CN115422463A

CN115422463A - Big data-based user analysis push processing method and system

Info

Publication number: CN115422463A
Application number: CN202211179150.4A
Authority: CN
Inventors: 高冬; 王莉
Original assignee: Individual
Current assignee: Ear Pattern Yuan Intelligent Technology Guangdong Co ltd
Priority date: 2022-09-27
Filing date: 2022-09-27
Publication date: 2022-12-02
Anticipated expiration: 2042-09-27
Also published as: CN115422463B

Abstract

According to the method and the system for processing the user analysis pushing based on the big data, provided by the embodiment of the invention, through the processing of identifying, matching, combining, screening, mining and the like on multiple types of behavior preference items in the user interaction activity information set, preference analysis and demand mining tasks under a complex user interaction activity scene can be effectively dealt with, and the upstream and downstream characteristics of the behavior preference items can be introduced to carry out mining and prediction on pushing decision requirements, so that the pushing decision requirements of the user interaction activity information set can be accurately and reasonably mined and predicted from the complex and various behavior preference items, and a credible analysis basis is provided for subsequent individuation and targeted pushing.

Description

Big data-based user analysis push processing method and system

Technical Field

The invention relates to the technical field of big data, in particular to a user analysis pushing processing method and system based on big data.

Background

The interest preference of the user is analyzed through a big data technology, the information matched with the interest preference is pushed, and the method has the characteristics of high efficiency, humanization, individuation and the like. It is conceivable that the current massive amount of information would be difficult to achieve effective dissemination without algorithmic intervention. At present, for the push repeat defect of big data, the following method can be generally used: avoid paying attention to information in a single aspect, and actively avoid low-popularity, simplification and entertainment contents pushed by a platform. Therefore, in order to improve personalized and targeted efficient pushing, it is very important to mine the pushing requirements of users, but it is difficult for related technologies to accurately and reasonably mine the pushing requirements in a complex environment.

Disclosure of Invention

In order to solve the technical problems in the related art, the invention provides a user analysis push processing method and system based on big data.

In a first aspect, an embodiment of the present invention provides a big data-based user analysis push processing method, which is applied to a user analysis push system, and the method includes: acquiring a user interaction activity information set which meets the requirement of big data mining; continuously mining and analyzing at least two types of behavior preference items in the user interaction activity information set to obtain directional mining information of each behavior preference item; determining the upstream and downstream characteristics between the at least two types of behavior preference items by combining the directional mining information of the at least two types of behavior preference items; determining service session interaction information corresponding to the behavior preference items according to set information extraction rules through the directional mining information of each behavior preference item; the information extraction rule is used for guiding the number of the service session interaction information with the same behavior preference item; and determining the push decision requirement of the user interaction activity information set by combining the upstream and downstream characteristics between the at least two types of behavior preference items and the service session interaction information of each behavior preference item.

By the design, through the identification, matching, combination, screening, mining and other processing of various behavior preference items in the user interaction activity information set, preference analysis and demand mining tasks in a complex user interaction activity scene can be effectively dealt with, and the upstream and downstream characteristics of the behavior preference items can be introduced to mine and predict the push decision demand, so that the push decision demand of the user interaction activity information set can be accurately and reasonably mined and predicted from the complex and various behavior preference items, and a credible analysis basis is provided for subsequent personalized and targeted push.

For some independently implementable embodiments, the obtaining a set of user interaction activity information that meets large data mining requirements comprises: acquiring not less than two groups of service interaction logs collected by an information acquisition unit; respectively processing one group of service interaction logs in the service interaction logs not lower than two groups by using a log processing algorithm not lower than two to obtain current user activity data in the service interaction logs of the corresponding group; and taking the current user activity data in the service interaction logs not lower than the two groups as the user interaction activity information set.

By the design, a group of user activity data is output in parallel by the service interaction logs in a plurality of log processing algorithms to form a user interaction activity information set, so that behavior preference items in the user interaction activity information set can be enriched, and the reliability and rationality of push decision demand mining are improved.

For some independently implementable embodiments, the directional mining information includes an identification core, a matching tag, and a matching record of the behavior preference item, and the performing continuous mining analysis on not less than two types of behavior preference items in the user interaction activity information set to obtain the directional mining information of each behavior preference item includes: performing behavior preference identification on at least two types of behavior preference items of the current user activity data in the user interaction activity information set to obtain an identification core of each behavior preference item; and matching the corresponding behavior preference items by combining the identification cores of the behavior preference items in the user interaction activity information set to obtain matching labels and matching records corresponding to the behavior preference items.

According to the design, firstly, multiple types of behavior preference items on the current user activity data are identified to obtain identification checks of all behavior preference items, and then the corresponding behavior preference items are matched based on the identification checks of the behavior preference items so as to ensure the identification accuracy of a single behavior preference item.

For some independently implementable embodiments, the no less than two types of behavior preference items include a mainstream behavior preference and an edge behavior preference; the mainstream behavior preference reflects a push demand theme; the performing behavior preference identification on at least two types of behavior preference items of the current user activity data in the user interaction activity information set to obtain an identification core of each behavior preference item includes: performing behavior preference identification on the mainstream behavior preference of the current user activity data in the user interaction activity information set to obtain an identification core of the mainstream behavior preference; performing behavior preference identification on the edge behavior preference in the current user activity data based on the current user activity data as selected user activity data to obtain a basic identification core of the edge behavior preference; the selected user activity data is extracted according to a set extraction step length; and performing preference matching processing on the edge behavior preference in the current user activity data based on the fact that the current user activity data is non-selected user activity data to obtain a candidate identification core of the edge behavior preference.

By the design, the selected user activity data and the non-selected user activity data are estimated for the edge behavior preference, abuse of an algorithm can be reduced, timeliness of behavior preference item identification is improved, and light weight of the whole scheme is achieved.

For some independently implementable embodiments, each current user activity data in the user interaction activity information set includes a digital authentication signature; the obtaining a candidate identification core of the edge behavior preference by performing preference matching processing on the edge behavior preference in the current user activity data based on the current user activity data as the non-selected user activity data includes: adjusting a first preference matching processing model by combining the basic identification core of the edge behavior preference; the base identification kernel is determined in selected user activity data prior to the current user activity data by the digital authentication signature; and estimating the distribution variable of the edge behavior preference in the current user activity data through the adjusted first preference matching processing model to obtain a candidate identification core of the edge behavior preference.

By the design, the first preference matching processing model is adjusted by using the basic recognition kernel of the edge behavior preference of the selected user activity data recognition, and the candidate recognition kernel of the edge behavior preference in the non-selected user activity data is estimated by using the first preference matching processing model, so that the advantages of good timeliness and high accuracy of preference matching processing are utilized, and the efficiency of pushing requirement mining analysis is ensured.

