CN114463072A

CN114463072A - E-business service optimization method based on business demand AI prediction and big data system

Info

Publication number: CN114463072A
Application number: CN202210290845.3A
Authority: CN
Inventors: 李云
Original assignee: Individual
Current assignee: Individual
Priority date: 2022-03-23
Filing date: 2022-03-23
Publication date: 2022-05-10

Abstract

The embodiment of the application discloses an E-business service optimization method based on business demand AI prediction and a big data system, wherein an E-business matching dictionary corresponding to each target business demand is constructed based on each target business demand of a specified subscriber and the demand type of each target business demand, the corresponding E-business service content information is pushed to the specified subscriber based on the E-business matching dictionary corresponding to each target business demand, the feedback activity data of the specified subscriber aiming at the E-business content information is obtained, the E-business matching dictionary of each target business demand is optimized and updated based on the feedback activity data, thus on the basis of business demand mining, the information is pushed by taking the demand type as the generation dimension of an E-business matching field, the optimized updating of the E-business matching dictionary is carried out based on the feedback activity, and the matching degree of a subsequent E-business matching dictionary and the specified subscriber is improved, and the content fixed-point pushing experience is improved.

Description

E-business service optimization method based on business demand AI prediction and big data system

Technical Field

The application relates to the technical field of big data mining, in particular to a power grid business optimization method based on business demand AI prediction and a big data system.

Background

Currently, internet e-commerce is gradually becoming the main way for people to shop, and the prosperity of e-commerce drives the continuous upgrade of related industries. In order to cater to the characteristics of the big data era and better hold the business opportunities of the big data era, each e-commerce service provider is expected to occupy the future market by taking full account of internet finance and chasing big data.

The electronic commerce service provider needs to realize the operation target, and can not conduct the business demand prediction of related users aiming at the legal big data analysis of wide users, so that the reference basis can be improved for the subsequent content pushing. In the related art, a corresponding user representation is usually generated for each user, and the user representations can be reflected in the form of an e-commerce matching dictionary, so how to effectively improve the matching degree of the e-commerce matching dictionary and a specified subscription user and improve the content fixed-point push experience is a technical problem to be solved in the field.

Disclosure of Invention

The application provides a E-business service optimization method based on business demand AI prediction and a big data system.

In a first aspect, an embodiment of the present application provides a method for optimizing e-commerce services based on AI prediction of service demand, which is applied to a big data system, and includes:

determining the demand type of each target service demand of a designated subscriber based on the service demand distribution of the designated subscriber;

constructing an e-commerce matching dictionary corresponding to each target service requirement based on each target service requirement of the specified subscription user and the requirement type of each target service requirement, wherein the e-commerce matching dictionary comprises a plurality of e-commerce matching pushing fields which are arranged according to priority;

pushing corresponding e-commerce service content information to the appointed subscription user based on the e-commerce matching dictionary corresponding to each target service requirement;

and feedback activity data of the appointed subscription user aiming at the e-commerce service content information is obtained, and the e-commerce matching dictionary of each target service requirement is optimized and updated based on the feedback activity data.

In an exemplary independent concept of the first aspect, the distribution of the business demand of the designated subscribing user is determined by:

each combined operation dimension data in the business combined operation big data of the appointed subscription user is walked, and different combined operation dimension data are obtained by carrying out data tracking on the appointed subscription user based on different business operation dimensions;

if a key wandering node which accords with key wandering characteristics is analyzed in the wandering combined operation dimension data of each wandering, the analyzed key wandering node is used as a reference key wandering node of the appointed subscriber, and time-space domain information of the reference key wandering node in the wandering combined operation dimension data and first wandering node information of the reference key wandering node analyzed from the wandering combined operation dimension data are determined;

acquiring contact combination operation dimension data of the walking combination operation dimension data from the rest combination operation dimension data except the walking combination operation dimension data in the service combination operation big data, wherein the walking combination operation dimension data and the contact combination operation dimension data have a link attribute;

tracking the reference key migration node in the contact combination operation dimensional data according to mapping contact information between the migration combination operation dimensional data and the contact combination operation dimensional data and time-space domain information of the reference key migration node in the migration combination operation dimensional data;

generating second wandering node information of the reference key wandering node based on tracking information and first wandering node information of the reference key wandering node, and if the behavior tendency probability in the second wandering node information is not smaller than a target behavior tendency probability, judging that the reference key wandering node has a service demand tendency;

if the behavior tendency probability in the second wandering node information is smaller than the target behavior tendency probability, judging that the reference key wandering node does not have a service demand tendency;

after all combined operation dimension data in the service combined operation big data are migrated, obtaining service demand tendency information of P reference key migration nodes analyzed from all combined operation dimension data, wherein P is a positive integer;

and determining the service demand distribution of the appointed subscription user based on the service demand tendency information of the P reference key walking nodes.

For example, in an exemplary independent concept of the first aspect, the mapping contact information between the walk-combine operation dimension data and the contact-combine operation dimension data is one of mapping contact information in a service-combine contact map of the service-combine operation big data;

wherein, the establishing process of any mapping contact information in the service combination contact map comprises the following steps:

selecting one combined operation dimension data from the service combined operation big data of the appointed subscription user as a first combined operation dimension data, and selecting the combined operation dimension data which has a common eigenvector with the first combined operation dimension data from the service combined operation big data as a second combined operation dimension data;

determining a plurality of first combined operation activities in the first combined operation dimension data and a plurality of second combined operation activities in the second combined operation dimension data; a first combined operation activity corresponds to a second combined operation activity, and the combined operation activity is: positioning the time-space domain information of one combined operation behavior of the appointed subscription user in the combined operation dimension data to obtain operation behavior activity;

and calculating mapping contact information between the first combined operation dimensional data and the second combined operation dimensional data based on the time-space-domain vector of each first combined operation activity and the corresponding time-space-domain vector of the second combined operation activity.

For example, in an exemplary independent concept of the first aspect, the calculating mapping relation information between the first combined operation dimension data and the second combined operation dimension data based on the spatio-temporal vector of each first combined operation activity and the spatio-temporal vector of the corresponding second combined operation activity includes:

acquiring mapping contact characteristics to be generated, wherein the mapping contact characteristics to be generated comprise a plurality of mapping contact parameters to be generated;

based on the mapping connection characteristics to be generated, mapping and connecting the spatio-temporal vector of each first combined operation activity to a high-dimensional relation scene from a low-dimensional relation scene where the first combined operation dimensional data is located to obtain a high-dimensional spatio-temporal vector of each first combined operation activity, wherein the high-dimensional spatio-temporal vector of each first combined operation activity comprises a plurality of mapping connection parameters;

mapping and linking the high-dimensional time-space domain vector of each first combined operation activity to a low-dimensional relation scene where the second combined operation dimensional data is located to obtain a mapping and linking vector of each first combined operation activity, wherein the mapping and linking vector of each first combined operation activity comprises a plurality of mapping and linking parameters;

based on the mapping contact vector of each first combined operation activity and the matching template of the time-space domain vector of the corresponding second combined operation activity, determining the configuration information of each mapping contact parameter in the mapping contact characteristics to obtain target mapping contact characteristics;

and based on the target mapping contact characteristics, the target mapping contact characteristics are used as mapping contact information between the first combined operation dimension data and the second combined operation dimension data.

For example, in an exemplary independent concept of the first aspect, the tracking the reference key walking node in the linkage combination operation dimension data according to mapping linkage information between the walking combination operation dimension data and the linkage combination operation dimension data and time-space domain information of the reference key walking node in the walking combination operation dimension data includes:

mapping and linking the time-space domain information of the reference key walking node in the walking combined operation dimensional data to the contact combined operation dimensional data according to the mapping and linking information between the walking combined operation dimensional data and the contact combined operation dimensional data to obtain a mapping and linking partition;

if W key wandering nodes are analyzed in the contact combination operation dimension data, determining the time-space domain information of each key wandering node in the W key wandering nodes in the contact combination operation dimension data, wherein W is a positive integer;

calculating the time-space domain information of each key walking node in the contact combination operation dimension data and the time-space domain relevancy between the key walking node and the mapping contact subarea;

searching a time-space domain correlation degree which is larger than the target time-space domain correlation degree in the time-space domain correlation degrees obtained through calculation;

if the time-space domain relevancy degree which is larger than the target time-space domain relevancy degree is found, determining that the reference key wandering node is tracked in the contact combination operation dimension data;

and if the time-space domain relevancy which is larger than the target time-space domain relevancy is not found, determining that the reference key wandering node is not tracked in the contact combination operation dimension data.

For example, in an exemplary independent concept of the first aspect, the time-space domain information of each key walking node in the connection combination operation dimension data is expressed based on a first defined data area, and the mapping connection partition is expressed based on a second defined data area;

the calculating the time-space domain information of each key walking node in the contact combination operation dimension data and the time-space domain relevancy between the mapping contact partitions comprises the following steps:

calculating the time-space domain information of the W key wandering node in the contact combination operation dimension data, and the ratio of the influence parameter values between the W key wandering node and the mapping contact subarea, wherein W belongs to [1, W ];

and determining the time-space domain relevancy between the time-space domain information of the w key walking node in the contact combination operation dimension data and the mapping contact partition based on the influence parameter value ratio.

For example, in an exemplary independent concept of the first aspect, the first wander node information includes a behavior tendency probability of the reference key wander node, and the generating of the second wander node information of the reference key wander node based on the tracking information and the first wander node information of the reference key wander node includes:

if the tracking information represents that the reference key walking node is not tracked in the contact combination operation dimension data, reducing the behavior tendency probability in the first walking node information by a preset amplitude;

and taking the key wandering node information obtained after the reduction of the preset amplitude as second wandering node information of the reference key wandering node.

