic AI network model by combining the first model quality characteristic until the to-be-debugged first basic AI network model reaches a relive debugging index.

Under an independently implementable design concept, the target debugging strategy comprises a re-excitation debugging strategy; the debugging of the second basic AI network model to be debugged by means of the target debugging strategy comprises the following steps: loading the plurality of to-be-debugged digital service item sets to a second basic AI network model to be debugged to obtain quantitative hit evaluation of each to-be-debugged digital service item set; and determining a second model quality characteristic by combining the quantitative hit evaluation of each to-be-debugged digital service item set and the contribution index of the to-be-debugged second basic AI network model, and improving the model variable of the to-be-debugged second basic AI network model by combining the second model quality characteristic until the to-be-debugged second basic AI network model reaches a reliving debugging index.

Under an independently implementable design concept, the target debugging strategy comprises a debugging strategy with an authentication example; the debugging of the first basic AI network model to be debugged by means of the target debugging strategy comprises the following steps: determining a second debugging example, wherein the second debugging example covers a plurality of digital service items to be debugged and a plurality of real derivative digital service item sets of each digital service item to be debugged; and debugging the first basic AI network model to be debugged based on the authentication example by combining the second debugging example.

Under an independently implementable design concept, the target debugging strategy comprises a debugging strategy with an authentication example; the debugging of the second basic AI network model to be debugged by means of the target debugging strategy comprises the following steps: determining a real derivative digital service item set of the digital service item to be debugged; determining the digital service item common values of the real derivative digital service item set and each to-be-debugged digital service item set, and taking the to-be-debugged digital service item set corresponding to the digital service item common value which reaches the specified judgment value condition in the plurality of digital service item common values as a debugging keyword; and debugging a second basic AI network model to be debugged based on an authentication example through a plurality of to-be-debugged digital service item sets covering the debugging keywords.

The method and the device are applied to the embodiment, the undetermined digital service item set to be debugged which needs to be improved is determined by combining the common value, and different detail information of each undetermined digital service item set can be guaranteed, so that the richness of the undetermined digital service item sets is reserved.

For a traditional AI network model, a machine learning idea is generally used to find a quantization index (score) for each digital service item, and then the digital service item with the selected quantization index ahead or the quantization index greater than a certain decision value is taken as a derivative digital service item, and a labeled learning idea is used to improve the related functions of the network model. In order to improve the model operation quality of the AI network model, the embodiment of the disclosure debugs the AI network model by fusing a labeled learning idea (a debugging strategy with an authentication example) and a reengineering debugging strategy (reinforcement learning), and provides a big data analysis method applied to digital space interaction. Wherein the first AI network model and the second AI network model are cascaded.

Further, the first AI network model estimates a quantitative hit rating for each coarse-screening type digital service item in the respective set of basic digital service items based on the first session activity description content and the second session activity description content, and the second AI network model estimates a quantitative hit rating (i.e., a confidence coefficient) for each set of pending digital service items. Therefore, the processing processes of the first AI network model and the second AI network model are matched with the statistical morphology (probability distribution) in the reentering debugging strategy, and the processing time consumption of the first AI network model and the second AI network model is not prolonged along with the increase of the number of the marks.

A big data analytics system, comprising: a memory for storing an executable computer program, a processor for implementing the above method when executing the executable computer program stored in the memory.

A computer-readable storage medium, on which a computer program is stored which, when executed, performs the above-described method.

Drawings

FIG. 1 is a schematic diagram illustrating one communication configuration of a big data analytics system in which embodiments of the present disclosure may be implemented.

Fig. 2 is a flow diagram illustrating a big data analysis method applied to digital space interaction, in which an embodiment of the present disclosure may be implemented.

Fig. 3 is an architecture diagram illustrating an application environment of a big data analysis method applied to digital space interaction, in which an embodiment of the present disclosure may be implemented.

Detailed Description

The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. For the purpose of making the purpose, technical solutions and advantages of the present disclosure clearer, the present disclosure will be described in further detail with reference to the accompanying drawings, the described embodiments should not be construed as limiting the present disclosure, and all other embodiments obtained by a person of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present disclosure. In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing embodiments of the disclosure only and is not intended to be limiting of the disclosure.

Fig. 1 is a block diagram illustrating one communication configuration of a big data analyzing system 100 that can implement an embodiment of the present disclosure, the big data analyzing system 100 including a memory 101 for storing an executable computer program, and a processor 102 for implementing the big data analyzing method applied to digital space interaction in the embodiment of the present disclosure when executing the executable computer program stored in the memory 101.

Fig. 2 is a flowchart illustrating a big data analysis method applied to digital space interaction, which may implement an embodiment of the present disclosure, and the big data analysis method applied to digital space interaction may be implemented by the big data analysis system 100 shown in fig. 1, and further may include the technical solutions described in the following related steps.

The overall design idea of the embodiment of the disclosure is as follows: determining a coarse screen type digital service project through a target digital service project; obtaining session activity description content based on the target digital service item and the coarse screening type digital service item, and determining a pending digital service item set according to the session activity description content; and obtaining a derivative digital service item of the target digital service item according to the pending digital service item set.