For some embodiments that can be implemented independently, the matching the corresponding behavior preference item in combination with the recognition core of each behavior preference item in the user interaction activity information set to obtain a matching tag and a matching record corresponding to the behavior preference item includes: loading the identification cores of all the behavior preference items in the user interaction activity information set to a second preference matching processing model to obtain a matching label of each behavior preference item; and determining a matching record corresponding to the behavior preference item based on the identification core corresponding to the same behavior preference item in a group of service interaction logs and the matching label corresponding to the behavior preference item.

By the design, the matching labels of the behavior preference items are output through the second preference matching processing model, the matching records are determined, the single identification window mined from each group of activity information can be associated and treated as the behavior preference item bound with the same matching label, and resource overhead is saved in the subsequent behavior preference item processing process.

For some independently implementable embodiments, the no less than two types of behavior preferences include the following: the method comprises the following steps of browsing items of a virtual mall, cross-border e-commerce concern items, VR service preference items, MR service preference items and hot topic preference items; the determining the upstream and downstream characteristics between the at least two types of behavior preference items by combining the directional mining information of the at least two types of behavior preference items comprises: determining cross-border mall associated information corresponding to the same e-commerce preference item in combination with a distribution variable joint analysis result between the identification core of the virtual mall browsing item and the identification core of the cross-border e-commerce concern item; determining somatosensory demand description characteristics between the cross-border e-commerce concern item and the VR service preference item by combining the result of the joint analysis of distribution variables between the recognition core of the cross-border e-commerce concern item and the recognition core of the VR service preference item; determining a first item participation feature or a second item participation feature between the hot topic preference and the MR service preference in combination with distribution variables of the respective identification cores of the hot topic preference and the MR service preference and the respective matching records of the hot topic preference and the MR service preference.

By the design, the virtual mall browsing items and the cross-border e-commerce attention items corresponding to the same business object are subjected to linkage analysis by using the identification cores of the behavior preference items, and the cross-border e-commerce attention items and the VR service preference items with somatosensory requirement description characteristics are subjected to linkage analysis; meanwhile, the cross-border e-commerce attention items and the MR service preference items are subjected to linkage analysis by combining the identification cores and the matching records of the behavior preference items, so that the service interaction logs are analyzed and determined as push decision requirements.

For some embodiments that can be implemented independently, the directional mining information further includes a quality coefficient of the behavior preference item, and the determining, according to a set information extraction rule, service session interaction information corresponding to the behavior preference item through the directional mining information of each behavior preference item includes: determining initial session interaction information by combining an identification core of a first behavior preference item based on the fact that the quality coefficient of the first behavior preference item in the current user activity data reaches a first quality score; the first behavior preference item is any one behavior preference item in the no less than two types of behavior preference items; recording the initial session interaction information into a temporary storage space of the first behavior preference item; and determining that the session interaction information recorded in the temporary storage space is the service session interaction information corresponding to the first behavior preference item based on that the temporary storage space of the first behavior preference item meets the preset screening starting requirement.

By means of the design, for each behavior preference item, whether the quality index recognized in the current user activity data reaches the first quality score or not is evaluated, namely screening conditions are determined, then initial session interaction information of the behavior preference item is determined and recorded into a temporary storage space, and finally service session interaction information with the best evaluation value recorded in the temporary storage space is screened according to a preset screening starting requirement, so that the space utilization rate is improved, and meanwhile the accuracy of subsequent knowledge vector mining is improved.

For some independently implementable embodiments, said entering the initial session interaction information into a staging space of the first behavioral preference comprises: and on the basis that the number of the session interaction information recorded in the temporary storage space of the first behavior preference item does not reach a storage limit value, unconditionally recording the initial session interaction information into the temporary storage space of the first behavior preference item.

By the design, the initial session interaction information is recorded into the temporary storage space on the basis of judging that the temporary storage space of the behavior preference item is not saturated by comparing the number of the session interaction information recorded in the temporary storage space with the storage limit value, so that the omission of the initial session interaction information of the behavior preference item is avoided.

For some independently implementable embodiments, the entering the initial session interaction information into the staging space of the first behavioral preference comprises: removing the first session interaction information in the temporary storage space on the basis that the number of the session interaction information recorded in the temporary storage space of the first behavior preference item reaches the storage limit value; the merit coefficient of the first session interaction information is lower than that of the initial session interaction information; and inputting the initial session interaction information into a temporary storage space of the first behavior preference item.

By means of the design, the number of the session interaction information recorded in the temporary storage space is compared with the storage limit value, the session interaction information with poor recognition evaluation value in the temporary storage space is removed on the basis of judging the saturation of the temporary storage space of the behavior preference item, and then the initial session interaction information is recorded in the temporary storage space, so that the overload of the temporary storage space of the behavior preference item is reduced, and the loss of the session interaction information with high evaluation value is caused.

For some independently implementable embodiments, the method further comprises: initializing the temporary storage space of the first behavior preference item based on the temporary storage space of the first behavior preference item reaching the screening start requirement.

By the design, the screening task is completed when the temporary storage space of the first action preference item meets the screening starting requirement, and the temporary storage space is initialized so that the screening task of the next user interaction activity information set can be normally performed.

For some independently implementable embodiments, the preset screening initiation requirement includes at least one of: screening timeliness requirements, screening intermittence requirements, matching time-consuming screening conditions, quality grade screening requirements and matching termination screening requirements; the screening timeliness requirement reflects that the duration for uninterruptedly positioning the first behavior preference reaches a first set duration; the screening intermittence requirement reflects that a gap time period matched with the first behavior preference reaches a set gap value; the matching time-consuming screening condition reflects that the total time length for matching the first behavior preference item reaches a second set time length; the second set time length is not less than the first set time length; the screening requirement of the goodness score reflects that session interaction information with a goodness coefficient reaching a second goodness score exists in the temporary storage space of the first behavior preference item; the second goodness score is greater than the first goodness score; the match termination filtering requirement reflects an overall level match of the first behavior preference until the collected service interaction log is completed.

By the design, the screening starting requirements of each behavior preference item can be flexibly set, so that the number of service session interaction information which is used for analyzing the same behavior preference item is limited based on multiple aspects, the evaluation value of screening the behavior preference items is improved, and the processing efficiency is improved.

For some embodiments that can be implemented independently, the determining, by combining the upstream and downstream features between the at least two types of behavior preference items and the service session interaction information of each of the behavior preference items, a push decision requirement of the user interaction activity information set includes: performing knowledge vector mining and/or element vector mining on the service session interaction information of each behavior preference item to obtain demand prediction information corresponding to the behavior preference item; combining the demand prediction information of the at least two types of behavior preference items by combining the upstream and downstream characteristics between the at least two types of behavior preference items; and determining a push decision requirement of the user interaction activity information set based on the combined requirement prediction information of the at least two types of behavior preference items.

By the design, the screened business session interaction information is subjected to big data mining, and the requirement prediction information is combined by combining the upstream and downstream characteristics among different behavior preference items to determine the push decision requirement of the user interaction activity information set, so that the precision and the rationality of the push decision requirement mining and prediction in a complex session environment are improved.