For example, in an exemplary independent concept of the first aspect, the generating second wander node information of the reference key wander node based on tracking information and first wander node information of the reference key wander node includes:

if the tracking information represents that the reference key wandering node is tracked in the contact combination operation dimension data, the key wandering node information of the reference key wandering node analyzed in the contact combination operation dimension data is used as contact wandering node information;

aggregating the first wandering node information of the reference key wandering node and the contact wandering node information to obtain second wandering node information of the reference key wandering node;

the first wandering node information comprises a behavior tendency probability of the reference key wandering node, and the contact wandering node information comprises a behavior tendency probability of the reference key wandering node;

the aggregating the first wandering node information of the reference key wandering node and the contact wandering node information to obtain second wandering node information of the reference key wandering node includes:

acquiring a first influence parameter value of the first walking node information and a second influence parameter value of the contact walking node information;

aggregating the behavior tendency probability in the first walking node information and the behavior tendency probability in the contact walking node information based on the first influence parameter value and the second influence parameter value to obtain a target behavior tendency probability;

loading the target behavior tendency probability to second wandering node information of the reference key wandering node;

or the first wandering node information includes the wandering node tag attribute of the reference key wandering node, the number of the contact combination operation dimensional data is N, the contact wandering node information of each contact combination operation dimensional data includes the wandering node tag attribute of the reference key wandering node, and N is an integer greater than 1;

if the attribute of the wandering node label in the first wandering node information is the same as the attribute of the wandering node label in each piece of contact wandering node information, taking the attribute of the wandering node label as the attribute of a target wandering node label, and loading the target wandering node label to second wandering node information of the reference key wandering node;

if the label attribute of the wandering node in at least one piece of contact wandering node information is different from the label attribute of the wandering node in the first wandering node information, counting the number of the label attributes of each wandering node, determining the label attribute of the wandering node with the largest number as the label attribute of the target wandering node, and loading the label attribute of the target wandering node to second wandering node information of the reference key wandering node;

or, the first wandering node information includes a behavior tendency probability of the reference key wandering node and a wandering node tag attribute of the reference key wandering node, and the contact wandering node information includes a wandering node tag attribute of the reference key wandering node, and the aggregating the first wandering node information of the reference key wandering node and the contact wandering node information to obtain second wandering node information of the reference key wandering node includes:

if the attribute of the wandering node tag in the first wandering node information is the same as the attribute of the wandering node tag in the contact wandering node information, improving the behavior tendency probability in the first wandering node information by a preset amplitude, and taking the wandering node information obtained after the preset amplitude is improved as second wandering node information of the reference key wandering node;

if the attribute of the wandering node tag in the first wandering node information is different from the attribute of the wandering node tag in the contact wandering node information, reducing the behavior tendency probability in the first wandering node information by a preset amplitude, and taking the wandering node information obtained after the reduction of the preset amplitude as second wandering node information of the reference key wandering node;

or the number of the contact combination operation dimensional data is N, wherein N is an integer greater than 1, and one piece of contact combination operation dimensional data corresponds to one piece of contact walking node information;

performing historical voting degree analysis on the first walking node information of the reference key walking node and the N pieces of contact walking node information to obtain the voting degree of the first walking node information and the voting degree of each piece of contact walking node information;

and selecting the key walking node information with the maximum voting degree from the first walking node information and the N pieces of contact walking node information as second walking node information of the reference key walking node.

In a second aspect, an embodiment of the present application provides a big data system, including:

a processor;

a memory, in which a computer program is stored, and when executed, the computer program implements the method for optimizing the e-commerce service based on the AI prediction of the service demand according to the first aspect.

Compared with the prior art, the E-commerce matching dictionary corresponding to each target service requirement is constructed based on each target service requirement of a specified subscriber and the requirement type of each target service requirement, corresponding E-commerce content information is pushed to the specified subscriber based on the E-commerce matching dictionary corresponding to each target service requirement, feedback activity data of the specified subscriber for the E-commerce content information is obtained, and the E-commerce matching dictionary corresponding to each target service requirement is optimized and updated based on the feedback activity data, so that information is pushed by taking the requirement type as the generation dimension of the E-commerce matching field on the basis of service requirement mining, the E-commerce matching dictionary is optimized and updated based on the feedback activity, the matching degree of a subsequent E-commerce matching dictionary and the specified subscriber is improved, and content fixed-point pushing experience is improved.

Drawings

Fig. 1 is a schematic flowchart illustrating steps of a e-commerce service optimization method based on AI prediction of service demand according to an embodiment of the present application;

fig. 2 is a schematic block diagram of a structure of a big data system for executing the e-commerce service optimization method based on the service demand AI prediction in fig. 1 according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments obtained by a person of ordinary skill in the art without any inventive step based on the embodiments in the present application are within the scope of protection of the present application.

Step S110, determining a requirement type of each target service requirement of the designated subscriber based on the service requirement distribution of the designated subscriber.

In this embodiment, the service demand distribution of the designated subscriber may represent the trigger frequency of each target service demand of the designated subscriber in each service plate, and then determine the demand type of each target service demand based on the trigger frequency, for example, the corresponding relationship between different trigger frequency intervals and corresponding demand types may be set, such as sufficient demand, irregular demand, and drop demand.

Step S120, constructing an e-commerce matching dictionary corresponding to each target service requirement based on each target service requirement of the specified subscription user and the requirement type of each target service requirement, wherein the e-commerce matching dictionary comprises a plurality of e-commerce matching pushing fields arranged according to priority.

In this embodiment, the e-commerce matching push fields for each target service requirement may be determined based on the requirement type of each target service requirement, if the requirement is sufficient, all the relevant e-commerce matching push fields under the target service requirement may be read, and arranged according to the heat priority, if the requirement is irregular, the e-commerce matching push fields relevant to the current time period under the target service requirement may be read, and arranged according to the heat priority, as the requirement is lowered, the e-commerce matching push fields relevant to the continuous increase of the current heat under the target service requirement may be read, and arranged according to the heat priority, which is not particularly limited.

And step S130, pushing corresponding e-commerce service content information to the appointed subscription user based on the e-commerce matching dictionary corresponding to each target service requirement.

In this embodiment, the e-commerce content information corresponding to the e-commerce matching push fields arranged according to the priority can be respectively obtained from different e-commerce service data sources based on the plurality of e-commerce matching push fields under each target service requirement, and pushed to the specified subscription user.

Step S140, obtaining feedback activity data of the appointed subscription user aiming at the e-commerce service content information, and carrying out optimization updating on the e-commerce matching dictionary of each target service demand based on the feedback activity data.

In this embodiment, after the e-commerce service content information is pushed, a series of feedback activities can be performed on the e-commerce matching dictionary by the related specified subscription user, so that the e-commerce matching dictionary of each target service requirement can be optimized and updated based on the feedback activity data, the matching degree of the subsequent e-commerce matching dictionary and the specified subscription user is improved, and the content fixed-point pushing experience is improved.

Based on the above steps, the present embodiment determines the requirement type of each target service requirement of the specified subscriber based on the service requirement distribution of the specified subscriber. Constructing an e-commerce matching dictionary corresponding to each target service requirement based on each target service requirement of the specified subscriber and the requirement type of each target service requirement, pushing corresponding e-commerce service content information to the specified subscriber based on the e-commerce matching dictionary corresponding to each target service requirement, and acquiring feedback activity data of the specified subscriber for the e-commerce service content information, and based on the feedback activity data, the E-commerce matching dictionary of each target service demand is optimized and updated, therefore, on the basis of business demand mining, information is pushed by taking the demand type as the generation dimension of the E-commerce matching field, the E-commerce matching dictionary is optimized and updated based on feedback activities, therefore, the matching degree of the subsequent e-commerce matching dictionary and the specified subscription user is improved, and the content fixed-point pushing experience is improved.

In an exemplary independently contemplated embodiment, step S140 may be implemented by the following exemplary steps.

Step A110, performing feature mining on a first feedback activity set of the feedback activity data to obtain feedback preference features of each feedback activity node in the first feedback activity set, where the first feedback activity set indicates a point of interest feedback activity.

In an exemplary design approach, the first feedback activity set may be obtained by arranging feedback activity nodes in the point of interest feedback activity according to a trigger order in the point of interest feedback activity set.

In an exemplary design idea, in order to make the preference generated for the point of interest feedback activity interpretable, a preference attribute corresponding to the point of interest feedback activity may also be used as a reference for generating the preference. In an exemplary design concept, before step a110, the method further includes: and performing feedback activity aggregation on the interest point feedback activity and the preference attribute corresponding to the interest point feedback activity to obtain the first feedback activity set. In addition, in order to facilitate distinguishing the feedback activity nodes in the point of interest feedback activity from the feedback activity nodes in the attribute representing preference, a connection feature may be further added to the point of interest feedback activity and the attribute representing preference, so that the first feedback activity set further includes the connection feature between the point of interest feedback activity and the attribute representing preference.

The preference attribute corresponding to the interest point feedback activity represents a preferred category attribute corresponding to the interest point feedback activity, and in an exemplary design idea, the preferred category attribute may include a positive category attribute indicating that the preference is positive and a negative category attribute indicating that the preference is negative; in other embodiments, the preferred category attribute may include an uncertain category attribute indicating that the preference attribute of the point of interest feedback activity is uncertain in addition to the positive category attribute and the negative category attribute.