It can be understood that by determining the derivative digital service items of the target digital service items in the undetermined digital service item set bound with the multiple types of item detail features, the determination interval of the derivative digital service items can be effectively expanded, and the precision and the credibility of the determined derivative digital service items can be guaranteed to a certain extent, so that the derivative digital service items can be determined for the target digital service items marked by the digital virtual interaction device in the coarse-screening digital service items with the multiple types of item detail features, and by the design, the derivative digital service items associated with the target digital service items can be completely and accurately obtained as much as possible, and therefore, corresponding service interaction software and hardware measures are deployed according to the derivative digital service items in advance to improve the service interaction intelligentization degree of subsequent digital virtual interaction devices, and the service interaction waiting time of the digital virtual interaction equipment is reduced.

The embodiment of the present disclosure is exemplarily illustrated by the following embodiments for the steps "determining a coarse-screening type digital service item through a target digital service item," obtaining a session activity description content based on the target digital service item and the coarse-screening type digital service item, and determining a pending digital service item set according to the session activity description content "and" obtaining a derived digital service item of the target digital service item according to the pending digital service item set "in the above design concept.

Step 101, positioning a target digital service item marked by a digital virtual interactive device, and determining a plurality of coarse-screen digital service items with the target digital service item and the upstream and downstream constraint conditions.

For the disclosed embodiment, each of the coarse-screening digital service item bindings has at least one type of the upstream and downstream constraint, and the plurality of coarse-screening digital service item bindings has a plurality of types of the upstream and downstream constraint. Further, a target digital service item marked by a digital virtual interaction device (such as an intelligent user terminal, a virtual service client, a digital service device and the like) has a relationship with an actual service type of the big data analysis method applied to digital space interaction. In some exemplary aspects, the target digital service item may be a set of project tasks covering specific office software, such as a project task including a WPS thread, based on the actual business type being digital office business. On the basis that the actual business type is cloud game business, the target digital service items can be any group of game functions marked by the digital virtual interaction equipment.

For the embodiment of the present disclosure, the above upstream and downstream constraints may be understood as corresponding situations, such as corresponding situations (association relations) between the target digital service item and several coarse-screening digital service items. For example, for digital office business, the corresponding situation may be understood as a digital service item that has a certain relation with the WPS thread in the coarse-screening digital service item, for example, the WPS thread itself, hyperlink software of the WPS thread, a cloud service item of the WPS thread, and the like. For the cloud game service, the corresponding situation can be understood as a game function keyword. For the embodiment of the present disclosure, for different target digital service items, different upstream and downstream constraints may be applied, and different numbers of coarse-screening digital service items may be applied.

And 103, mining the session activity description content of the target digital service item to obtain a first session activity description content, and mining the session activity description content of each coarse screening type digital service item to obtain a second session activity description content.

For the embodiment of the present disclosure, the session activity description content of the target digital service item may be mined according to the above upstream and downstream constraints, and the session activity description content of each coarse-screening type digital service item may be mined according to the above upstream and downstream constraints (which may be understood as a feature vector or a feature map, and is used to record the detailed features or key content of different digital service items).

Step 105, determining a plurality of sets of pending digital service items by combining the first session activity description content and the second session activity description content; each pending digital service item set covers at least a part of the coarse-screening digital service items in the plurality of coarse-screening digital service items, and each pending digital service item set is bound with at least one type of upstream and downstream constraint conditions.

For the embodiment of the present disclosure, each set of pending digital service items covers at least a part of the coarse-screening digital service items, and the coarse-screening digital service items included in different sets of pending digital service items may be consistent or may not be consistent. And binding the least one type of upstream and downstream constraint conditions aiming at each pending digital service item set, wherein the least one type of upstream and downstream constraint conditions corresponding to different pending digital service item sets are not completely consistent or not consistent.

Viewed from some exemplary perspective, for the cloud game service, several upstream and downstream constraints are: game type, pay or not, game developer. In this case, the several pending digital service item sets may include three pending digital service item sets, such as "game type" for the upstream and downstream constraint corresponding to the pending digital service item set item _ set _ a, "pay-or-not" for the upstream and downstream constraint corresponding to the pending digital service item set item _ set _ B, "game developer" and "pay-or-not" for the upstream and downstream constraint corresponding to the pending digital service item set item _ set _ C.

It can be understood that the coarse-screening type digital service items in each set of pending digital service items are coarse-screening type digital service items (for example, digital service items to be processed or paired) sampled from a plurality of coarse-screening type digital service items according to upstream and downstream constraint conditions corresponding to the set of pending digital service items.

And 107, screening a target undetermined digital service item set reaching a set index in the plurality of undetermined digital service item sets, and determining coarse-screening digital service items in the target undetermined digital service item set as derived digital service items of the target digital service items.

For the embodiment of the present disclosure, the confidence coefficient may be understood as a possibility that the coarse-screened digital service item in each set of pending digital service items is used as a derivative digital service item of the target digital service item and is issued to the digital virtual interaction device (in other words, a quantitative hit evaluation of the set of pending digital service items).

In the practical application process, the confidence coefficient of each pending digital service item set can be determined, and then the coarse-screening digital service items in the target pending digital service item set, which reach the set index in the determined confidence coefficients, are determined as the derivative digital service items of the target digital service items.

It can be understood that the number of the target pending digital service item sets may be one or several, and the number of the target pending digital service item sets is not limited in the embodiments of the present disclosure, and may be flexibly selected.

For the embodiment of the disclosure, first, a target digital service item marked by a digital virtual interactive device is located, and a plurality of coarse-screen digital service items having various upstream and downstream constraint conditions with the target digital service item are determined; furthermore, session activity description contents of the target digital service item and the coarse screening type digital service item can be respectively mined, and then a plurality of pending digital service item sets are determined according to the mined session activity description contents, wherein each pending digital service item set is bound with at least one type of upstream and downstream constraint conditions. Further, the coarse-screening digital service items in the pending digital service item sets (in other words, the target pending digital service item set) which reach the set index in the pending digital service item sets can be determined as the derived digital service items of the target digital service item.