For some independently implementable embodiments, the behavior preference is a virtual mall browsing item, the number of the business session interaction information of the virtual mall browsing item is not less than 2; performing knowledge vector mining and/or element vector mining on the service session interaction information of each behavior preference item to obtain demand prediction information corresponding to the behavior preference item, including: respectively carrying out knowledge vector mining on the service session interaction information not less than two of the virtual mall browsing items to obtain not less than two knowledge description blocks; splicing the knowledge description blocks to obtain a linkage descriptor of the browsing item of the virtual mall; performing element vector mining on service session interaction information with the highest quality index in the service session interaction information not lower than the two pieces of service session interaction information to obtain an element description block of the virtual mall browsing item; and determining demand forecasting information of the virtual mall browsing item based on the linkage descriptor of the virtual mall browsing item and the element description block of the virtual mall browsing item.

By means of the design, the feature output quality of the descriptors can be improved by vector mining and splicing the service session interaction information of the browsing items of the virtual shopping malls, and meanwhile, element vector mining is performed on the service session interaction information with the best evaluation value, so that the reliability of demand mining prediction can be improved.

For some independently implementable embodiments, the method runs with the set rule by no less than two executing nodes in the log processing algorithm; the setting rule reflects that the generation result of the previous execution node is used as the input information of the next execution node between every two adjacent execution nodes through an information transition strategy.

By the design, a plurality of execution nodes are designed by adopting a multi-log processing algorithm to execute a big data-based user analysis push processing method, and an information transition strategy is matched between every two adjacent execution nodes, so that the processing efficiency of the whole scheme can be improved.

In a second aspect, the present invention further provides a user analysis pushing system, which includes a processor and a memory; the processor is in communication with the memory, and the processor is configured to read the computer program from the memory and execute the computer program to implement the method described above.

In a third aspect, the present invention also provides a server, including a processor and a memory; the processor is connected with the memory in communication, and the processor is used for reading the computer program from the memory and executing the computer program to realize the method.

Drawings

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

Fig. 1 is a schematic flow chart of a big data-based user analysis push processing method according to an embodiment of the present invention.

Fig. 2 is a schematic communication architecture diagram of an application environment of a big data-based user analysis push processing method according to an embodiment of the present invention.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings in which the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the invention, as detailed in the appended claims.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

The method provided by the embodiment of the invention can be executed in a user analysis pushing system, a computer device or a similar operation device. Taking the example of running on a user analytics push system, the user analytics push system 10 may include one or more processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, and optionally may also include a transmission device 106 for communication functions. It will be understood by those skilled in the art that the above structure is only illustrative and not limiting for the structure of the user analysis push system. For example, the user analytics push system 10 may also include more or fewer components than shown above, or have a different configuration than shown above.

The memory 104 may be used to store a computer program, for example, a software program and a module of application software, such as a computer program corresponding to a big data-based user analysis push processing method in an embodiment of the present invention, and the processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, so as to implement the method described above. The memory 104 may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory remotely located from the processor 102, which may be connected to the user analytics push system 10 over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission device 106 is used to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the user analysis push system 10. In one example, the transmission device 106 includes a Network adapter (NIC), which can be connected to other Network devices through a base station so as to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

Based on this, please refer to fig. 1, fig. 1 is a schematic flow chart of a big data-based user analysis pushing processing method according to an embodiment of the present invention, where the method is applied to a user analysis pushing system, and further includes the following technical solutions.

And a PROCESS110, obtaining a user interaction activity information set which meets the requirement of big data mining.

In the embodiment of the present invention, the user interaction activity information set meeting the big data mining requirement may be a user interaction activity information set in a set time period or a user interaction activity information set correspondingly determined according to a demand analysis instruction, and the user interaction activity information set includes two or more user interaction activity information. Furthermore, the user interaction activity information acquisition mode is not limited, and the user interaction activity information can be directly acquired and acquired through a user analysis pushing system and can also be received from other systems. In addition, the user interaction activity information may be activity records corresponding to different digital services, including but not limited to e-commerce, digital office, teleconferencing, smart city services, and the like.

And a PROCESS120, which is used for continuously mining and analyzing at least two types of behavior preference items in the user interaction activity information set to obtain directional mining information of each behavior preference item.

In the embodiment of the present invention, the at least two types of behavior preference items are important behavior preference items or behavior preference items with a higher attention degree in the pushing demand mining process, such as a virtual mall browsing item, a cross-border e-commerce attention item, a VR service preference item, an MR service preference item, a hot topic preference item, and the like, which are based on the user pushing demand mining. In the actual implementation process, various behavior preference items in each user interaction activity information in the user interaction activity information set are mined and uninterruptedly positioned, in other words, various behavior preference items can be continuously mined and tracked.

Further, the process of behavior preference identification may be understood as utilizing a DNN network to perform distributed variable identification and clustering operations of multiple classes of behavior preferences in user interaction information. The directional mining information of the behavior preference may include a recognition core (i.e., a recognition window) of the behavior preference, a category of the behavior preference, a matching tag (such as a location tag of the continuous analysis), a matching record (a location track of the continuous analysis), and the like. In addition, the behavior preference item can also be understood as a user behavior preference event, and the user behavior preference event can reflect the preference or tendency of the user, so that mining analysis of related push requirements can be performed based on the user behavior preference event.

For some embodiments which can be implemented independently, before performing behavior preference identification on the user interaction activity information, the obtained user interaction activity information set can be subjected to standardization processing, population characteristics of the user interaction activity information are weakened, and personalized characteristics of the user interaction activity information are highlighted.

And the PROCESS130 determines the upstream and downstream characteristics between the no less than two types of behavior preference items by combining the directional mining information of the no less than two types of behavior preference items.

In the embodiment of the invention, after directional mining information of each behavior preference item in the same user interaction activity information is obtained, the upstream and downstream characteristics between the virtual mall browsing item and the cross-border e-commerce attention item corresponding to the same e-commerce preference item can be mined according to the identification core and the matching tag of the behavior preference item; the upstream and downstream characteristics corresponding to the somatosensory demand description characteristics can be mined according to the relationship between the identification kernels of the two behavior preference items; and mining the upstream and downstream characteristics of the project participation characteristics according to the distribution variables and the matching records of the identification window.

And the PROCESS140 determines the service session interaction information corresponding to the behavior preference item according to the set information extraction rule through the directional mining information of each behavior preference item.

In the embodiment of the invention, after all the behavior preference items identified in the user interaction activity information set are matched and the upstream and downstream characteristics are determined, the identification evaluation value of the behavior preference items can be calculated, and further the service session interaction information with the quality coefficient meeting the requirements is determined to be subjected to subsequent processing analysis.

Wherein the information extraction rule is used for guiding the number of the service session interaction information with the same behavior preference item; the screening can be understood as that after the priority and the disadvantage coefficients of the service session interaction information in each group of activity information of the designated matching behavior preference items in the whole matching recording process are sequentially adjusted, the behavior preference item on the service session interaction information with the highest evaluation value is selected for processing.