In an exemplary design idea, the preference attribute corresponding to the feedback activity of the point of interest may be obtained by labeling the feedback activity of the point of interest by a user, that is, a user knowing the preference of the feedback activity of the point of interest labels the preference attribute based on the preference of the feedback activity of the point of interest, so as to obtain the preference attribute corresponding to the feedback activity of the point of interest.

In an exemplary design idea, the interest point feedback activity, the preference attribute corresponding to the interest point feedback activity, and the preference basis feedback activity corresponding to the interest point feedback activity may all be used as reference bases for generating the preference. In an exemplary design concept, before step a110, the method further includes: and performing feedback activity aggregation on the interest point feedback activity, the preference attribute corresponding to the interest point feedback activity and the preference corresponding to the interest point feedback activity according to the feedback activity to obtain the first feedback activity set. Further, in order to distinguish the point of interest feedback activities, the preference attributes, and the preference-based feedback activities, a connection feature may be added between two adjacent ones (e.g., between the point of interest feedback activities and the preference attributes, between the preference attributes and the preference-based feedback activities), so that the added connection feature is also included in the first feedback activity set.

In an exemplary design concept, step a110 includes: and performing feature mining on the first feedback activity set through a feedback preference feature extraction unit to obtain feedback preference features of feedback activity nodes in the first feedback activity set.

Step a120, based on the feedback preference characteristics of each feedback activity node in the first feedback activity set and the predicted interest point characteristics of the interest point prediction model at the real-time online service stage, determining an interest metric value corresponding to each feedback activity node in the first feedback activity set at the real-time online service stage.

The interest point prediction model is used for predicting based on the feedback preference characteristics of all feedback activity nodes in the first feedback activity set to obtain predicted interest point characteristics, and the predicted interest point characteristics are used for determining preference attention activities of corresponding business online stages. The preference attention activity refers to a feedback activity node in a preference event. In the process of generating the preference event, in order to ensure the dependency relationship between the feedback activity node and the feedback activity node in the preference event, the preference attention activity in the preference event is generated according to the service online stage.

Correspondingly, for the interest point prediction model, the predicted interest point characteristics of each service online stage are generated correspondingly according to the service online stage, so that the predicted interest point characteristics of the real-time service online stage are ensured to be used for determining preference attention activities corresponding to the real-time service online stage. That is, the interest point prediction model determines the predicted interest point characteristics corresponding to the preference attention activities in the preference events one by one, and a process of determining a preference attention activity is referred to as an online business stage. Because the interest point prediction model determines the predicted interest point characteristics corresponding to the preference attention activity according to the priority of the preference attention activity in the preference event, the predicted interest point characteristics generated by the interest point prediction model in the service online stage t (or the tth service online stage, wherein t is more than or equal to 1, and t is a positive integer) are the tth feedback activity nodes corresponding to the preference event.

In an exemplary design concept, before step a120, the method further includes: the transmission characteristics of the interest point prediction model in the real-time service online stage are obtained, wherein the transmission characteristics of the interest point prediction model in the real-time service online stage comprise the predicted interest point characteristics of the interest point prediction model in the last service online stage, and the transmission characteristics of the interest point prediction model in the first service online stage comprise potential characteristic information corresponding to a trigger tag; and processing the interest point prediction model based on the transmission characteristics of the real-time service online stage to generate the predicted interest point characteristics of the real-time service online stage.

In an exemplary design idea, the potential feature information corresponding to the trigger tag may be obtained by inputting an interest weight feature corresponding to the trigger tag into an interest point prediction model (e.g., single-layer LSTM), and generating, by the interest point prediction model, the potential feature information corresponding to the trigger tag based on the interest weight feature corresponding to the trigger tag. The interest weight feature corresponding to the trigger tag may be generated through a shared Embedding (Embedding) layer.

In an exemplary design idea, if the feedback preference feature extraction unit is constructed according to a long-time memory network, the input of the feedback preference feature extraction unit at each service online stage includes, in addition to the predicted interest point feature of the interest point prediction model at the previous service online stage, an interest point component generated at the previous service online stage, so that the feedback preference feature extraction unit performs processing according to the predicted interest point feature of the previous service online stage and the interest point component of the previous service online stage to obtain the predicted interest point feature of the real-time service online stage and the interest point component of the real-time service online stage; repeating the process, the interest point prediction model can correspondingly generate the predicted interest point characteristics corresponding to each service online stage.

In an exemplary design idea, an interest point component input by the interest point prediction model in a first service online stage may be an interest point component generated by the feedback preference feature extraction unit in a last service online stage in an encoding process, where the interest point component generated by the feedback preference feature extraction unit in the last service online stage in the encoding process represents feature mining information of the first feedback activity set.

In an exemplary design idea, the predicted interest point feature generated by the interest point prediction model in the online stage of the real-time service may represent a low-dimensional vector representation of a preferred interest activity corresponding to the online stage of the real-time service, and it may also be understood that the predicted interest point feature generated in the online stage of the real-time service indicates semantics of the preferred interest activity corresponding to the online stage of the real-time service, and the preferred interest activity generated in the online stage of the final real-time service is related to the predicted interest point feature generated by the interest point prediction model in the online stage of the real-time service.

The interest metric values corresponding to the feedback activity nodes in the first feedback activity set in the online stage of the real-time service are used for representing interest weights distributed to the feedback activity nodes in the first sequence of feedback activities for determining the preference attention activities in the online stage of the real-time service. It should be noted that the preference attention activity in the preference event is determined according to the feedback activity nodes in the first feedback activity set, and the influence weights of the feedback activity nodes in the first feedback activity set on different preference attention activities in the preference event are different, so that the interest metric values corresponding to the feedback activity nodes in the first feedback activity set at different on-line business stages are also different. Therefore, in the present solution, the interest metric value of each feedback activity node in the first feedback activity set at the corresponding on-line service stage is correspondingly determined according to the on-line service stage of the preference interest activity.

It should be noted that, because there is a difference in the influence weight of each feedback activity node in the first feedback activity set on the preference attention activity in each online service stage, in each online service stage for generating the preference attention activity, according to the step a120, the interest metric value of each feedback activity node in the first feedback activity set in each online service stage is determined in a targeted manner by combining the feedback preference characteristic of each feedback activity node in the first feedback activity set and the predicted interest point characteristic of the interest point prediction model in each online service stage, so as to ensure that the determined interest metric value accurately reflects the influence weight on the preference attention activity in each online service stage.

Step A130, determining first support degree information based on the interest metric value corresponding to each feedback activity node in the first feedback activity set in the real-time service online stage, the predicted interest point characteristic of the interest point prediction model in the real-time service online stage, and the transfer characteristic of the interest point prediction model in the real-time service online stage, where the first support degree information represents a first support degree of a preferred interest activity corresponding to the real-time service online stage from a historical frequent activity set and a second support degree from the first feedback activity set.

Because the preference corresponding to the interest point feedback activity is derived from the preference basis feedback activity, the support degree of the feedback activity node in the preference event corresponding to the interest point feedback activity derived from the preference basis feedback activity is higher, and therefore, under the condition that the preference basis feedback activity corresponding to the interest point feedback activity is also used as a reference basis for generating the preference event corresponding to the interest point feedback activity, the feedback activity node can also be called from the preference basis feedback activity corresponding to the interest point feedback activity as the feedback activity node in the preference event.

The historical frequent activity set may be selected based on actual needs.

In an exemplary design concept, step a130 includes:

step a210, based on the feedback preference characteristics corresponding to each feedback activity node in the first feedback activity set and the interest metric value corresponding to each feedback activity node in the first feedback activity set at the real-time service online stage, determining the interest join characteristics corresponding to the real-time service online stage.

In an exemplary design concept, step a210 includes: and taking the interest metric value corresponding to each feedback activity node in the first feedback activity set in the online stage of the real-time service as the connection attribute of the corresponding feedback activity node, and performing knowledge entity connection on the feedback preference characteristics of all the feedback activity nodes in the first feedback activity set to obtain the interest connection characteristics corresponding to the online stage of the real-time service.

Step A220, aggregating the interest connection characteristics corresponding to the real-time service online stage, the predicted interest point characteristics of the interest point prediction model in the real-time service online stage, and the transmission characteristics of the interest point prediction model in the real-time service online stage to obtain first aggregation characteristics.

Step a230, performing a support degree decision on the first aggregated feature to obtain the first support degree.

In an exemplary design approach, a first support degree may be generated by performing a non-linear feature transformation on a first aggregated feature by a fully connected layer.

Step A240, determining the second support degree based on the first support degree, wherein the sum of the first support degree and the second support degree is 1.

The preference attention activity in the preference event is derived from either the first feedback activity set or the historical frequent activity set, so the sum of the first support degree and the second support degree is set to be 1, and after the first support degree is determined, the difference between 1 and the first support degree is the second support degree.

Step A140, based on the first support degree and the second support degree, determining a preference interest activity corresponding to the online stage of the real-time service in the historical frequent activity set and the first feedback activity set, and performing optimization updating on the e-commerce matching dictionary of each target service requirement based on e-commerce preference fields related to the preference interest activity corresponding to the online stage of the real-time service and the preference interest degree corresponding to each e-commerce preference field; the preference attention activity is used for determining a preference event corresponding to the interest point feedback activity.