In combination with the related art, in view of that each set of pending digital service items corresponds to at least one upstream and downstream constraint, different sets of pending digital service items correspond to different item detail characteristics. By determining the derivative digital service items of the target digital service items in the undetermined digital service item set bound with the multi-class item detail characteristics, the precision and the credibility of the determined derivative digital service items can be guaranteed to a certain extent, and thus the derivative digital service items can be determined for the target digital service items marked by the digital virtual interaction equipment in the coarse screening digital service items with the multi-class item detail characteristics.

Under some possible design ideas, the actual service types are described by taking cloud game services as examples. In combination with the related art, for the embodiment of the present disclosure, a target digital service item marked by a digital virtual interactive device is first located, and a number of coarse-screening digital service items having an upstream and downstream constraint with the target digital service item are determined.

For the cloud game service, the target digital service item can be understood as a game function concerned by the digital virtual interaction device, and the upstream and downstream constraint conditions can be understood as information such as game type, game running load and payment. The plurality of coarse screening type digital service items can be understood as coarse screening type game function items which are determined according to the upstream and downstream constraint conditions and are associated with the target digital service item.

Further to determining the target digital service item and the coarse-screening digital service items, the saliency descriptions (session activity description/feature vector/feature map) of the target digital service item can be mined, and the session activity description of each coarse-screening digital service item can be mined, where an exemplary mining process includes the following steps: and mining the significance description of the target digital service item according to the upstream and downstream constraint conditions to obtain first session activity description content, and mining the session activity description content of each coarse-screen type digital service item according to the upstream and downstream constraint conditions to obtain second session activity description content.

In an actual application process, the service item operation data of the target digital service item and the service item operation data of each coarse screen type digital service item can be obtained. Then, the first service item running data matched with each upstream and downstream constraint condition is determined in the service item running data of the target digital service item, and the second service item running data matched with the upstream and downstream constraint condition is determined in the service item running data of each coarse screening type digital service item. Further, the session activity description content of the first service project operation data can be mined to obtain the first session activity description content; and mining the session activity description content of the second service item operation data to obtain the second session activity description content.

It is understood that the service item operation data of the above target digital service item can be understood as a digital service item element (attribute) of the target digital service item, for example, for a cloud game service, the digital service item element may be: the payment of the target digital service item, the game running load, the game type, the game developer and other factors. Similarly, the service item operating data of the coarse-screening type digital service item and the digital service item element which can be understood as the coarse-screening type digital service item are not expanded.

When the number of the upstream and downstream constraints is several, a plurality of first session activity description contents and a plurality of second session activity description contents can be mined according to the upstream and downstream constraints, wherein each first session activity description content and each second session activity description content correspond to different upstream and downstream constraints. Further, the session activity description contents mined by the above mining ideas can be combined to determine a set of pending digital service projects corresponding to a plurality of upstream and downstream constraints (or project detail characteristics).

Further to determining the first session activity description and the second session activity description, a number of sets of pending digital service items can be determined in conjunction with the first session activity description and the second session activity description.

For some design ideas that can be implemented independently, the step 103 of determining a plurality of sets of pending digital service items in combination with the first session activity description content and the second session activity description content can be implemented as follows.

Step 201, determining a basic digital service item set covering a plurality of coarse-screening digital service items corresponding to each upstream and downstream constraint condition.

Wherein the underlying set of digitized service items can be understood as the initial set of digitized service items.

Step 202, combining the first session activity description content and the second session activity description content, determining a quantitative hit evaluation of each coarse-screening type digital service item in each basic digital service item set.

Step 203, combining the quantitative hit evaluation to clean the coarse-screening digital service items which do not reach the hit evaluation index in each basic digital service item set, and determining the plurality of undetermined digital service item sets according to each basic digital service item set which is cleaned.

In cloud game business, after paying attention to a game function, the digital virtual interaction device continues to pay attention to the game function related to the game function, so that two parts of contents are needed for multiple types of marking problems of the cloud game business, namely: service item execution data of all the coarse-screen type game function items (also in other words, coarse-screen type digitalized service items), and service item execution data of the game functions marked by the digital virtual interactive device (also in other words, target digitalized service items). Further, after the service item operation data are determined, corresponding session activity description contents can be mined for the coarse screening type game function items and the marked game functions respectively according to the upstream and downstream constraint conditions.

In some exemplary aspects, a session activity description may be created for each game function according to keywords of the game function, such as game type, game running load, and whether to pay a fee.

In mining the session activity description content of the coarse-screening type game function item (also in other words, the coarse-screening type digitalized service item) and the marked game function (also in other words, the target digitalized service item), the session activity description content of the above service item operation data can be mined through the machine learning model. In an actual application process, the first service item operation data may be loaded into one machine learning model to obtain a first session activity description content, and the second service item operation data may be loaded into another machine learning model to obtain a second session activity description content.

For the embodiment of the present disclosure, for each upstream and downstream constraint condition, a basic digital service item set may be set in advance, and the basic digital service item set may include the above several coarse-screening digital service items.

Further to determining the basic digital service item sets, quantitative hit ratings of the respective coarse-screen digital service items in the respective basic digital service item sets can be determined in combination with the first session activity description content and the second session activity description content.