Further, the matching behavior preference item is a behavior preference item with a matching tag, and exists continuously in a series of user interaction activity information, and the service session interaction information is a partial information set segmented from the user interaction activity information based on the identified behavior preference item identification core, and includes behavior preference item information (including the identification core of the behavior preference item, the behavior preference item information set, and other identification information) of the matching behavior preference item in a certain user interaction activity information, and is used for performing knowledge vector mining, element description block analysis, and the like.

It can be understood that, the good and bad coefficients are respectively determined for each behavior preference item, and the thought groups determined for different behavior preference items can be flexibly selected, and only the good and bad coefficient interval of the behavior preference item needs to be ensured to be [0,1]. Therefore, the service session interaction information of the best evaluation value can be conveniently selected for feature mining analysis after adjustment according to the quality coefficient sequence.

And the PROCESS150 determines the push decision requirement of the user interaction activity information set by combining the upstream and downstream characteristics between the at least two types of behavior preference items and the service session interaction information of each behavior preference item.

In the embodiment of the invention, the service session interaction information of each behavior preference item, which is screened and output, is subjected to knowledge vector mining and element description block analysis, and the processing results are combined to be used as the push decision requirement of a user interaction activity information set in combination with the upstream and downstream characteristics among different behavior preference items, and the prediction processing or regression analysis processing can also be carried out according to the upstream and downstream characteristics among no less than two types of behavior preference items and the service session interaction information of each behavior preference item, so that the push decision requirement is obtained, and the push decision requirement can be used for guiding the large data push aiming at the user, so that the push precision and pertinence are improved. .

Further, after the extraction processing of the knowledge description block and the element description block of the service session interaction information of all the behavior preference items output by screening is completed, all the behavior preference items are combined, and then each service session interaction information of each group of service interaction logs is processed in an iterative manner until all the processing is completed.

It can be understood that processes 110 to 150 according to the embodiment of the present invention may run in a loop, that is, the latest set of user interaction activity information is continuously obtained for processing until all the service session interaction information processing is completed or a stop instruction is received.

In the embodiment of the invention, through carrying out identification, matching, combination, screening, mining and other processing on a plurality of types of behavior preference items in the user interaction activity information set, preference analysis and demand mining tasks in a complex user interaction activity scene can be effectively responded, and upstream and downstream characteristics of the behavior preference items can be introduced to carry out mining and prediction on push decision requirements, so that the push decision requirements of the user interaction activity information set can be accurately and reasonably mined and predicted from the complex and various behavior preference items, and credible analysis basis is provided for subsequent personalized and targeted push.

For some independently implementable embodiments, the processes 110 to 150 run with set rules (e.g., synchronous parallel processing) by no less than execution nodes (e.g., as related function modules) in two log processing algorithms (e.g., as computer threads); the set rule reflects that between every two adjacent execution nodes, the generation result of the previous execution node is used as the input information of the next execution node through an information transition policy (for example, a cache space can be set for data information transmission in the node processing process).

Due to the design, different execution nodes operate according to different log processing algorithms, and the processing efficiency of the whole scheme can be improved. For example, a plurality of execution nodes are configured based on a multi-log processing algorithm to run the big-data-based user analysis push processing method, and information transition policies (such as setting a cache space) are matched between every two adjacent execution nodes, so that the processing efficiency of the whole scheme can be improved.

In some examples, the directional mining information includes an identification core, a matching tag, and a matching record for the behavior preference. The following is a related exemplary description of a big data-based user analysis push processing method according to an embodiment of the present invention.

And a PROCESS210, which obtains no less than two groups of service interaction logs collected by the information acquisition unit.

And the PROCESS220 is used for processing one service interaction log in the service interaction logs in the first group through a log processing algorithm in the second group to obtain the current user activity data in the service interaction logs in the corresponding group.

In the embodiment of the present invention, in each log processing algorithm, one data extractor is used to process a group of service interaction logs or interaction log texts (for example, an activity data extraction operation), for example, a log processing algorithm _ a is used to process a 1 st group of service interaction logs, and a log processing algorithm _ B is used to process a 2 nd group of service interaction logs, so that multiple groups of service interaction logs can be processed simultaneously by using multiple log processing algorithms, and current user activity data in each group of service interaction logs can be obtained simultaneously.

And processing 230, using the current user activity data in the service interaction logs not lower than the two groups as the user interaction activity information set.

In the embodiment of the invention, a plurality of groups of service interaction logs extract current user activity data in a plurality of log processing algorithms simultaneously to form a data set, namely a user interaction activity information set.

And the PROCESS240 is used for carrying out behavior preference identification on not less than two types of behavior preference items of the current user activity data in the user interaction activity information set to obtain an identification core of each behavior preference item.

In the embodiment of the invention, the identification core of the behavior preference item is an identification window. The behavior preference identification can be carried out on a group of user interaction activity information through a unitary regression component network, and the behavior preference identification can also be carried out on a group of user interaction activity information through a multivariate regression component network.

For some embodiments that can be implemented independently, for each behavior preference, behavior preference recognition is performed sequentially using a univariate regression component network to obtain a recognition core of the behavior preference. For example, for a virtual mall browsing item with a unique identifiable user tag, using a univariate regression network to identify the virtual mall browsing item in each user interaction activity information so as to ensure the identification precision of the virtual mall browsing item.

In other examples, the multiple classes of behavior preference items are simultaneously subjected to behavior preference identification by using a multivariate regression component network, and identification cores of the behavior preference items in a group of user interaction activity information are obtained. For example, for edge behavior preferences such as VR service preferences, MR service preferences and hot topic preferences, a multivariate regression sub-network is used to identify the behavior preferences in the user interaction activity information selected according to a certain period, thereby improving the behavior preference identification efficiency.

And the PROCESS250 is used for matching the corresponding behavior preference items by combining the identification cores of the behavior preference items in the user interaction activity information set to obtain matching labels and matching records corresponding to the behavior preference items.

In the embodiment of the invention, the matching labels of the behavior preference items are output through the second preference matching processing model, and the behavior preference items in the previous and subsequent information have the same and unique matching labels, so that the behavior preference items can be used for performing behavior preference item association or screening analysis subsequently.

In the actual implementation process, firstly, multiple types of behavior preference items on current user activity data are identified to obtain identification cores of all behavior preference items, then corresponding behavior preference items are matched based on the identification cores of the behavior preference items, and a single identification window mined from each group of activity information is associated through a matching tag to form a matching record, so that the identification accuracy for the single behavior preference item is ensured.

And the PROCESS260 determines the upstream and downstream characteristics between the no less than two types of behavior preference items by combining the directional mining information of the no less than two types of behavior preference items.

In the embodiment of the invention, the directional mining information comprises an identification core, a matching tag and a matching record of the behavior preference item. The upstream and downstream characteristics between the at least two types of behavior preference items can include cross-border mall associated information, somatosensory demand description characteristics, first item participation characteristics or second item participation characteristics and the like, so that the upstream and downstream characteristics can be understood as an association relationship.