The first support degree indicates the support degree of the preference attention activity corresponding to the online stage of the real-time service from the historical frequent activity set; the second support degree indicates the support degree of the preference interest activity corresponding to the real-time service online stage from the first feedback activity set, so that the target support degree of the preference interest activity corresponding to the real-time service online stage of the feedback activity nodes in the first feedback activity set and each feedback activity node in the historical frequent activity set can be further calculated according to the first support degree and the second support degree, and then the preference interest activity corresponding to the real-time service online stage is determined in the historical frequent activity set and the first feedback activity set according to the target support degree.

It should be noted that the same feedback active node may exist in the first feedback active set and the historical frequent active set, for example, a feedback active node (assumed to be a feedback active node E) exists in both the first feedback active set and the historical frequent active set, and then the target support degree corresponding to the feedback active node E in the real-time service online stage needs to be calculated by combining the first support degree and the second support degree at the same time.

In an exemplary design concept, step a140 includes:

step A310, obtaining second support degree information, where the second support degree information indicates that each feedback activity node in the historical frequent activity set is a reference support degree of a preference attention activity corresponding to the real-time service online stage.

In an exemplary design concept, step A310 includes: aggregating the interest connection characteristics corresponding to the real-time service online stage and the predicted interest point characteristics of the interest point prediction model in the real-time service online stage to obtain second aggregation characteristics; and carrying out nonlinear feature conversion on the second aggregation feature, and carrying out support degree decision according to a nonlinear feature conversion result to obtain second support degree information.

Step a320, based on the first support degree and the second support degree, performing knowledge entity linkage on the second support degree information and the interest metric value corresponding to each feedback activity node in the first feedback activity set in the online stage of the real-time service, and determining target support degree information, where the target support degree information indicates that each feedback activity node in the historical frequent activity set and the first feedback activity set is a target support degree of a preferred interest activity corresponding to the online stage of the real-time service.

For example, in step a320, the first support degree is used as a connection attribute of a second support degree information item, the second support degree is used as a connection attribute of an interest metric value item corresponding to each feedback activity node in the first feedback activity set in the real-time service online stage, and the second support degree information and the interest metric value corresponding to each feedback activity node in the first feedback activity set in the real-time service online stage are weighted to obtain target support degree information.

In an exemplary design idea, an interest metric value corresponding to a feedback activity node in a first feedback activity set at a real-time online service stage may be regarded as a reference support degree of a preference interest activity corresponding to the real-time online service stage of the feedback activity node in the first feedback activity set, and therefore, if a feedback activity node E1 only exists in the first feedback activity set and the history frequent activity set does not include the feedback activity node E1, the target support degree of the preference interest activity corresponding to the real-time online service stage of the feedback activity node E1 is equal to a fusion value of a second support degree and the reference support degree of the preference interest activity corresponding to the real-time online service stage of the feedback activity node E1; similarly, if a feedback active node E2 exists only in the historical frequent active set, and the first feedback active set does not include the feedback active node E2, the target support degree of the preference interest activity corresponding to the real-time service online stage by the feedback active node E2 may be equal to the fusion value of the reference support degrees corresponding to the feedback active node E2 indicated by the first support degree and the second support degree information; if a feedback activity node E3 is not only located in the first feedback activity set, but also exists in the historical frequent activity set, the feedback activity node E3 is that the target support degree of the preference interest activity corresponding to the real-time online service stage is equal to the sum of the first target support degree and the second target support degree, where the first target support degree is equal to the fusion value of the second support degree and the interest metric value of the preference interest activity corresponding to the real-time online service stage by the feedback activity node E3; the second target support degree is equal to the fusion value of the first support degree and the reference support degree corresponding to the feedback active node E3 indicated by the second support degree information.

Step A330, based on the target support degree, screening in the historical frequent activity set and the first feedback activity set, and determining a preference attention activity corresponding to the real-time online service stage.

Through the step a320, the target support degree of the preference attention activity corresponding to the real-time online service stage of each feedback activity node in the historical frequent activity set and the first feedback activity set can be calculated, and thus, the feedback activity node with the highest target support degree in the first feedback activity set and the historical frequent activity set can be determined as the preference attention activity corresponding to the real-time online service stage.

In the scheme, a first support degree that the preferred concerned activities corresponding to the real-time service online stage are from a historical frequent activity set and a second support degree that the preferred concerned activities corresponding to the real-time service online stage are from the first feedback activity set are determined based on the interest metric value of each feedback activity node in the first feedback activity set for the preferred concerned activities corresponding to the real-time service online stage, the predicted interest point characteristics of an interest point prediction model in the real-time service online stage and the transmission characteristics of the interest point prediction model in the real-time service online stage, and determining the preference attention activity corresponding to the online stage of the real-time service from the historical frequent activity set and the first feedback activity set according to the first support degree and the second support degree, determining a preference event corresponding to the interest point feedback activities indicated by the first feedback activity set according to the preference attention activities; in the scheme, the inventor finds that the preference for the interest point feedback activity is from the characteristic of high support degree of the interest point feedback activity, so that the association between the preference event and the interest point feedback activity is enhanced according to the mode of generating the preference event by the preference attention activity determined by the first support degree and the second support degree, and the accuracy of the generated preference event is effectively improved.

In an exemplary design concept, step a120 includes:

step A410, aggregating the feedback preference characteristics of each feedback activity node in the first feedback activity set with the predicted interest point characteristics of the interest point prediction model in the real-time online service stage, so as to obtain third aggregation characteristics corresponding to each feedback activity node in the first feedback activity set.

Step a420, performing nonlinear characteristic conversion on each third aggregation characteristic to obtain a nonlinear conversion characteristic corresponding to each feedback activity node in the first feedback activity set.

Step A430, performing activation processing on each nonlinear conversion characteristic to obtain an initial interest metric value of each feedback activity node in the first feedback activity set for a preference interest activity corresponding to a real-time service online stage.

Step A440, performing regularized conversion on each initial interest metric value to obtain an interest metric value of a preference interest activity corresponding to a real-time service online stage by each feedback activity node in the first feedback activity set.

In an exemplary design idea, each initial interest metric value may be subjected to regularization conversion through a softmax function, and a result of the regularization conversion is used as an interest metric value of a preference interest activity corresponding to a real-time service online stage by a corresponding feedback activity node.

Through the steps A410-S440, the interest metric value of the preference interest activity corresponding to each feedback active node in the first feedback active set in the real-time online business stage is calculated on the basis of the feedback preference characteristic of each feedback active node in the first feedback active set and the predicted interest point characteristic of the interest point prediction model in the real-time online business stage.

After the target support degree is obtained, the feedback activity node with the maximum target support degree in the first feedback activity set and the historical frequent activity set can be determined as the preference attention activity corresponding to the service online stage. And then, combining the preference attention activities corresponding to the online phases of the businesses by the preference event generation layer to obtain the preference events corresponding to the interest point feedback activities.

In an exemplary design concept, after step a140, the method further comprises: associating the interest point feedback activity, the preference attribute corresponding to the interest point feedback activity and the preference event corresponding to the interest point feedback activity to obtain interest point preference data; and loading the interest point preference data to an interest point preference database.

The interest point feedback activities and the preference attributes corresponding to the interest point feedback activities are input into a preference generation model, and the preference generation model processes the interest point feedback activities and the preference attributes corresponding to the interest point feedback activities according to the method of the application to obtain preference events corresponding to the interest point feedback activities. The preference generation model may then generate a data combination of the point of interest feedback activities, preference attributes, preference events, and then store the data combination in a point of interest preference database.

In an exemplary design concept, the method further comprises:

step A510, receiving a point of interest preference calling instruction, wherein the point of interest preference calling instruction indicates a target point of interest.

Step A520, performing interest point matching in the interest point preference database based on the target interest point, and determining target interest point preference data of which the interest point feedback activity is matched with the target interest point.

In an exemplary design idea, the interest point matching may be to calculate a matching degree between an interest point feedback activity in the interest point preference data and the target interest point, and then determine the target interest point preference data according to the calculated matching degree.

In an exemplary design idea, the interest point preference data where the interest point feedback activity matching with the target interest point to the highest degree is located may be determined as the target interest point preference data. In other embodiments, the interest point preference data where the interest point feedback activity with the matching degree with the target interest point exceeding the matching degree threshold may also be determined as the target interest point preference data. In other embodiments, the interest point preference data may be sorted according to the matching degree from high to low, and the interest point preference data located in the top set number in the sorting may be determined as the target interest point preference data.

Step A530, returning the preference event in the target interest point preference data to the initiator of the interest point preference calling instruction. And returning the preference event in the target interest point preference data to the initiator of the interest point preference calling instruction as a search result of the target interest point.

In an exemplary independent concept, the distribution of the service demand of the designated subscriber in the foregoing step S110 may be obtained by the following exemplary steps.

And step B101, migrating each combined operation dimension data in the service combined operation big data of the appointed subscription user.

The big data system carries out key wandering node mining on each combined operation dimension data in the business combined operation big data of the appointed subscription user one by one, and different combined operation dimension data are obtained by carrying out data tracking on the appointed subscription user based on different business operation dimensions. In an exemplary design idea, performing key walk node mining on each combined operation dimension data refers to: performing key wandering node mining on each combined operation dimension data through a key wandering node mining network to obtain key wandering node mining information of each combined operation dimension data; the key wandering node mining network is obtained by performing network weight optimization on the AI structure network by referring to a training data sequence (marking a training data sequence of a relevant key wandering node). For example, reference training data is input into the AI structural network, and network weight information of the AI structural network is adjusted based on a difference value between decision information of the AI structural network and labeling information of the reference training data, so as to obtain a key wandering node mining network.