It is understood that for each coarse-screening type of digital service, a quantitative hit score can be determined for each underlying set of digital services. This quantitative hit evaluation can be understood as: and the coarse screen type digital service items are matched with the possibility of the digital virtual interactive equipment according to the upstream and downstream constraint conditions corresponding to each basic digital service item set.

Further, for each basic digital service item set, the coarse-screening digital service items which do not reach the hit evaluation index in the basic digital service item set can be cleaned, and the cleaned basic digital service item set is determined as the pending digital service item set.

In the practical application process, the target coarse screening type digital service items of which the quantitative hit evaluation is not more than a first judgment value in each basic digital service item set can be determined; then, the target coarse-screening type digital service item is cleaned in a plurality of coarse-screening type digital service items of the basic digital service item set.

For the embodiment of the present disclosure, a first determination value may be set (for example, the first determination value may be set to 0.6). Furthermore, the coarse-screening digital service items with quantitative hit evaluation not greater than 0.6 can be cleaned, so that the coarse-screening digital service items reaching the hit evaluation index are obtained, and the coarse-screening digital service items reaching the hit evaluation index are determined as the to-be-determined digital service item set.

The method and the device are applied to the embodiment, the thought of determining the plurality of undetermined digital service item sets according to the quantitative hit evaluation of each coarse-screen digital service item in each basic digital service item set can accurately and credibly determine the undetermined digital service item sets corresponding to different detailed information, and therefore derived digital service items can be determined for the digital virtual interaction device more efficiently. Through comparison of the quantitative hit evaluation and the first judgment value, the idea of cleaning the coarse-screening type digital service items in the basic digital service item set according to the comparison condition can be used for timely and accurately cleaning the coarse-screening type digital service items which do not have a pairing relation with upstream and downstream constraint conditions of the basic digital service item set.

For some design ideas that can be implemented independently, the step 202 of determining the quantitative hit rating of each of the coarse-screening type digital services in each of the basic digital services sets in combination with the first session activity description content and the second session activity description content can be implemented as follows.

And 11, determining a binary operation result between the first session activity description content and the second session activity description content to obtain a target session activity description content.

The binary operation result can be understood as a vector product.

And step 12, determining the quantitative hit evaluation of each coarse-screening type digital service item in each basic digital service item set by combining the target session activity description content.

For the disclosed embodiments, the first session activity description content and the second session activity description content may be input into a first AI network model, whose behavior includes the following, so as to determine the above quantitative hit evaluation according to the output result of the first AI network model: and setting the first session activity description content and the second session activity description content to be U-dimensional eigenvectors, setting the number of the plurality of pending items to be V, and performing product processing on the first session activity description content and the second session activity description content to obtain the target session activity description content. The target session activity description content further gets the full _ coxected _ layers with the size of 2 x V through an input size of U, and then the V group pending digital service item set can be obtained. The data in each set of the to-be-determined digital service item sets can be processed through a set of logistic regression, and the quantitative hit evaluation probability _1 and the probability _2 of being marked of each coarse-screen type digital service item can be obtained. And aiming at each set of the pending digital service items, covering the coarse-screening digital service items belonging to the set of the pending digital service items, and the quantitative hit evaluation and the unmarked possibility corresponding to each coarse-screening digital service item.

If the probability _1>0.6, the coarse-screened digital service item will be included in the set of pending digital service item sets. In combination with the above ideas, each pending digital service item set covers a plurality of different coarse-screening digital service items, and there are differences in detail information between the different pending digital service item sets, for example, a first pending digital service item set may include a coarse-screening digital game function similar to the target digital service item, a second pending digital service item set may include a coarse-screening digital service item consistent with the target digital game function, and so on.

The method is applied to the embodiment, and the thought of quantitative hit evaluation of the coarse-screen type digital service projects in each basic digital service project set can be determined quickly, accurately and credibly by combining the ideas.

In cloud game services of game functions, most of digital virtual interaction devices continue to pay attention to game functions similar to a game function after one game function is ordered, and a small part of digital virtual interaction devices want to pay attention to game functions of the same game type or front-end and back-end game functions. However, in the learning thought with labels, if all previous examples are taken for debugging, small example samples are ignored, so that the debugging quality of the model is difficult to guarantee. In view of this, in the big data analysis method applied to digital space interaction provided in the embodiments of the present disclosure, by generating a plurality of pending digital service item sets for a target digital service item, detailed information (features) of small case data can be absorbed, and the richness of the plurality of pending digital service item sets is expanded.

Further, a target undetermined digital service item set reaching a set index can be screened from the undetermined digital service item sets.

For some design ideas that can be implemented independently, step 107, selecting a target pending digital service item set that meets a set index from the plurality of pending digital service item sets may be implemented in the following manner.

Step 301, determining the overall description content of each to-be-determined digital service item set respectively by combining the session activity description content of each coarse-screen digital service item in each to-be-determined digital service item set.

Step 302, obtaining the description content of the interaction behavior of the digital virtual interaction device.

The interactive behavior description content can be understood as a feature vector of the digital virtual interactive device at the device level in the interactive process.

Step 303, determining a confidence coefficient of each pending digital service item set by combining the overall description content and the interactive behavior description content, and determining the pending digital service item set of which the confidence coefficient reaches a set index as the target pending digital service item set.