Furthermore, the cross-border mall related information mainly refers to linkage analysis of virtual mall browsing items and cross-border e-commerce attention items corresponding to the same business object; the somatosensory requirement description characteristics refer to upstream and downstream characteristics of cross-border e-commerce attention items and VR service preference items, and are upstream and downstream characteristics corresponding to the somatosensory requirement description characteristics respectively; the first item participation characteristic and the second item participation characteristic both refer to upstream and downstream characteristics between the cross-border e-commerce concern item and the MR service preference item.

Further, the upstream and downstream characteristics may be determined by the following idea: determining cross-border mall associated information corresponding to the same e-commerce preference item in combination with a distribution variable joint analysis result between the identification core of the virtual mall browsing item and the identification core of the cross-border e-commerce concern item; determining somatosensory demand description characteristics between the cross-border e-commerce concern item and the VR service preference item by combining the result of the joint analysis of distribution variables between the recognition core of the cross-border e-commerce concern item and the recognition core of the VR service preference item; and determining a first item participation characteristic or a second item participation characteristic between the hot topic preference item and the MR service preference item by combining distribution variables (such as position information) of the identification cores of the hot topic preference item and the MR service preference item and the matching records of the hot topic preference item and the MR service preference item.

By means of the design, the virtual mall browsing items and the cross-border e-commerce attention items corresponding to the same business object are subjected to linkage analysis by using the identification cores of the behavior preference items, and the cross-border e-commerce attention items and the VR service preference items with somatosensory requirement description characteristics are subjected to linkage analysis. And further combining the identification core and the matching record of each behavior preference item, performing linkage analysis on the cross-border E-commerce concern item and the MR service preference item, and finally performing prediction mining processing of pushing decision requirements.

And the PROCESS270 determines the service session interaction information corresponding to the behavior preference item according to the set information extraction rule through the directional mining information of each behavior preference item.

In the embodiment of the present invention, the information extraction rule is used to guide the number of the service session interaction information of the same behavior preference.

And the PROCESS280 determines the push decision requirement of the user interaction activity information set by combining the upstream and downstream characteristics between the at least two types of behavior preference items and the service session interaction information of each behavior preference item.

In the embodiment of the invention, a group of user activity data is output in parallel in a plurality of log processing algorithms for a plurality of groups of service interaction logs to form a user interaction activity information set, then behavior preference identification is carried out on the user interaction activity information set at the same time to obtain identification checks of all behavior preference items, and corresponding behavior preference items are matched based on the identification checks of the behavior preference items, so that the identification accuracy of a single behavior preference item is ensured, the accurate identification of a plurality of types of behavior preference items is realized, and the accuracy and the reliability of demand mining are improved.

For some examples, the no less than two types of behavior preferences include mainstream behavior preferences and edge behavior preferences; the mainstream behavior preferences reflect push demand topics. Based on this, the PROCESS120 "continuously mining and analyzing not less than two types of behavior preference items in the user interaction activity information set to obtain the directional mining information of each behavior preference item" may be implemented by the following technical solution.

And a PROCESS310, which is used for obtaining the mainstream behavior preference of the current user activity data in the user interaction activity information set to perform behavior preference identification, and obtaining an identification core of the mainstream behavior preference.

The mainstream behavior preference refers to a behavior preference item which can only reflect the pushing demand theme in the current user activity data. For some e-commerce services, virtual mall browsing items may be preferred as mainstream behavior, with unique identification of user tags.

In the actual implementation process, a unitary regression component network is used for identifying the virtual mall browsing items for each group of activity information in the service interaction log, and an identification core of the virtual mall browsing items in each group of activity information, namely an item capturing window, is obtained, so that the accuracy of the virtual mall browsing items is ensured.

And a PROCESSS 320, which is used for performing behavior preference identification on the edge behavior preference in the current user activity data based on the current user activity data as the selected user activity data to obtain a basic identification core of the edge behavior preference.

Further, the selected user activity data is extracted according to a set extraction step size. The edge behavior preference is other behavior preference items of the service session interaction information except the mainstream behavior preference, such as cross-border e-commerce attention items, VR service preference items, MR service preference items and hot topic preference items related to behavior attention events.

In the selected user activity data, a multiple regression partial network is used for identifying various edge behavior preferences at the same time, and a basic identification core of each edge behavior preference is obtained. The simultaneous identification of multiple behavior preferences is typically only performed for the case where the current user activity data is selected user activity data, thereby avoiding overuse of resources.

And a PROCESS330, based on that the current user activity data is non-selected user activity data, performing preference matching processing on the edge behavior preference in the current user activity data to obtain a candidate identification core of the edge behavior preference.

Further, each current user activity data in the user interaction activity information set includes a digital authentication signature, and after the basic identification kernel of the edge behavior preference in the selected user activity data is further identified, the candidate identification kernel (updated identification kernel) of each edge behavior preference in the non-selected user activity data after estimation is performed by using the basic identification kernel of each edge behavior preference. Therefore, abuse of the algorithm can be reduced for the edge behavior preference, timeliness of behavior preference item identification is improved, and light weight of the whole scheme is achieved.

For some examples, the first preference matching processing model is adjusted in conjunction with a base recognition kernel of the edge behavior preferences; the base identification kernel is determined in selected user activity data prior to the current user activity data by the digital authentication signature; and estimating the distribution variable of the edge behavior preference in the current user activity data through the adjusted first preference matching processing model to obtain a candidate identification core of the edge behavior preference.

And the PROCESS340 loads the identification cores of all the behavior preference items in the user interaction activity information set to a second preference matching processing model to obtain matching labels of all the behavior preference items.

In the embodiment of the present invention, after the mainstream behavior preference identification of PROCESS310 and the identification and combined estimation of edge behavior preferences in processes 320 to 330 are performed on all current user activity data in the user interaction activity information set, the distribution variables and categories of all behavior preferences on the user interaction activity information are determined, and a second preference matching processing model (multi-classification model) can be used to perform multi-line preference matching to obtain a matching label of each recognition window.

And a processing 350, determining a matching record corresponding to the behavior preference item based on the identification core corresponding to the same behavior preference item in a group of service interaction logs and the matching label corresponding to the behavior preference item.

In the embodiment of the invention, the identification cores which are mined from the data of each current user activity and carry the same matching labels are correlated and are used as the matching records of the behavior preference items for processing, thereby facilitating the subsequent screening analysis.

For example, the concept of processes 320-330 employing selected user activity data identification and non-selected user activity data estimation for edge behavior preferences is due to the fact that the large-scale processing of the multivariate regression component network used across border-e-provider concerns, VR service preferences, MR service preferences, and hot topic preferences is slow and time consuming. The concept of using selected user activity data identification and non-selected user activity data estimation takes advantage of the high efficiency of preference matching processing to improve processing efficiency. The data group interval of the selected user activity data in the preference matching process is generally 4-8 groups, the data group interval is too small to speed up, the accuracy of single behavior preference estimation is reduced when the data group interval is too large, and the error of the identification core of the estimated behavior preference is larger.