And step B102, if the key wandering node which accords with the key wandering characteristics is analyzed in the wandering combined operation dimension data of each wandering, taking the analyzed key wandering node as a reference key wandering node of a specified subscription user, and determining the time-space domain information of the reference key wandering node in the wandering combined operation dimension data and the first wandering node information of the reference key wandering node analyzed from the wandering combined operation dimension data.

The first wandering node information of the reference key wandering node is used for characterizing related information of the reference key wandering node, and the first wandering node information of the reference key wandering node may include, but is not limited to, a behavior tendency probability of the reference key wandering node and a wandering node label attribute of the reference key wandering node.

If the big data system analyzes a key wandering node in the wandering combined operation dimension data of each wandering (if the key wandering node is analyzed in the wandering combined operation dimension data output by the key wandering node mining network), the analyzed key wandering node is used as a reference key wandering node of a specified subscriber, time-space domain information of the reference key wandering node in the wandering combined operation dimension data and first wandering node information of the reference key wandering node analyzed from the wandering combined operation dimension data are determined based on the key wandering node mining information; for example, the key wandering node mining information carries a time-space domain of the key wandering node, the big data system determines time-space domain information of a reference key wandering node in the wandering combination operation dimension data based on the time-space domain, and summarizes relevant information such as behavior tendency probability of the reference key wandering node, wandering node label attribute and the like as first wandering node information of the reference key wandering node.

And step B103, acquiring the connection combination operation dimension data of the wandering combination operation dimension data from the rest combination operation dimension data except the wandering combination operation dimension data in the service combination operation big data.

The connection combination operation dimension data of the wandering combination operation dimension data refers to: and in the rest combined operation dimension data except the wandering combined operation dimension data in the service combined operation big data, the combined operation dimension data which has the connection attribute with the wandering combined operation dimension data. The mapping contact information is determined through service combination analysis of service combination operation big data, and the mapping contact information between the wandering combination operation dimensional data and the contact combination operation dimensional data is one of the mapping contact information in the service combination contact map of the service combination operation big data. In an exemplary design idea, the process of establishing any mapping contact information in a business combination contact map includes:

selecting one combined operation dimensional data from the service combined operation big data of the appointed subscriber as a first combined operation dimensional data, and selecting the combined operation dimensional data which has a feature vector in common with the first combined operation dimensional data from the service combined operation big data as a second combined operation dimensional data (the common feature vector refers to the feature vector which appears in the two combined operation dimensional data at the same time); determining a plurality of first combined operation activities in the first combined operation dimension data and a plurality of second combined operation activities in the second combined operation dimension data; a first combined operation activity corresponds to a second combined operation activity, and the combined operation activity is: positioning time-space domain information of a combined operation behavior of a designated subscriber in combined operation dimension data to obtain operation behavior activity; for example, if the first combined operation dimension data and the second combined operation dimension data both include combined operation behavior 1-combined operation behavior 4 of a specified subscription user, the time-space domain information of the combined operation behavior 1 in the first combined operation dimension data is labeled as first combined operation activity 1, the time-space domain information of the combined operation behavior 1 in the second combined operation dimension data is labeled as second combined operation activity 1, and the first combined operation activity 1 corresponds to the second combined operation activity 1; similarly, a first combined operation activity 2-a first combined operation activity 4 can be calibrated according to the time-space domain information of the combined operation activity 2-the combined operation activity 4 in the first combined operation dimensional data, and a second combined operation activity 2-a second combined operation activity 4 can be calibrated according to the time-space domain information of the combined operation activity 2-the combined operation activity 4 in the second combined operation dimensional data. Calculating mapping contact information between the first combination operation dimensional data and the second combination operation dimensional data based on the time-space-domain vector of each first combination operation activity and the corresponding time-space-domain vector of the second combination operation activity; for example, the mapping connection characteristics between the first combination operation dimension data and the second combination operation dimension data may be determined by the plurality of time-space-domain vectors of the first combination operation activities and the time-space-domain vectors of the second combination operation activities corresponding to the respective combination operation flows, so as to obtain the mapping connection information between the first combination operation dimension data and the second combination operation dimension data.

And step B104, tracking the reference key migration node in the contact combination operation dimensional data according to the mapping contact information between the migration combination operation dimensional data and the contact combination operation dimensional data and the time-space domain information of the reference key migration node in the migration combination operation dimensional data.

And the big data system maps and links the time-space domain information of the reference key walking node in the walking combined operation dimensional data to the contact combined operation dimensional data according to the mapping contact information between the walking combined operation dimensional data and the contact combined operation dimensional data to obtain a mapping contact partition. The big data system searches the connection combination operation dimension data according to the mapping connection partition to determine whether a reference key wandering node can be found in the connection combination operation dimension data, for example:

if W key wandering nodes are analyzed in the contact combination operation dimension data, determining the time-space domain information of each key wandering node in the W key wandering nodes in the contact combination operation dimension data, wherein W is a positive integer; calculating the time-space domain information of each key walking node in the contact combination operation dimension data and the time-space domain relevancy between the key walking nodes and the mapping contact partitions; the time-space domain relevancy can be specifically obtained by calculating an influence parameter value ratio of the time-space domain information of each key walking node in the connection combination operation dimensional data to the time-space domain information of the mapping connection partition (the same time-space domain of the time-space domain corresponding to the time-space domain information of each key walking node in the connection combination operation dimensional data and the time-space domain corresponding to the mapping connection partition is divided by the merged time-space domain), and the like. After the time-space domain information of each key wandering node in the contact combination operation dimension data and the time-space domain relevancy between the mapping contact partitions are obtained, the big data system searches the time-space domain relevancy which is greater than the target time-space domain relevancy in the time-space domain relevancy obtained through calculation; if the time-space domain relevancy degree which is larger than the target time-space domain relevancy degree is found, determining that a reference key wandering node is tracked in the contact combination operation dimension data; and if the time-space domain relevancy which is larger than the target time-space domain relevancy is not found, determining that the reference key wandering node is not tracked in the contact combination operation dimension data.

Accordingly, if the key wander node is not resolved in the contact combination operation dimension data (i.e., the key wander node is not present in the contact combination operation dimension data), it is determined that the reference key wander node is not tracked in the contact combination operation dimension data.

And step B105, generating second wandering node information of the reference key wandering node based on the tracking information and the first wandering node information of the reference key wandering node, and deciding the service demand tendency of the reference key wandering node according to the second wandering node information.

The tracking information is used to characterize whether a reference key wandering node is tracked in the contact combination operation dimension data, and if the tracking information characterizes that the reference key wandering node is tracked in the contact combination operation dimension data, the tracking information may further include key wandering node information (such as behavior tendency probability of the reference key wandering node, wandering node label attribute, and the like) of the reference key wandering node.

In an exemplary design idea, the first wandering node information includes a behavior tendency probability of a reference key wandering node, if the tracking information represents that the reference key wandering node is not tracked in the contact combination operation dimension data, the big data system reduces the behavior tendency probability in the first wandering node information by a preset amplitude (for example, by 0.1), and uses the key wandering node information obtained after the reduction of the preset amplitude as the second wandering node information of the reference key wandering node. Correspondingly, if the tracking information represents that the reference key wandering node is tracked in the contact combination operation dimension data, the big data system takes the first wandering node information of the reference key wandering node analyzed in the contact combination operation dimension data as second wandering node information, and aggregates the first wandering node information of the reference key wandering node and the contact wandering node information (for example, aggregates the behavior tendency probability of the reference key wandering node), so as to obtain the second wandering node information of the reference key wandering node.

In another exemplary design idea, the first wandering node information includes a wandering node tag attribute referring to a key wandering node, the number of contact combination operation dimension data is N, N is an integer greater than 1, and the tracking information of each contact combination operation dimension data includes contact wandering node information; performing historical voting degree analysis on first walking node information of a reference key walking node and N pieces of contact walking node information to obtain the voting degree of the first walking node information and the voting degree of each piece of contact walking node information; and selecting the key walking node information with the maximum voting degree from the first walking node information and the N pieces of contact walking node information as second walking node information of the reference key walking node.

After second wandering node information of the reference key wandering node is obtained, the big data system judges whether the behavior tendency probability in the second wandering node information is larger than the target behavior tendency probability. If the behavior tendency probability in the second wandering node information is not smaller than the target behavior tendency probability, judging that the reference key wandering node has a service demand tendency; and if the behavior tendency probability in the second wandering node information is smaller than the target behavior tendency probability, judging that the reference key wandering node does not have the service demand tendency.

According to the embodiment of the application, when the key wandering node is analyzed in the wandering combined operation dimension data of each wandering, the analyzed key wandering node can be used as a reference key wandering node of a specified subscriber, and further, the reference key wandering node is tracked in the contact combined operation dimension data according to mapping contact information between the wandering combined operation dimension data and related contact combined operation dimension data and time-space domain information of the reference key wandering node in the wandering combined operation dimension data, so that second wandering node information of the reference key wandering node is generated on the basis of the tracking information and first wandering node information of the reference key wandering node tracked from the wandering combined operation dimension data, and service requirement tendency of the reference key wandering node is decided according to the second wandering node information. By the scheme of analyzing the combined key wandering node by connecting the combined operation dimension data to the reference key wandering node in the wandering combined operation dimension data, whether the reference key wandering node is the key wandering node with the actual service demand mining value or not can be accurately analyzed, so that the error of mining the key wandering node is reduced, and the accuracy of mining the service demand tendency is improved.