For the embodiment of the present disclosure, a second AI network model may be called to screen a target pending digital service item set that reaches a set index in a plurality of pending digital service item sets, and an exemplary design idea is as follows: further to determining a number of sets of pending digital service items, it is also necessary to determine that the last set of pending digital service items (such as the target set of pending digital service items) is paired with the digital virtual interaction device. In this case, the positioning of the set of pending digital service items requires two parts of content, respectively: item detail features and digital virtual interaction device saliency descriptions for each set of pending digital service items.

In an actual application process, the session activity description content of each coarse-screen digital service item in each to-be-determined digital service item set may be determined in combination with the session activity description content of each coarse-screen digital service item in each to-be-determined digital service item set, so as to obtain the session activity description content of the to-be-determined digital service item set, where the session activity description content of the to-be-determined digital service item set may represent a type of a coarse-screen digital service item included in the to-be-determined digital service item set, where the type is intended to express an item detail feature of the to-be-determined digital service item set.

Each digital virtual interactive device has different attention tendencies aiming at different types of digital service items, and at the moment, the intention of the digital virtual interactive device needs to be acquired. Further, the interactive behavior description content can be determined from the digital virtual interactive device intent.

After the interactive behavior description content and the overall description content are determined, the product processing of the interactive behavior description content and the overall description content can be realized, the confidence coefficient of the digital virtual interactive device for pairing the service item set to be determined is obtained, and finally the service item set to be determined, of which the confidence coefficient reaches the set index, is selected to serve as the last service item set to be determined (in other words, the target service item set to be determined).

For some design ideas which can be independently implemented, the set of digital service items to be determined with the largest confidence coefficient can be used as the last set of digital service items to be determined. Under another possible design idea, one or more sets of pending digital service items with confidence coefficients not less than the second determination value may be used as the last set of pending digital service items. Under another possible design idea, the undetermined digital service item set with the first U maximum confidence coefficients in the confidence coefficients can be used as the last undetermined digital service item set.

The method is applied to the embodiment, by combining the interactive behavior description content and the session activity description content of each pending digital service item set and determining the thought of the last pending digital service item set in a plurality of pending digital service item sets, the derived digital service items which are adapted to the interactive preference of the digital virtual interactive device as much as possible can be positioned, so that the differentiated marking preference of different digital virtual interactive devices can be achieved.

In combination with the related technical content, for the embodiment of the present disclosure, a first AI network model may be invoked to process the first session activity description content and the second session activity description content, so as to obtain the plurality of pending digital service item sets; and calling a second AI network model to screen a target undetermined digital service item set which reaches a set index in the undetermined digital service item sets.

For the disclosed embodiments, it is proposed to improve the above multi-class labeling problem by two AI network models (such as a first AI network model and a second AI network model). Illustratively, a first AI network model is used to determine a number of sets of pending digital service items and a second AI network model is used to determine a last set of pending digital service items in the number of sets of pending digital service items.

Before determining a plurality of sets of service items to be digitalized and a final set of service items to be digitalized through the first AI network model and the second AI network model, the first AI network model and the second AI network model need to be debugged, and an exemplary debugging process includes the following steps: with the help of a target debugging strategy, debugging a first basic AI network model to be debugged to obtain the first AI network model, and debugging a second basic AI network model to be debugged to obtain the second AI network model, wherein the target debugging strategy comprises: a debug policy with authentication examples and a reactivation debug policy.

In combination with the above related technical content, for a traditional AI network model, a machine learning idea is generally used to find a quantization index for each digital service item, and then the digital service item with the selected quantization index ahead or the quantization index greater than a certain decision value is taken as a derivative digital service item, and the related functions of the network model are improved by using a labeled learning idea. The method and the device are applied to the embodiment of the disclosure, and the model operation quality of the AI network model can be improved by combining the label learning idea and the reliving debugging strategy. For the embodiment of the present disclosure, a big data analysis method applied to digital space interaction is provided, which can improve the problem of positioning of multiple types of data, and the big data analysis method applied to digital space interaction employs two AI network models, such as a first AI network model and a second AI network model. Wherein the first AI network model and the second AI network model are cascaded.

For the disclosed embodiments, a first AI network model estimates the quantitative hit rating of each coarse-screening type digitized service item in each set of underlying digitized service items from the first session activity description content and the second session activity description content, and a second AI network model estimates the quantitative hit rating (i.e., confidence coefficient) of each set of pending digitized service items. Therefore, the processing processes of the first AI network model and the second AI network model are matched with the statistical morphology in the reengineering debugging strategy, and the processing time consumption of the first AI network model and the second AI network model is not prolonged along with the expansion of the number of the marks, so that the time consumption of the digital service item marks can be saved by the big data analysis method applied to the digital space interaction, and the digital service item mining efficiency is improved.

The above debugging process of the first and second AI network models is explained below according to different scenarios.

Scene 1: the target debug policy includes a reengineering debug policy.

In the scenario 1, the first basic AI network model to be debugged is debugged by means of the target debugging strategy, which can be implemented in the following manner.

(1) Determining a first debugging example; the first debugging example comprises a plurality of digital service items to be debugged and a plurality of target derivative digital service items of each digital service item to be debugged.

(2) Loading the first debugging example into a first basic AI network model to be debugged to obtain quantitative hit evaluation of each target derivative digital service item; and determining a to-be-debugged digital service item set of the to-be-debugged digital service item by combining the quantitative hit evaluation of the target derivative digital service item.

(3) And determining a first model quality characteristic by combining quantitative hit evaluation of the to-be-debugged target derived digital service item set and the contribution index of the to-be-debugged first basic AI network model, and improving the model variable of the to-be-debugged first AI network model by combining the first model quality characteristic until the to-be-debugged first basic AI network model reaches a relive debugging index.