In the actual implementation process, after the mainstream behavior preference of each user interaction activity information is identified to obtain the identification core of the mainstream behavior preference, the current service session interaction information is judged to be selected user activity data or non-selected user activity data, and the identification core of the edge behavior preference is obtained according to the set estimated by the selected user activity data + the non-selected user activity data. In some examples, implementations of the behavior preference identification matching process provided for the examples of the present invention may include the following.

PROCESS311, an identification core identifying the mainstream behavior preference.

PROCESS312, determining whether the current user activity data is the selected user activity data. If the judgment result is yes, processing 313 is executed; if the judgment result is negative, processing 315 is executed.

A PROCESSS 313 for identifying a basic identification core of the edge behavior preference;

a PROCESS314, adjusting the first preference matching processing model;

here, the first preference matching process model (single classification model) refers to a process of estimating candidate recognition kernels of behavior preferences in the next several groups of user activity data based on the basic recognition kernels of behavior preferences in the selected user activity data, wherein the process of adjusting the first preference matching process model is a process of taking distribution variables of recognition kernels in the selected user activity data as basic distribution variables required by the first preference matching process model to match new distribution variables of the estimated behavior preferences.

And a PROCESS315, estimating candidate identification kernels with edge behavior preference through the first preference matching processing model.

Further, the same first preference matching processing model is used for matching different edge behavior preferences in the current user activity data, and the first preference matching processing model can match and estimate a plurality of different recognition window distribution variables at one time. The first preference matching processing model carries out targeted matching processing on the edge behavior preferences of the types.

And a PROCESS316, which loads the identification cores of all the identified behavior preference items to a second preference matching processing model, and respectively matches each behavior preference item in the current user activity data.

In the embodiment of the invention, the second preference matching processing model is used for performing linkage analysis on the identification cores corresponding to the same behavior preference items in the adjacent groups of user activity data and configuring a unique matching label, the main process is to input the identification cores of all the behavior preference items in the current user activity data and generate the matching label of the identification core corresponding to each behavior preference item, and the same behavior preference items in the previous and next information have the same and unique matching label.

In the embodiment of the invention, the first preference matching processing model is adjusted by utilizing the basic identification core of the edge behavior preference identified by the selected user activity data, and the candidate identification core of which the edge behavior preference is in the non-selected user activity data is estimated by utilizing the first preference matching processing model, so that the advantages of good timeliness and high accuracy of preference matching processing are utilized, and the efficiency of pushing requirement mining analysis is ensured. By the design, the matching labels of the behavior preference items are output through the second preference matching processing model, the matching records are determined, the single identification window mined from each group of activity information can be associated and treated as the behavior preference item bound with the same matching label, and resource overhead is saved in the subsequent behavior preference item processing process.

For some examples, the directional mining information further includes a goodness coefficient (a quality score or a quality assessment value of a behavior preference) of the behavior preference. Based on this, the PROCESS140 "determines the service session interaction information corresponding to the behavior preference item according to the set information extraction rule by mining the information oriented to each behavior preference item" may include the following contents.

And a PROCESS410, determining initial session interaction information by combining the identification cores of the first behavior preference items based on the first goodness score achieved by the goodness coefficient of the first behavior preference item in the current user activity data.

In the embodiment of the present invention, the first behavior preference item is any behavior preference item of the at least two types of behavior preference items, and the first goodness score reflects a lowest goodness coefficient as a filtering condition, and may be 0.3, for example.

For some examples, if the goodness coefficient of the identified first behavioral preference is not below the first goodness score, the local information set corresponding to the identified core of the first behavioral preference is determined from the current user activity data as the initial session interaction information.

For other examples, the identification of the first behavioral preference is deleted if the goodness coefficient of the identified first behavioral preference is lower than the first goodness score.

And a PROCESS420, which records the initial session interaction information into a temporary storage space of the first behavior preference item.

Whether the temporary storage space of the first behavior preference item is saturated or not can be judged first, and whether the initial session interaction information is directly recorded in the temporary storage space or the initial session interaction information is overwritten by the recorded session interaction information with the lowest evaluation value is determined based on the judgment result.

For some examples, the initial session interaction information is unconditionally entered into the staging space for the first behavioral preference based on the number of session interaction information recorded in the staging space for the first behavioral preference not reaching a storage limit. Therefore, by comparing the number of the session interaction information recorded in the temporary storage space with the storage limit value, the initial session interaction information is recorded into the temporary storage space on the basis of judging that the temporary storage space of the behavior preference item is not saturated, and the deletion of the initial session interaction information of the behavior preference item is effectively reduced.

For other examples, the first action preference is a preference for a first action in the temporal space, and the first action preference is a preference for a second action in the temporal space; the merit coefficient of the first session interaction information is lower than that of the initial session interaction information; and inputting the initial session interaction information into a temporary storage space of the first action preference item. Therefore, by comparing the number of the session interaction information recorded in the temporary storage space with the storage limit value, the session interaction information with poor identification evaluation value in the temporary storage space is removed on the basis of judging that the temporary storage space of the behavior preference item is saturated, and then the initial session interaction information is recorded in the temporary storage space, so that the overload of the temporary storage space of the behavior preference item is effectively reduced, and the loss of the session interaction information with good evaluation value is caused.

And a PROCESS430, determining that the session interaction information recorded in the temporary storage space is the service session interaction information corresponding to the first behavior preference item, based on that the temporary storage space of the first behavior preference item meets a preset screening start requirement.

In an embodiment of the present invention, the preset screening start requirement includes at least one of the following items: screening timeliness requirements, screening intermittence requirements, matching time-consuming screening conditions, quality grade screening requirements and matching termination screening requirements. And through the flexible screening starting requirements of multiple categories, noise interference is reduced, and the screening timeliness and the screening evaluation value of the behavior preference item are improved.

In an embodiment of the present invention, the screening timeliness requirement reflects a length of time to uninterruptedly locate the first action preference up to a first set length of time. For example, given a duration, i.e., a first set duration, the screening output is enabled once the behavioral preference is located uninterrupted beyond that duration. The screening intermittency requirement reflects that a gap period matching the first behavioral preference reaches a set gap value. And the matching time-consuming screening condition reflects that the total time length for matching the first behavior preference item reaches a second set time length. The screening requirement of the goodness score reflects that session interaction information with the goodness coefficient reaching a second goodness score exists in the temporary storage space of the first behavior preference item; the second goodness score is greater than the first goodness score. For example, when the quality coefficient of the evaluation value session interaction information recorded in the behavior preference item reaches the second quality score, the screening output is performed, and then the screening is not performed. Thus, behavior preference items with certain high evaluation value can be extracted. The match termination filtering requirement reflects an overall level of matching the first behavior preference until the collected service interaction log is complete. In other words, after the matching of the behavior preference item is completed, the session interaction information of the best evaluation value in the whole course of the matching record of the behavior preference item is filtered and output, and is usually used as an initial scheme. For the general behavior preference item, selecting the best service session interaction information output from the whole matching process is a preferred scheme.