According to the above description of the business demand prediction method using big data deep mining, the embodiment of the present application provides another business demand prediction method using big data deep mining, which may include the following steps B201 to B208:

and step B201, performing service combination analysis on the service combination operation big data of the appointed subscription user.

In an exemplary design idea, selecting one combination operation dimension data from service combination operation big data of a specified subscription user as first combination operation dimension data, and selecting combination operation dimension data having a common feature vector with the first combination operation dimension data from the service combination operation big data as second combination operation dimension data; determining a plurality of first combined operation activities in the first combined operation dimension data and a plurality of second combined operation activities in the second combined operation dimension data; a first combined operation activity corresponds to a second combined operation activity, and the combined operation activity is: positioning time-space domain information of a combined operation behavior of a designated subscriber in combined operation dimension data to obtain operation behavior activity; for example, if the first combined operation dimension data and the second combined operation dimension data both include combined operation behavior 1-combined operation behavior 4 of a specified subscription user, the time-space domain information of the combined operation behavior 1 in the first combined operation dimension data is labeled as first combined operation activity 1, the time-space domain information of the combined operation behavior 1 in the second combined operation dimension data is labeled as second combined operation activity 1, and the first combined operation activity 1 corresponds to the second combined operation activity 1; similarly, a first combined operation activity 2-a first combined operation activity 4 may be calibrated according to the time-space domain information of the combined operation activity 2-the combined operation activity 4 in the first combined operation dimension data, and a second combined operation activity 2-a second combined operation activity 4 may be calibrated according to the time-space domain information of the combined operation activity 2-the combined operation activity 4 in the second combined operation dimension data.

Determining a plurality of first combined operation activities in the first combined operation dimensional data, and after determining a plurality of second combined operation activities in the second combined operation dimensional data, acquiring a mapping contact characteristic to be generated by the big data system, wherein the mapping contact characteristic to be generated comprises a plurality of mapping contact parameters to be generated; based on the mapping connection characteristics to be generated, mapping and connecting the time-space domain vector of each first combined operation activity to a high-dimensional relation scene from a low-dimensional relation scene where the first combined operation dimension data is located to obtain a high-dimensional time-space domain vector of each first combined operation activity, wherein the high-dimensional time-space domain vector of each first combined operation activity comprises a plurality of mapping connection parameters; mapping and linking the high-dimensional time-space domain vector of each first combined operation activity to a low-dimensional relation scene where second combined operation dimensional data is located to obtain a mapping and linking vector of each first combined operation activity, wherein the mapping and linking vector of each first combined operation activity comprises a plurality of mapping and linking parameters; and matching the mapping contact vector of each first combined operation activity with the matching template of the time-space domain vector of the corresponding second combined operation activity, determining the configuration information of each mapping contact parameter in the mapping contact characteristics, obtaining target mapping contact characteristics, and taking the target mapping contact characteristics as the mapping contact information between the first combined operation dimensional data and the second combined operation dimensional data.

For example, the combined operation behavior in the common feature vector of the first combined operation dimension data may be calibrated to obtain a first combined operation activity 1-a first combined operation activity 4, and the combined operation behavior in the common feature vector of the second combined operation dimension data may be calibrated to obtain a second combined operation activity a-a second combined operation activity d; wherein the first combined operation activity 1 corresponds to the second combined operation activity a, the first combined operation activity 2 corresponds to the second combined operation activity b, the first combined operation activity 3 corresponds to the second combined operation activity c, and the first combined operation activity 4 corresponds to the second combined operation activity d; and determining the mapping relation characteristics between the first combination operation dimension data and the second combination operation dimension data according to each first combination operation activity and the second combination operation activity corresponding to the first combination operation activity, and obtaining the mapping relation information between the first combination operation dimension data and the second combination operation dimension data.

According to the above description of step B201, it is assumed that the service composition operation big data of the designated subscriber includes composition operation dimension data 1-composition operation dimension data 5. After the service combination analysis is completed, the combination operation dimension data 1-5 in the service combination operation big data of the appointed subscription user can be associated through mapping contact characteristics; for example, the point map in the combined operation dimension data 1 may be related to the combined operation dimension data 2 by the map related feature M12, and the point map in the combined operation dimension data 2 may also be related to the combined operation dimension data 1 by the map related feature M21.

And step B202, migrating each combined operation dimension data in the business combined operation big data of the appointed subscription user.

Step B203, if a key wandering node conforming to the key wandering characteristics is analyzed in the wandering combination operation dimension data of each wandering, taking the analyzed key wandering node as a reference key wandering node of the designated subscriber, and determining time-space domain information of the reference key wandering node in the wandering combination operation dimension data and first wandering node information of the reference key wandering node analyzed from the wandering combination operation dimension data.

And step B204, acquiring the connection combination operation dimension data of the migration combination operation dimension data from the rest combination operation dimension data except the migration combination operation dimension data in the service combination operation big data.

The specific implementation schemes of step B204 and step B205 may refer to the implementation schemes of step B102 and step B103, and are not described herein again.

And step B205, tracking the reference key migration node in the contact combination operation dimensional data according to the mapping contact information between the migration combination operation dimensional data and the contact combination operation dimensional data and the time-space domain information of the reference key migration node in the migration combination operation dimensional data.

According to mapping contact information between the walk combination operation dimensional data and the contact combination operation dimensional data, the big data system maps and contacts the time-space domain information of the reference key walk node in the walk combination operation dimensional data to the contact combination operation dimensional data to obtain a mapping contact partition. For example, mapping relation is performed on the basis of mapping relation information between the walking combined operation dimensional data and the connection combined operation dimensional data by referring to time-space domain information of a key walking node in the walking combined operation dimensional data, and a mapping relation partition of the key walking node in the connection combined operation dimensional data is obtained.

The big data system searches the connection combination operation dimension data according to the mapping connection partition to determine whether a reference key wandering node can be found in the connection combination operation dimension data, for example:

if W key wandering nodes are analyzed in the contact combination operation dimensional data, determining the time-space domain information of each key wandering node in the W key wandering nodes in the contact combination operation dimensional data, wherein W is a positive integer; calculating the time-space domain information of each key walking node in the contact combination operation dimension data and the time-space domain relevancy between the key walking nodes and the mapping contact partitions; the time-space domain relevancy can be obtained by calculating the ratio of the influence parameter values of the time-space domain information of each key wandering node in the contact combination operation dimension data and the time-space domain information between the mapping contact partitions.

In an exemplary design idea, time-space domain information of each key walking node in the W key walking nodes in the connection combination operation dimensional data is expressed based on a first limited data area, a mapping connection partition is expressed based on a second limited data area, the big data system calculates a ratio of an influence parameter value of the second limited data area to each first limited data area, and determines the ratio of the influence parameter value of the second limited data area to each first limited data area as the time-space domain information of the key walking node in the connection combination operation dimensional data and the time-space domain correlation degree between the key walking node and the mapping connection partition.

In another exemplary design idea, time-space domain information of each key walking node in the W key walking nodes in the contact combination operation dimension data is expressed based on a first defined data area, a mapping contact partition is expressed based on a second defined data area, the big data system calculates the intersection degree of the second defined data area and each first defined data area, and determines the intersection degree of the second defined data area and each first defined data area as the time-space domain information of the key walking node in the contact combination operation dimension data and the time-space domain association degree between the key walking node and the mapping contact partition.

In another exemplary design idea, the big data system obtains a collection time-space domain of each key walking node in the contact combination operation dimensional data and a collection time-space domain of the mapping contact partition, calculates a distance between the collection time-space domain of each key walking node and the collection time-space domain of the mapping contact partition, and determines the distance between the collection time-space domain of the mapping contact partition and the collection time-space domain of each key walking node as time-space domain information of the key walking node in the contact combination operation dimensional data and a time-space domain correlation degree between the key walking node and the mapping contact partition.

After the time-space domain information of each key walking node in the contact combination operation dimension data and the time-space domain relevancy between the mapping contact subareas are obtained, if the time-space domain relevancy is determined based on the influence parameter value ratio of the second limited data area to each first limited data area or the intersection degree of the second limited data area and each first limited data area, the big data system searches the time-space domain relevancy which is greater than the target time-space domain relevancy in the calculated time-space domain relevancy; if the time-space domain relevancy degree which is larger than the target time-space domain relevancy degree is found, determining that a reference key wandering node is tracked in the contact combination operation dimension data; and if the time-space domain relevancy which is larger than the target time-space domain relevancy is not found, determining that the reference key wandering node is not tracked in the contact combination operation dimension data.

If the time-space domain relevancy is determined based on the distances between the aggregated time-space domain of the mapping contact partition and the aggregated time-space domain of each key wandering node, the big data system searches the time-space domain relevancy which is smaller than the target time-space domain relevancy from the calculated time-space domain relevancy; if the time-space domain relevancy degree which is larger than the target time-space domain relevancy degree is found, determining that a reference key wandering node is tracked in the contact combination operation dimension data; and if the time-space domain relevancy which is larger than the target time-space domain relevancy is not found, determining that the reference key wandering node is not tracked in the contact combination operation dimension data.