In the refitting strategy, for a first basic AI network model to be debugged, a label of each target derivative digital service item is understood as an isolated variable (independent event), and assuming that the number of the target derivative digital service items is x, at this time, quantitative hit evaluation of each target derivative digital service item can be determined through the first basic AI network model to be debugged. The set-up for the target derived digital services project in the first debugging paradigm includes a number of specified upstream and downstream constraints. In this way, the quantitative hit evaluation of the target derived digital service item can be understood as the possibility that the target derived digital service item is marked in each set of pending digital service items to be debugged, where the set of pending digital service items to be debugged is a set of digital service items including at least a part of the target derived digital service item determined in combination with the first debugging example and the specified upstream and downstream constraints.

Further, after the quantitative hit evaluation of the target derived digital service item is determined, a pending digital service item set to be debugged corresponding to each specified upstream and downstream constraint condition can be determined, and the overall determination possibility of the target derived digital service item in each pending digital service item set to be debugged is determined. After the overall likelihood whole _ P is determined, a way to be able to find the max contribution index (e.g., minimize the first model quality feature) can be found by reentering the gradient idea in the debugging strategy. Further, when the strategy is determined, the debugging process of the reactivation debugging strategy is finished, and at this time, the first AI network model reaching the reactivation debugging index can be obtained.

Scene 2: the target debug policy includes a reengineering debug policy.

In scenario 2, with the aid of a target debugging strategy, the second basic AI network model to be debugged is debugged, which can be implemented in the following manner.

(1) And loading the plurality of to-be-debugged digital service item sets to a second basic AI network model to be debugged to obtain quantitative hit evaluation of each to-be-debugged digital service item set.

(2) And determining a second model quality characteristic by combining the quantitative hit evaluation of each to-be-debugged digital service item set and the contribution index of the to-be-debugged second basic AI network model, and improving the model variable of the to-be-debugged second basic AI network model by combining the second model quality characteristic until the to-be-debugged second basic AI network model reaches a relive debugging index.

In combination with the related technical content, in scene 1, quantitative hit evaluation of the target derivative digital service item can be determined, and further, a plurality of pending digital service item sets to be debugged can be determined according to the quantitative hit evaluation. Then, a plurality of to-be-debugged digital service item sets are put into a second basic AI network model to be debugged, and quantitative hit evaluation (in other words, a confidence coefficient) of each to-be-debugged digital service item set is obtained; and then, determining a second model quality characteristic by combining the quantitative hit evaluation of the to-be-debugged digital service item set and the contribution index of a second basic AI network model to be debugged, and improving the model variable of the second basic AI network model according to the second model quality characteristic until determining the second AI network model reaching the relive debugging index. Wherein the model quality characteristic can be understood as a loss function.

Scene 3: the target debug policy includes a debug policy with authentication examples.

In scenario 3, with the aid of a target debugging strategy, debugging a first basic AI network model to be debugged can be implemented in the following manner.

(1) Determining a second debugging example, wherein the second debugging example covers a plurality of digital service items to be debugged and a real derivative digital service item set of each digital service item to be debugged.

(2) And debugging the first basic AI network model to be debugged based on the authentication example by combining the second debugging example.

In the debugging strategy with the certification example, a second debugging example can be set in advance, wherein the second debugging example covers a plurality of digital service items to be debugged and a real derivative digital service item set of each digital service item to be debugged. Furthermore, the second debugging example can be loaded into the first basic AI network model to be debugged for debugging, and the first AI network model meeting the requirements of the labeled learning idea can be obtained through debugging.

Scene 4: the target debug policy includes a debug policy with authentication examples.

In scenario 4, with the aid of the target debugging strategy, the second basic AI network model to be debugged is debugged, which can be implemented in the following manner.

(1) And determining a real derivative digital service item set of the digital service item to be debugged.

(2) And determining the digital service item common values of the real derivative digital service item set and each to-be-debugged digital service item set, and taking the to-be-debugged digital service item set corresponding to the digital service item common value which reaches the specified judgment value condition in the plurality of digital service item common values as a debugging keyword.

(3) And debugging the second basic AI network model to be debugged based on the authentication example through the plurality of to-be-debugged digital service item sets covering the debugging key words.

With reference to the foregoing related technology content, for the embodiment of the present disclosure, a commonality analysis algorithm (e.g., analyzing an aggregate and a union) may be used to determine a commonality value of the digital service items of the real derived digital service item set and each pending digital service item set to be debugged, and the pending digital service item set to be debugged with the largest commonality value of the digital service items is screened as a debugging keyword (training tag) to debug the second basic AI network model. The undetermined digital service item sets to be debugged which need to be improved are determined by combining the common value, so that each undetermined digital service item set proposed by the AI network model has different detail information, and the richness of the undetermined digital service item sets is ensured.

In an independently implementable design, after obtaining a derivative digital service item of the target digital service item, the method can further include: running the derived digital service item in combination with the digital virtual interaction equipment, and acquiring service feedback big data aiming at the derived digital service item through the digital virtual interaction equipment; determining service expectation information of the digital virtual interaction device for the derived digital service item based on the service feedback big data; and upgrading the derivative digital service item by combining the service expectation information.

For the embodiment of the disclosure, the big data analysis system may communicate with the digital virtual interaction device, and activate running of the derivative digital service item to implement service interaction communication between the big data analysis system and the plurality of digital virtual interaction devices, and the service feedback big data may be feedback of the digital virtual interaction device for the derivative digital service item in the interaction process, further, the service expectation information may be understood as a potential requirement of the digital virtual interaction device.