And the PROCESS440 initializes the temporary storage space of the first behavior preference item based on that the temporary storage space of the first behavior preference item reaches the screening starting requirement.

In the embodiment of the invention, the screening task is completed when the temporary storage space of the first behavior preference item meets the screening starting requirement, and the initialization of the temporary storage space is convenient for the normal implementation of the screening task of the next user interaction activity information set.

In the embodiment of the invention, for each behavior preference item, whether the quality index identified in the current user activity data reaches the first quality score or not is estimated, namely a screening condition is determined, then the initial session interaction information of the behavior preference item is determined and recorded in the temporary storage space, and finally the service session interaction information with the best evaluation value recorded in the temporary storage space is determined according to the preset screening starting requirement, so that the space utilization rate is improved and the accuracy of subsequent knowledge vector mining is improved. Meanwhile, various information extraction rules are used, the screening requirements of each application environment of each behavior preference item are flexibly set, and the resource utilization rate is improved.

In some independent embodiments, the PROCESS150 "determines the push decision requirement of the user interaction activity information set by combining the upstream and downstream characteristics between the at least two types of behavior preference items and the service session interaction information of each of the behavior preference items" may include the following.

And a PROCESS510, performing knowledge vector mining and/or element vector mining on the service session interaction information of each behavior preference item to obtain demand prediction information corresponding to the behavior preference item.

In the embodiment of the invention, knowledge vector mining (feature extraction) is carried out on behavior preference items related to a user through a knowledge vector mining model to obtain a knowledge description block (feature vector) with a certain dimension, and element vector mining is carried out through an element vector mining model to obtain element description blocks such as an e-commerce behavior preference element description block and a VR behavior preference element description block; and for other unimportant behavior preference items such as VR service preference items, performing element vector mining only through an element vector mining model to obtain corresponding element description blocks.

In the actual implementation process, different types of behavior preference items are analyzed respectively. For virtual mall browsing items, respectively performing knowledge vector mining on no less than two service session interaction information of the virtual mall browsing items to obtain no less than two knowledge description blocks; splicing the knowledge description blocks to obtain a linkage descriptor of the browsing item of the virtual mall; performing element vector mining on service session interaction information with the highest quality index in the service session interaction information not lower than the two pieces of service session interaction information to obtain an element description block of the virtual mall browsing item; and determining demand forecasting information of the virtual mall browsing item based on the linkage descriptor of the virtual mall browsing item and the element description block of the virtual mall browsing item. Therefore, the feature output quality of the descriptor can be improved by vector mining and splicing the service session interaction information of the browsing items of the virtual malls; meanwhile, element vector mining is carried out on the service session interaction information with the best evaluation value, and the reliability of demand mining prediction can be improved.

For cross-border e-commerce attention items, generally, business session interaction information is screened and output, knowledge vector mining of VR behavior preference is performed directly by using a DNN (digital network), then element vector mining is performed on VR behavior preference by using a VR behavior preference element description block model, and then features and element description blocks of the VR behavior preference are used as demand prediction information of the cross-border e-commerce attention items. For the VR service preference item, element vector mining is carried out by using a feature pyramid network, and then an interactive topic element description block such as topic viewpoint distribution variables and topic viewpoint contents is used as demand prediction information of the VR service preference item. And for the MR service preference item, performing mixed reality element vector mining by using an element description block model, and using the element description block model as the demand prediction information of the MR service preference item.

A PROCESS520, combining the demand prediction information of the at least two types of behavior preference items with the upstream and downstream characteristics between the at least two types of behavior preference items;

in the embodiment of the invention, the demand prediction information of each behavior preference item is combined through the upstream and downstream characteristics among different behavior preference items, so that excessive data processing pressure can be avoided.

And a processing 530, determining a push decision requirement of the user interaction activity information set based on the combined requirement prediction information of the at least two types of behavior preference items.

In the embodiment of the invention, after mining of knowledge descriptors and element description blocks of all kinds of service session interaction information which is screened and output is completed, the push decision-making requirement of the user interaction activity information set is determined jointly based on the requirement prediction information of all behavior preference items.

In the embodiment of the invention, the screened business conversation interaction information is subjected to big data mining, and the demand forecasting information is integrated by combining the upstream and downstream characteristics among different behavior preference items to serve as the push decision demand of the user interaction activity information set, so that the precision and the rationality of the push decision demand mining and forecasting under the complex conversation environment are improved.

Based on the same or similar inventive concepts, please refer to fig. 2 in combination, and a schematic structural diagram of an application environment 30 of a big data-based user analysis pushing processing method is also provided, which includes a user analysis pushing system 10 and a service client 20 that communicate with each other, and the user analysis pushing system 10 and the service client 20 implement or partially implement the technical solutions described in the above method embodiments when running.

Further, a readable storage medium is provided, on which a program is stored, which when executed by a processor implements the method described above.

Further, the invention also provides a server, comprising a processor and a memory; the processor is connected with the memory in communication, and the processor is used for reading the computer program from the memory and executing the computer program to realize the method.

In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus and method embodiments described above are illustrative only, as the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, the functional modules in the embodiments of the present invention may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention or a part thereof, which essentially contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a media service server 10, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a … …" does not exclude the presence of another identical element in a process, method, article, or apparatus that comprises the element.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A big data-based user analysis push processing method is applied to a user analysis push system, and comprises the following steps:

obtaining the user interaction activity information set which meets the big data mining requirement, and continuously mining and analyzing not less than two types of behavior preference items in the user interaction activity information set to obtain directional mining information of each behavior preference item;

determining the upstream and downstream characteristics between the at least two types of behavior preference items by combining the directional mining information of the at least two types of behavior preference items; determining service session interaction information corresponding to the behavior preference items according to set information extraction rules through the directional mining information of each behavior preference item; wherein the information extraction rule is used for guiding the number of the service session interaction information with the same behavior preference item;

and determining the push decision requirement of the user interaction activity information set by combining the upstream and downstream characteristics between the at least two types of behavior preference items and the service session interaction information of each behavior preference item.

2. The method of claim 1, wherein obtaining a set of user interaction activity information that meets big data mining requirements comprises:

acquiring not less than two groups of service interaction logs collected by an information acquisition unit;

respectively processing one group of service interaction logs in the service interaction logs not lower than two groups by using a log processing algorithm not lower than two to obtain current user activity data in the service interaction logs of the corresponding group;

and taking the current user activity data in the service interaction logs not lower than the two groups as the user interaction activity information set.

3. The method of claim 2, wherein the directionally mined information includes identification cores, matching labels, and matching records of the behavior preferences, and wherein the continuously mining not less than two types of behavior preferences in the user interaction activity information set to obtain directionally mined information for each of the behavior preferences includes:

performing behavior preference identification on at least two types of behavior preference items of the current user activity data in the user interaction activity information set to obtain an identification core of each behavior preference item;

and matching the corresponding behavior preference items by combining the identification cores of the behavior preference items in the user interaction activity information set to obtain matching labels and matching records corresponding to the behavior preference items.