Taking the analysis of the interest points as an example, mapping and linking the time-space domain information of the candidate interest points in the migration combined operation dimensional data to the linkage combined operation dimensional data according to the mapping and linking information between the migration combined operation dimensional data and the linkage combined operation dimensional data to obtain mapping and linking partitions of the candidate interest points; if the candidate interest points are analyzed in the contact combination operation dimensional data, determining the time-space domain information of the candidate interest points in the contact combination operation dimensional data; calculating the time-space domain information of the candidate interest points in the contact combination operation dimension data and the time-space domain correlation degree between the candidate interest points and the mapping contact partitions to obtain tracking information; for example, if the time-space domain relevance between the time-space domain information of the candidate interest point in the contact combination operation dimensional data and the mapping contact partition is greater than the target time-space domain relevance, finding the time-space domain relevance greater than the target time-space domain relevance, and judging that the candidate interest point is analyzed in the contact combination operation dimensional data; correspondingly, if the time-space domain relevance between the time-space domain information of the candidate interest point in the contact combination operation dimension data and the mapping contact partition is less than or equal to the target time-space domain relevance, the time-space domain relevance larger than the target time-space domain relevance is not found, and the candidate interest point is judged not to be resolved in the contact combination operation dimension data.

Further, if the key wander node is not resolved in the contact combining operation dimension data (i.e., the key wander node is not present in the contact combining operation dimension data), it is determined that the reference key wander node is not tracked in the contact combining operation dimension data.

Step B206, generating second wandering node information of the reference key wandering node based on the tracking information and the first wandering node information of the reference key wandering node, and deciding the service demand tendency of the reference key wandering node according to the second wandering node information.

In an exemplary design idea, the first wandering node information includes a behavior tendency probability of a reference key wandering node, if the tracking information represents that the reference key wandering node is not tracked in the contact combination operation dimension data, the big data system reduces the behavior tendency probability in the first wandering node information by a preset amplitude (for example, by 0.1), and uses the key wandering node information obtained after the reduction of the preset amplitude as the second wandering node information of the reference key wandering node. Correspondingly, if the tracking information represents that the reference key wandering node is tracked in the contact combination operation dimensional data, the big data system takes the key wandering node information of the reference key wandering node analyzed in the contact combination operation dimensional data as contact wandering node information, and aggregates the first wandering node information of the reference key wandering node and the contact wandering node information to obtain second wandering node information of the reference key wandering node. For example, the first walking node information and the contact walking node information both include behavior tendency probabilities of reference key walking nodes, the big data system acquires a first influence parameter value of the first walking node information and a second influence parameter value of the contact walking node information, and aggregates the behavior tendency probability in the first walking node information and the behavior tendency probability in the contact walking node information based on the first influence parameter value and the second influence parameter value to obtain target behavior tendency probabilities, and loads the target behavior tendency probabilities to the second walking node information of the reference key walking nodes; for example, the behavior tendency probability of the reference key walking node in the first walking node information is set to be 0.8, and the first influence parameter value is set to be 1; the behavior tendency probability of the reference key walking node in the information of the contact walking nodes is 0.4, the second influence parameter value is 0.2, and the target behavior tendency probability =0.8 × 1+0.4 × 0.2= 0.88.

In another exemplary design idea, the number of the contact combination operation dimension data is N, where N is an integer greater than 1, the first wandering node information and the contact wandering node information of each contact combination operation dimension data further include a wandering node tag attribute referring to a key wandering node, and if the wandering node tag attribute in the first wandering node information and the wandering node tag attribute in each contact wandering node information are the same, the wandering node tag attribute in the first wandering node information is taken as a target wandering node tag attribute and loaded to second wandering node information referring to the key wandering node; and if the label attribute of the wandering node in at least one piece of contact wandering node information is different from the label attribute of the wandering node in the first wandering node information, counting the number of the label attributes of each wandering node, determining the label attribute of the wandering node with the largest number as the label attribute of the target wandering node, and loading the label attribute of the target wandering node into second wandering node information of the reference key wandering node.

By combining the two implementation schemes, the first wandering node information and the contact wandering node information of each contact combination operation dimension data both include behavior tendency probability and wandering node label attribute, and the first wandering node information and the contact wandering node information are aggregated to obtain second wandering node information, namely the second wandering node information includes both target behavior tendency probability and target wandering node label attribute.

In still another exemplary design concept, the first wandering node information includes behavior tendency probabilities of the reference key wandering nodes and wandering node tag attributes of the reference key wandering nodes, and the contact wandering node information includes wandering node tag attributes of the reference key wandering nodes. If the attribute of the wandering node tag in the first wandering node information is the same as the attribute of the wandering node tag in the contact wandering node information, improving the behavior tendency probability in the first wandering node information by a preset amplitude (for example, increasing the behavior tendency probability by 0.1), and taking the description information (including the increased behavior tendency probability and the wandering node tag attribute of the reference key wandering node) obtained after the improvement of the preset amplitude as the second wandering node information of the reference key wandering node; if the attribute of the wandering node tag in the first wandering node information is different from the attribute of the wandering node tag in the contact wandering node information, reducing the behavior tendency probability in the first wandering node information by a preset amplitude (for example, reducing the behavior tendency probability by 0.1), and taking the wandering node information (including the reduced behavior tendency probability and the wandering node tag attribute of the reference key wandering node) obtained after the reduction of the preset amplitude as second wandering node information of the reference key wandering node; for example, in the first wandering node information of the wandering combination operation dimension data, the behavior tendency probability of the reference key wandering node 1 is 0.7, and the wandering node label attribute is class 1; if the walk node label attribute of the reference key walk node 1 in the contact walk node information is class 1, the behavior tendency probability of the reference key walk node 1= 0.7+0.1= 0.8; if the walk node tag attribute of the reference key walk node 1 in the contact walk node information is category 1, the behavior tendency probability of the reference key walk node 1= 0.7-0.1= 0.6. And taking the behavior tendency probability after the reference key walking node 1 is updated as second walking node information of the reference key walking node 1. It should be noted that the above numerical values are only examples and do not constitute practical limitations of the present application.

In yet another exemplary design idea, the first wandering node information includes a wandering node tag attribute referring to a key wandering node, the number of contact combination operation dimension data is N, N is an integer greater than 1, and the tracking information of each contact combination operation dimension data includes contact wandering node information; performing historical voting degree analysis on first walking node information of a reference key walking node and N pieces of contact walking node information to obtain the voting degree of the first walking node information and the voting degree of each piece of contact walking node information; and selecting the key wandering node information with the maximum voting degree from the first wandering node information and the N pieces of contact wandering node information as second wandering node information of the reference key wandering node.

In an exemplary design idea, the first wandering node information and the contact wandering node information both include behavior tendency probabilities of reference key wandering nodes, and if the behavior tendency probability of the reference key wandering node in the first wandering node information or the contact wandering node information is greater than or equal to the target behavior tendency probability, the reference key wandering node is approved to have a service requirement tendency; if the behavior tendency probability of the reference key walking node in the first walking node information or the contact walking node information is smaller than the target behavior tendency probability, the reference key walking node is rejected to have a service requirement tendency, and second walking node information of the reference key walking node is generated based on a voting result; for example, if the target behavior tendency probability is set to be 0.6 and the behavior tendency probability of the reference key walking node in the first walking node information is set to be 0.7, the vote is complied with according to the first walking node information; and if the behavior tendency probability of the reference key walking node in the contact walking node information 1 is 0.3, a rejection ticket is thrown according to the contact walking node information 1. After second wandering node information of the reference key wandering node is obtained, if the historical voting degree for agreeing that the reference key wandering node has the service demand tendency is greater than or equal to the historical voting degree for disagreeing that the reference key wandering node has the service demand tendency, the reference key wandering node is judged to have the service demand tendency; and if the historical voting degree agreeing that the reference key walking node has the service demand tendency is less than the historical voting degree disagreeing that the reference key walking node has the service demand tendency, judging that the reference key walking node does not have the service demand tendency.

In another exemplary design idea, the first walking node information and the contact walking node information both include a walking node tag attribute of a reference key walking node, a historical voting degree of each walking node tag attribute is counted, a walking node tag attribute with the highest historical voting degree is determined as a target walking node tag attribute, and the target walking node tag attribute and a historical voting rate of the target walking node tag attribute are used as second walking node information of the reference key walking node; for example, assuming that the historical voting degree of the migration node tag attribute 1 is 10 votes and the historical voting degree of the migration node tag attribute 2 is 30 votes, the migration node tag attribute 2 is determined as the target migration node tag attribute. Further, if the historical voting rate of the target wandering node tag attribute is greater than or equal to the historical voting rate threshold, determining that the reference key wandering node has a service demand tendency, and determining the category of the reference key wandering node as the target wandering node tag attribute; if the historical voting rate of the target wandering node label attribute is smaller than the historical voting rate threshold value, judging that the reference key wandering node does not have a service demand tendency; for example, assuming that the historical voting rate threshold is 60%, if the historical voting rate of the target wandering node tag attribute is 75%, it is determined that the reference key wandering node has a service demand tendency, and the category of the reference key wandering node is the target wandering node tag attribute; accordingly, if the historical voting rate of the target wander node tag attribute is 55%, it is determined that the reference critical wander node does not have a traffic demand tendency.

In another exemplary design idea, the first wandering node information and the contact wandering node information include a behavior tendency probability of a reference key wandering node and a wandering node tag attribute of the reference key wandering node, and a voting is performed by integrating the behavior tendency probability of the reference key wandering node and the wandering node tag attribute in the first wandering node information or the contact wandering node information (for example, if the behavior tendency probability of the reference key wandering node is greater than a target behavior tendency probability and the wandering node tag attribute belongs to a preset wandering node tag attribute, the reference key wandering node is voted to have a service demand tendency, otherwise, the reference key wandering node is objected to have a service demand tendency), and second wandering node information of the reference key wandering node is generated based on a voting result. After second wandering node information of the reference key wandering node is obtained, if the historical voting degree for agreeing that the reference key wandering node has the service demand tendency is greater than or equal to the historical voting degree for disagreeing that the reference key wandering node has the service demand tendency, the reference key wandering node is judged to have the service demand tendency; and if the historical voting degree agreeing that the reference key walking node has the service demand tendency is less than the historical voting degree disagreeing that the reference key walking node has the service demand tendency, judging that the reference key walking node does not have the service demand tendency.