Under an independently implementable design idea, determining the service expectation information of the digital virtual interactive device for the derivative digital service item based on the service feedback big data may include the following contents: identifying a target significant service feedback event from the hot activity feedback record of the service feedback big data through a deployed user demand mining model, and obtaining the identified target significant service feedback event information; the user demand mining model is obtained by association deployment according to part of prior hot activity feedback records in the second prior example set and prior hot activity feedback records in the third prior example set; the second prior example set is obtained by performing mainstream interaction link annotation processing on prior hot activity feedback records which are not subjected to annotation processing in the first prior example set according to a user requirement mining model to be deployed, and the third prior example set comprises a plurality of prior hot activity feedback records which are subjected to annotation processing; determining value type feedback data and non-value type feedback data in the hot activity feedback record according to the target significant service feedback event information and a significant service feedback event constraint condition set in advance; and carrying out demand mining processing on the value type feedback data to obtain service expectation information of the digital virtual interaction equipment aiming at the derivative digital service project.

For the embodiment of the present disclosure, the user requirement mining model may be an LSTM model, the significant service feedback event may be a more core feedback event, the prior example set may be understood as a sample, the association deployment may be understood as joint training, and the mainstream interaction link may be understood as a core or key service interaction link. The design can quickly distinguish the value type feedback data and the non-value type feedback data in the hot activity feedback records by considering the constraint conditions of the significant service feedback events, so that the value type feedback data can be accurately and timely subjected to demand mining processing to obtain service expected information of the digital virtual interaction equipment for the derived digital service projects, and extra resource consumption caused by mining the non-value type feedback data is avoided.

Under an independently implementable design concept, the determining, according to the target significant service feedback event information and a significant service feedback event constraint condition set in advance, the value-type feedback data and the non-value-type feedback data in the hot activity feedback record includes: pairing at least part of feedback content sets in the constraint condition of the significant service feedback event with the determined target significant service feedback event information; determining comparison data between semantic tags of the significant service feedback events in the significant service feedback event constraint conditions and semantic tags of the target significant service feedback events in the hot activity feedback records according to pairing results; updating the semantic labels of the significant service feedback events in the constraint conditions of the significant service feedback events according to the comparison data; and projecting the updated semantic tag of the significant service feedback event into the hot activity feedback record to obtain the valuable feedback data including the target significant service feedback event in the hot activity feedback record and the non-valuable feedback data including at least part of the non-valuable feedback data. By the design, the value type feedback data and the non-value type feedback data can be accurately distinguished.

Fig. 3 is an architecture diagram illustrating an application environment of a big data analysis method applied to digital space interaction in which the big data analysis system 100 and the digital virtual interactive device 200, which communicate with each other, may be implemented according to an embodiment of the present disclosure. Based on this, the big data analysis system 100 and the digital virtual interaction device 200 implement or partially implement the big data analysis method applied to digital space interaction of the embodiment of the present disclosure at runtime.

The above description is only for the preferred embodiment of the present disclosure, and is not intended to limit the scope of the present disclosure.

Claims (10)

1. A big data analysis method applied to digital space interaction is characterized by being applied to a big data analysis system, and the method at least comprises the following steps:

determining a coarse screen type digital service project through a target digital service project;

acquiring session activity description contents based on the target digital service item and the coarse screening type digital service item, and determining a set of pending digital service items according to the session activity description contents;

and obtaining the derivative digital service item of the target digital service item according to the pending digital service item set.

2. The method of claim 1, wherein determining a coarse screen type digitized service item from a target digitized service item comprises: positioning a target digital service item marked by a digital virtual interactive device, and determining a plurality of coarse-screen digital service items with upstream and downstream constraint conditions existing in the target digital service item; wherein each coarse screening type digital service item is bound with at least one type of the upstream and downstream constraint condition, and the plurality of coarse screening type digital service items are bound with a plurality of types of the upstream and downstream constraint conditions;

the obtaining session activity description content based on the target digital service item and the coarse screening type digital service item, and determining a set of pending digital service items according to the session activity description content, includes: mining the session activity description content of the target digital service project to obtain a first session activity description content, and mining the session activity description content of each coarse screening type digital service project to obtain a second session activity description content; determining a plurality of sets of pending digital service items in combination with the first session activity description and the second session activity description; each pending digital service item set covers at least a part of coarse screening digital service items in the plurality of coarse screening digital service items, and each pending digital service item set is bound with at least one type of upstream and downstream constraint conditions;

the obtaining of the derivative digital service item of the target digital service item according to the pending digital service item set includes: and screening a target undetermined digital service item set reaching a set index in the plurality of undetermined digital service item sets, and determining coarse-screening digital service items in the target undetermined digital service item set as derivative digital service items of the target digital service items.

3. The method of claim 2, wherein said determining a number of sets of pending digital service items in conjunction with said first session activity description content and said second session activity description content comprises:

determining a basic digital service item set which covers a plurality of coarse screening digital service items and corresponds to each upstream and downstream constraint condition;

determining a quantitative hit evaluation of each of the coarse-screening type digital service items in each of the basic digital service item sets in combination with the first session activity description content and the second session activity description content;

and combining the quantitative hit evaluation to clean coarse-screening digital service items which do not reach the hit evaluation index in each basic digital service item set, and determining the plurality of undetermined digital service item sets according to each basic digital service item set which is cleaned.