4. The method of claim 1, wherein the no less than two categories of behavior preference items include a mainstream behavior preference and an edge behavior preference; the mainstream behavior preference reflects a push demand topic; the performing behavior preference identification on at least two types of behavior preference items of the current user activity data in the user interaction activity information set to obtain an identification core of each behavior preference item includes: performing behavior preference identification on the mainstream behavior preference of the current user activity data in the user interaction activity information set to obtain an identification core of the mainstream behavior preference; performing behavior preference identification on the edge behavior preference in the current user activity data based on the current user activity data as selected user activity data to obtain a basic identification core of the edge behavior preference; the selected user activity data is extracted according to a set extraction step length; based on the fact that the current user activity data are non-selected user activity data, carrying out preference matching processing on the edge behavior preference in the current user activity data to obtain a candidate identification core of the edge behavior preference;

wherein each current user activity data in the user interaction activity information set comprises a digital authentication signature; the obtaining a candidate identification core of the edge behavior preference by performing preference matching processing on the edge behavior preference in the current user activity data based on the current user activity data as the non-selected user activity data includes: adjusting a first preference matching processing model by combining the basic identification core of the edge behavior preference; the base identification kernel is determined in selected user activity data prior to the current user activity data by the digital authentication signature; estimating the distribution variable of the edge behavior preference in the current user activity data through the adjusted first preference matching processing model to obtain a candidate identification core of the edge behavior preference;

wherein, the matching the corresponding behavior preference item by combining the identification cores of the behavior preference items in the user interaction activity information set to obtain a matching tag and a matching record corresponding to the behavior preference item includes: loading the identification cores of all the behavior preference items in the user interaction activity information set to a second preference matching processing model to obtain a matching label of each behavior preference item; and determining a matching record corresponding to the behavior preference item based on the identification core corresponding to the same behavior preference item in a group of service interaction logs and the matching label corresponding to the behavior preference item.

5. The method of claim 1, wherein the no less than two types of behavior preferences include: the method comprises the following steps of browsing items of a virtual mall, cross-border e-commerce concern items, VR service preference items, MR service preference items and hot topic preference items;

the determining the upstream and downstream characteristics between the at least two types of behavior preference items by combining the directional mining information of the at least two types of behavior preference items comprises:

determining cross-border mall associated information corresponding to the same e-commerce preference item in combination with a distribution variable joint analysis result between the identification core of the virtual mall browsing item and the identification core of the cross-border e-commerce concern item;

determining somatosensory demand description characteristics between the cross-border e-commerce concern item and the VR service preference item by combining the result of the joint analysis of distribution variables between the recognition core of the cross-border e-commerce concern item and the recognition core of the VR service preference item;

determining a first item participation feature or a second item participation feature between the hotspot topic preferences and the MR service preferences in conjunction with distribution variables of the respective identification kernels of the hotspot topic preferences and the MR service preferences and respective match records of the hotspot topic preferences and the MR service preferences.

6. The method of claim 2, wherein the directional mining information further includes a quality coefficient of the behavior preference item, and the determining, according to a set information extraction rule, service session interaction information corresponding to the behavior preference item through the directional mining information of each behavior preference item includes: determining initial session interaction information by combining an identification core of a first behavior preference item based on the fact that the quality coefficient of the first behavior preference item in the current user activity data reaches a first quality score; wherein the first behavior preference is any behavior preference of the no less than two types of behavior preferences; recording the initial session interaction information into a temporary storage space of the first behavior preference item; determining that the session interaction information recorded in the temporary storage space is the service session interaction information corresponding to the first behavior preference item based on that the temporary storage space of the first behavior preference item meets the preset screening starting requirement;

wherein the entering of the initial session interaction information into the scratchpad space of the first behavior preference includes: on the basis that the number of the session interaction information recorded in the temporary storage space of the first behavior preference item does not reach a storage limit value, unconditionally recording the initial session interaction information into the temporary storage space of the first behavior preference item;

wherein the entering of the initial session interaction information into the scratchpad space of the first behavior preference includes: removing the first session interaction information in the temporary storage space on the basis that the number of the session interaction information recorded in the temporary storage space of the first behavior preference item reaches the storage limit value; the quality coefficient of the first session interaction information is lower than that of the initial session interaction information; and inputting the initial session interaction information into a temporary storage space of the first behavior preference item.

7. The method of claim 6, further comprising:

initializing the temporary storage space of the first behavior preference item based on the temporary storage space of the first behavior preference item reaching the screening start requirement.

8. The method of claim 7, wherein the preset screening initiation requirements comprise at least one of: screening timeliness requirements, screening intermittence requirements, matching time-consuming screening conditions, quality grade screening requirements and matching termination screening requirements;

wherein the screening timeliness requirement reflects a duration of uninterruptedly locating the first behavior preference to a first set duration; the screening intermittence requirement reflects that a gap time period matched with the first behavior preference reaches a set gap value; the matching time-consuming screening condition reflects that the total time length for matching the first behavior preference item reaches a second set time length; the second set time length is not less than the first set time length; the screening requirement of the goodness score reflects that the session interaction information of which the goodness coefficient reaches a second goodness score exists in the temporary storage space of the first behavior preference item; the second goodness score is greater than the first goodness score; the match termination filtering requirement reflects an overall level match of the first behavior preference until the collected service interaction log is completed.

9. The method of claim 1, wherein the determining a push decision requirement of the user interaction activity information set by combining the upstream and downstream characteristics between the at least two types of behavior preferences and the service session interaction information of each of the behavior preferences comprises: performing knowledge vector mining and/or element vector mining on the service session interaction information of each behavior preference item to obtain demand prediction information corresponding to the behavior preference item; combining the demand prediction information of the at least two types of behavior preference items by combining the upstream and downstream characteristics between the at least two types of behavior preference items; determining a push decision requirement of the user interaction activity information set based on the combined requirement prediction information of the at least two types of behavior preference items;

the behavior preference item is a virtual mall browsing item, and the number of the service session interaction information of the virtual mall browsing item is not less than 2; performing knowledge vector mining and/or element vector mining on the service session interaction information of each behavior preference item to obtain demand prediction information corresponding to the behavior preference item, including: respectively carrying out knowledge vector mining on the service session interaction information not less than two of the virtual mall browsing items to obtain not less than two knowledge description blocks; splicing the knowledge description blocks to obtain a linkage descriptor of the virtual mall browsing item; performing element vector mining on service session interaction information with the highest quality index in the service session interaction information not lower than the two pieces of service session interaction information to obtain an element description block of the virtual mall browsing item; and determining demand prediction information of the virtual mall browsing item based on the linkage descriptor of the virtual mall browsing item and the element description block of the virtual mall browsing item.

10. A user analytics push system comprising a processor and a memory; the processor is connected in communication with the memory, and the processor is configured to read the computer program from the memory and execute the computer program to implement the method of any one of claims 1 to 9.

11. A server, comprising a processor and a memory; the processor is connected in communication with the memory, and the processor is configured to read the computer program from the memory and execute the computer program to implement the method of any one of claims 1 to 9.