And step B207, counting service demand tendency information of P reference key wandering nodes analyzed from all combined operation dimension data, wherein P is a positive integer.

And counting the number of the reference key wandering nodes analyzed from each combined operation dimension data in the service combined operation big data set, and obtaining the service requirement tendency of each reference key wandering node through the steps B202 to B206.

And step B208, determining the service demand distribution of the appointed subscription user based on the service demand tendency information of the P reference key walking nodes.

According to the same inventive concept, as shown in fig. 2, the embodiment of the present application further provides a big data system, and the big data system 100 may have a relatively large difference due to different configurations or performances, and may include one or more Central Processing Units (CPUs) 112 (e.g., one or more processors) and a memory 111. Wherein the memory 111 may be a transient storage or a persistent storage. The program stored in memory 111 may include one or more modules, each of which may include a sequence of instructions operating on the big data system 100. Further, central processor 112 may be configured to communicate with memory 111 to perform a series of instruction operations in memory 111 on big data system 100.

Big data system 100 may also include one or more power supplies, one or more communication units 113, one or more pass-to-output interfaces, and/or one or more operating systems, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, and the like.

In addition, a storage medium is provided in an embodiment of the present application, and the storage medium is used for storing a computer program, and the computer program is used for executing the method provided in the embodiment.

The embodiment of the present application also provides a computer program product including instructions, which when run on a computer, causes the computer to execute the method provided by the above embodiment.

Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium may be at least one of the following media: various media that can store program codes, such as Read-only Memory (ROM), RAM, magnetic disk, or optical disk.

It should be noted that, in the present specification, each embodiment is described in a progressive manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, they are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described embodiments of the apparatus and system are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement without inventive effort.

The above description is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A E-business service optimization method based on business demand AI prediction is applied to a big data system, and is characterized by comprising the following steps:

constructing an e-commerce matching dictionary corresponding to each target service requirement based on each target service requirement of the specified subscription user and the requirement type of each target service requirement, wherein the e-commerce matching dictionary comprises a plurality of e-commerce matching push fields which are arranged according to priority;

2. The E-commerce service optimization method based on AI prediction of service demand as claimed in claim 1, wherein said step of optimally updating E-commerce matching dictionaries of said respective target service demands based on said feedback activity data comprises:

performing feature mining on a first feedback activity set of the feedback activity data to obtain feedback preference features of feedback activity nodes in the first feedback activity set, wherein the first feedback activity set indicates interest point feedback activities;

based on the feedback preference characteristics of each feedback activity node in the first feedback activity set and the predicted interest point characteristics of the interest point prediction model in the real-time service online stage, determining an interest metric value corresponding to each feedback activity node in the first feedback activity set in the real-time service online stage;

determining first support degree information based on interest metric values corresponding to feedback activity nodes in the first feedback activity set in a real-time service online stage, predicted interest point characteristics of the interest point prediction model in the real-time service online stage and transmission characteristics of the interest point prediction model in the real-time service online stage, wherein the first support degree information represents a first support degree of preference interest activities corresponding to the real-time service online stage from a historical frequent activity set and a second support degree from the first feedback activity set;

determining preference attention activities corresponding to the real-time service online stage in the historical frequent activity set and the first feedback activity set based on the first support degree and the second support degree, and performing optimization updating on the e-commerce matching dictionaries of the target service demands based on e-commerce preference fields related to the preference attention activities corresponding to the real-time service online stage and preference attention degrees corresponding to each e-commerce preference field; the preference attention activity is used for determining a preference event corresponding to the interest point feedback activity.

3. The method according to claim 2, wherein the determining first support information based on the interest metric value corresponding to each feedback activity node in the first feedback activity set in the real-time online service stage, the predicted interest point characteristic of the interest point prediction model in the real-time online service stage, and the transfer characteristic of the interest point prediction model in the real-time online service stage includes:

determining interest connection characteristics corresponding to the real-time service online stage based on feedback preference characteristics corresponding to each feedback activity node in the first feedback activity set and interest metric values corresponding to each feedback activity node in the first feedback activity set in the real-time service online stage;

aggregating interest connection characteristics corresponding to the real-time service online stage, predicted interest point characteristics of the interest point prediction model in the real-time service online stage and transmission characteristics of the interest point prediction model in the real-time service online stage to obtain first aggregation characteristics;

carrying out support degree decision on the first aggregation characteristic to obtain the first support degree;

determining the second support degree based on the first support degree, wherein the sum of the first support degree and the second support degree is 1;

the determining, based on the feedback preference characteristics corresponding to the feedback activity nodes in the first feedback activity set and the interest metric values corresponding to the feedback activity nodes in the first feedback activity set at the online stage of the real-time service, the interest connection characteristics corresponding to the online stage of the real-time service includes:

and taking the interest metric value corresponding to each feedback activity node in the first feedback activity set in the online stage of the real-time service as the connection attribute of the corresponding feedback activity node, and performing knowledge entity connection on the feedback preference characteristics of all the feedback activity nodes in the first feedback activity set to obtain the interest connection characteristics corresponding to the online stage of the real-time service.

4. The method according to claim 2, wherein before determining the interest metric value corresponding to each feedback activity node in the first feedback activity set in the real-time online business stage based on the feedback preference characteristic of each feedback activity node in the first feedback activity set and the predicted interest point characteristic of the interest point prediction model in the real-time online business stage, the method further comprises:

the transmission characteristics of the interest point prediction model in the real-time service online stage are obtained, wherein the transmission characteristics of the interest point prediction model in the real-time service online stage comprise the predicted interest point characteristics of the interest point prediction model in the last service online stage, and the transmission characteristics of the interest point prediction model in the first service online stage comprise potential characteristic information corresponding to a trigger tag;

and processing the interest point prediction model based on the transmission characteristics of the real-time service online stage to generate the predicted interest point characteristics of the real-time service online stage.

5. The E-business service optimization method based on AI prediction of business needs of claim 2, wherein the determining of the preferred activities of interest corresponding to the online phase of real-time business in the historical frequent activity set and the first feedback activity set based on the first support degree and the second support degree comprises:

acquiring second support degree information, wherein the second support degree information represents the reference support degree of each feedback activity node in a historical frequent activity set, which is a preferred attention activity corresponding to the real-time service online stage;

performing knowledge entity linkage on the second support degree information and interest metric values corresponding to feedback activity nodes in the first feedback activity set in a real-time service online stage based on the first support degree and the second support degree, and determining target support degree information, wherein the target support degree information indicates that the feedback activity nodes in the historical frequent activity set and the first feedback activity set are target support degrees of preference interest activities corresponding to the real-time service online stage;

based on the target support degree, screening in the historical frequent activity set and the first feedback activity set, and determining preference attention activities corresponding to the online stage of the real-time service;

wherein the obtaining the second support degree information includes:

aggregating the interest connection characteristics corresponding to the real-time service online stage and the predicted interest point characteristics of the interest point prediction model in the real-time service online stage to obtain second aggregation characteristics;

and carrying out nonlinear feature conversion on the second aggregation feature, and carrying out support degree decision according to a nonlinear feature conversion result to obtain second support degree information.

6. The method of claim 2, wherein the determining the interest metric value corresponding to each feedback activity node in the first feedback activity set in the real-time online business stage based on the feedback preference characteristic of each feedback activity node in the first feedback activity set and the predicted interest point characteristic of the interest point prediction model in the real-time online business stage comprises:

aggregating the feedback preference characteristics of each feedback activity node in the first feedback activity set with the predicted interest point characteristics of the interest point prediction model in the real-time service online stage respectively to obtain third aggregation characteristics corresponding to each feedback activity node in the first feedback activity set;

performing nonlinear characteristic conversion on each third aggregation characteristic to obtain a nonlinear conversion characteristic corresponding to each feedback activity node in the first feedback activity set;

activating each nonlinear conversion characteristic to obtain an initial interest metric value of each feedback activity node in the first feedback activity set for a preference interest activity corresponding to a real-time service online stage;

and carrying out regularized conversion on each initial interest metric value to obtain an interest metric value corresponding to each feedback activity node in the first feedback activity set in the online stage of the real-time service.

7. The E-commerce service optimization method based on business demand AI prediction as claimed in claim 2, wherein said feature mining a first feedback activity set of the feedback activity data to obtain feedback preference features of feedback activity nodes in the first feedback activity set comprises:

and performing feedback preference feature extraction on the first feedback activity set through a feedback preference feature extraction unit to obtain feedback preference features of all feedback activity nodes in the first feedback activity set.

8. The operator service optimization method based on AI prediction, according to any of the claims 1 to 7, characterized in that the distribution of the service needs of the appointed subscriber is determined by the following steps:

9. The E-commerce service optimization method based on business demand AI prediction as claimed in claim 8, wherein the mapping contact information between the wandering combined operation dimension data and the contact combined operation dimension data is one of the mapping contact information in the business combined contact map of the business combined operation big data;

10. A big data system, comprising:

a processor;

a memory in which a computer program is stored, which when executed implements the method for optimizing electric business service based on AI prediction of service demand according to any one of claims 1 to 9.