4. The method of claim 3, wherein said determining a quantitative hit rating for each of said coarse screening type digital services items in each of said underlying digital services item sets in conjunction with said first session activity description content and said second session activity description content comprises:

determining a binary operation result between the first session activity description content and the second session activity description content to obtain a target session activity description content;

determining quantitative hit evaluation of each coarse-screening type digital service item in each basic digital service item set in combination with the target session activity description content;

wherein the cleaning of the coarse-screening type digital service items in each basic digital service item set which do not reach the hit evaluation index in combination with the quantitative hit evaluation comprises: determining a target coarse screen type digital service item of which the quantitative hit evaluation is not more than a first judgment value in each basic digital service item set; and cleaning the target coarse-screening type digital service item in a plurality of coarse-screening type digital service items of the basic digital service item set.

5. The method of claim 2, wherein the screening the set of pending digital service items for a set of target pending digital service items that meet a set criteria comprises: respectively determining the overall description content of each to-be-determined digital service item set by combining the session activity description content of each coarse-screen digital service item in each to-be-determined digital service item set; obtaining interactive behavior description content of the digital virtual interactive device; determining a confidence coefficient of each pending digital service item set by combining the overall description content and the interactive behavior description content, and determining the pending digital service item set with the confidence coefficient reaching a set index as the target pending digital service item set;

wherein the mining of the session activity description content of the target digital service item to obtain a first session activity description content, and the mining of the session activity description content of each coarse-screening digital service item to obtain a second session activity description content includes: acquiring service item operation data of the target digital service item and service item operation data of each coarse screen type digital service item; determining first service item operation data matched with each upstream and downstream constraint condition in the service item operation data of the target digital service item, and determining second service item operation data matched with the upstream and downstream constraint condition in the service item operation data of each coarse screening type digital service item; mining the session activity description content of the first service project operation data to obtain the first session activity description content; and mining the session activity description content of the second service item operation data to obtain the second session activity description content.

6. The method of claim 2, wherein said determining a number of sets of pending digital service items in conjunction with said first session activity description content and said second session activity description content comprises: calling a first AI network model to process the first session activity description content and the second session activity description content to obtain a plurality of pending digital service item sets;

the step of screening a target undetermined digital service item set which reaches a set index in the plurality of undetermined digital service item sets comprises the following steps: calling a second AI network model to screen a target undetermined digital service item set which reaches a set index in the plurality of undetermined digital service item sets;

wherein the method further comprises: with the help of a target debugging strategy, debugging a first basic AI network model to be debugged to obtain the first AI network model, and debugging a second basic AI network model to be debugged to obtain the second AI network model, wherein the target debugging strategy comprises: a debug policy with authentication examples and a reactivation debug policy.

7. The method of claim 6, wherein the target debugging policy comprises a reenergizing debugging policy; the debugging of the first basic AI network model to be debugged by means of the target debugging strategy comprises the following steps:

determining a first debugging paradigm; the first debugging example comprises a plurality of digital service items to be debugged and a plurality of target derivative digital service items of each digital service item to be debugged;

loading the first debugging example into a first basic AI network model to be debugged to obtain quantitative hit evaluation of each target derivative digital service item;

determining a to-be-debugged digital service item set of the to-be-debugged digital service item by combining the quantitative hit evaluation of the target derivative digital service item;

and determining a first model quality characteristic by combining the quantitative hit evaluation of the target derivative digital service item to be debugged in the to-be-debugged to the contribution index of the to-be-debugged first basic AI network model, and improving the model variable of the to-be-debugged first basic AI network model by combining the first model quality characteristic until the to-be-debugged first basic AI network model reaches a relive debugging index.

8. The method of claim 7, wherein the target debugging policy comprises a reenergizing debugging policy; the debugging of the second basic AI network model to be debugged by means of the target debugging strategy comprises the following steps:

loading the plurality of to-be-debugged digital service item sets to a second basic AI network model to be debugged to obtain quantitative hit evaluation of each to-be-debugged digital service item set;

and determining a second model quality characteristic by combining the quantitative hit evaluation of each to-be-debugged digital service item set and the contribution index of the to-be-debugged second basic AI network model, and improving the model variable of the to-be-debugged second basic AI network model by combining the second model quality characteristic until the to-be-debugged second basic AI network model reaches a reliving debugging index.

9. The method of claim 7, wherein the target debug policy comprises a debug policy having authentication instances; the debugging of the first basic AI network model to be debugged by means of the target debugging strategy comprises the following steps:

determining a second debugging example, wherein the second debugging example covers a plurality of digital service items to be debugged and a plurality of real derivative digital service item sets of each digital service item to be debugged;

debugging the first basic AI network model to be debugged based on the authentication example by combining the second debugging example;

wherein the target debug policy comprises a debug policy having an authentication instance; the debugging of the second basic AI network model to be debugged by means of the target debugging strategy comprises the following steps: determining a real derivative digital service item set of the digital service item to be debugged; determining the digital service item common values of the real derivative digital service item set and each to-be-debugged digital service item set, and taking the to-be-debugged digital service item set corresponding to the digital service item common value which reaches the specified judgment value condition in the plurality of digital service item common values as a debugging keyword; and debugging the second basic AI network model to be debugged based on the authentication example through a plurality of to-be-debugged digital service item sets covering the debugging key words.

10. A big data analytics system, comprising:

a memory for storing an executable computer program, a processor for implementing the method of any one of claims 1-9 when executing the executable computer program stored in the memory.

Images