CN113449012A - Internet service mining method based on big data prediction and big data prediction system - Google Patents
Internet service mining method based on big data prediction and big data prediction system Download PDFInfo
- Publication number
- CN113449012A CN113449012A CN202110682187.8A CN202110682187A CN113449012A CN 113449012 A CN113449012 A CN 113449012A CN 202110682187 A CN202110682187 A CN 202110682187A CN 113449012 A CN113449012 A CN 113449012A
- Authority
- CN
- China
- Prior art keywords
- data
- interest
- internet service
- mining
- big data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2471—Distributed queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Probability & Statistics with Applications (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the disclosure provides an internet service mining method based on big data prediction and a big data prediction system, wherein an interest prediction index is determined by referring to a mining index set of each key internet data of internet service big data in a reference data set corresponding to an interest business behavior prediction network, and a mining reference interval of the interest prediction index can be determined according to the mining index set corresponding to each key internet data, so that the interest prediction index corresponding to each target interest business behavior is better matched with the real interest tendency characteristics of the internet service big data corresponding to the target interest business behavior, noise errors generated when the mining index is determined for each key internet data through the interest prediction index are reduced, and the precision of the estimation of the mining index set in the internet service big data is improved.
Description
Technical Field
The disclosure relates to the technical field of big data prediction, in particular to an internet service mining method based on big data prediction and a big data prediction system.
Background
With the rapid development of social informatization, data research becomes more and more complex no matter the change rate of data or new types of data are continuously updated, which means that the era of big data comes. Data mining integrates multiple categories of theories and technologies, such as high-performance computing, machine learning, artificial intelligence, pattern recognition, statistics, data visualization, database technologies, expert systems, and the like. The big data era is both opportunistic and challenging for data mining, big data is analyzed, a proper system is established, continuous optimization is realized, and the decision accuracy is improved, so that the big data era is more beneficial to mastering and conforms to the multi-end change of the market. In the big data era, data mining is recognized in various fields as the most common data analysis means, and the application of technologies such as classification, optimization, recognition, prediction and the like in data mining in a plurality of fields is mainly researched at present.
Based on the above, the inventor researches and discovers that the current data mining scheme can generate a part of noise errors when mining indexes of each key internet data are determined through interest estimation indexes (such as estimated mining reference indexes), so that the accuracy of estimation of a mining index set in internet service big data is insufficient.
Disclosure of Invention
In order to overcome at least the above disadvantages in the prior art, the present disclosure is directed to providing an internet service mining method and a big data prediction system based on big data prediction.
In a first aspect, the present disclosure provides an internet service mining method based on big data prediction, which is applied to a big data prediction system, where the big data prediction system is in communication connection with a plurality of internet service terminals, and the method includes:
acquiring internet service big data of the internet service terminal;
performing feature extraction on the Internet service big data based on a network extraction node of an interest service behavior prediction network to obtain operation behavior feature distribution corresponding to the Internet service big data;
based on an interest prediction node of the interest business behavior prediction network, performing interest prediction on the operation behavior feature distribution to obtain target interest business behaviors corresponding to all key internet data in the internet service big data;
acquiring mining value indexes corresponding to all key internet data in the internet service big data based on target interest business behaviors corresponding to all key internet data in the internet service big data and interest prediction indexes corresponding to all the target interest business behaviors;
acquiring a mining index set corresponding to the internet service big data based on mining value indexes corresponding to all key internet data in the internet service big data;
the interest business behavior prediction network is an artificial intelligence learning network obtained by training by taking reference internet service big data in a reference data set as training data and taking reference interest business behaviors of all key internet data of the reference internet service big data as training attributes; and the interest pre-estimation indexes corresponding to the target interest business behaviors are determined based on the mining index set of each key internet data of the reference internet service big data.
The present disclosure provides an internet service mining method based on big data prediction, the method comprising: acquiring a reference data set, wherein the reference data set comprises reference internet service big data and reference interest business behaviors corresponding to all key internet data in the reference internet service big data respectively; performing feature extraction on the reference internet service big data based on a network extraction node of an interest service behavior prediction network to obtain reference operation behavior feature distribution corresponding to the reference internet service big data; based on the reference operation behavior feature distribution, performing interest prediction through an interest prediction node of an interest service behavior prediction network to obtain prediction results corresponding to each key internet data in the reference internet service big data; training the interest business behavior prediction network based on the prediction results corresponding to the key internet data in the reference internet service big data and the reference interest business behaviors corresponding to the reference key internet service data in the reference internet service big data; the trained interest business behavior prediction network is used for obtaining target interest business behaviors of all key internet data in the internet service big data according to the input internet service big data and determining a mining index set corresponding to the internet service big data according to interest estimation indexes corresponding to all the target interest business behaviors; and the interest prediction index corresponding to each target interest business behavior is determined based on a mining index set of each key internet data of the reference data set of the interest business behavior prediction network and the reference internet service big data.
In one embodiment, the method further comprises: acquiring class reference key internet data; the similar reference key internet data is key internet data which can be corresponding to similar reference interest business behaviors in the key internet data of each reference internet service big data in the reference data set; the class reference interest business behavior is any one of the reference interest business behaviors; and determining interest estimation indexes corresponding to the class reference interest business behaviors based on mining index sets corresponding to all key internet data in the class reference key internet data respectively.
In an embodiment, the determining an interest prediction index corresponding to the reference-like interest service behavior based on a mining index set corresponding to each piece of key internet data in the reference-like key internet data includes: and determining interest estimation indexes corresponding to the class reference interest business behaviors based on preset sorting range mining indexes of a mining index set corresponding to each key internet data in the class reference key internet data.
In a second aspect, the disclosed embodiment further provides an internet service mining system based on big data prediction, where the internet service mining system based on big data prediction includes a big data prediction system and a plurality of internet service terminals communicatively connected to the big data prediction system;
the big data prediction system is configured to:
acquiring internet service big data of the internet service terminal;
performing feature extraction on the Internet service big data based on a network extraction node of an interest service behavior prediction network to obtain operation behavior feature distribution corresponding to the Internet service big data;
based on an interest prediction node of the interest business behavior prediction network, performing interest prediction on the operation behavior feature distribution to obtain target interest business behaviors corresponding to all key internet data in the internet service big data;
acquiring mining value indexes corresponding to all key internet data in the internet service big data based on target interest business behaviors corresponding to all key internet data in the internet service big data and interest prediction indexes corresponding to all the target interest business behaviors;
acquiring a mining index set corresponding to the internet service big data based on mining value indexes corresponding to all key internet data in the internet service big data;
the interest business behavior prediction network is an artificial intelligence learning network obtained by training by taking reference internet service big data in a reference data set as training data and taking reference interest business behaviors of all key internet data of the reference internet service big data as training attributes; and the interest pre-estimation indexes corresponding to the target interest business behaviors are determined based on the mining index set of each key internet data of the reference internet service big data.
According to any one of the aspects, in the embodiment provided by the disclosure, the interest prediction index is determined by referring to the mining index set of each key internet data of the internet service big data in the reference data set corresponding to the interest service behavior prediction network, and the mining reference interval of the interest prediction index can be determined according to the mining index set corresponding to each key internet data, so that the interest prediction index corresponding to each target interest service behavior more matches the real interest tendency characteristics of the internet service big data corresponding to the target interest service behavior, the noise error generated when the mining index is determined for each key internet data through the interest prediction index is reduced, and the precision of the mining index set estimation in the internet service big data is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings that need to be called in the embodiments are briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present disclosure, and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
Fig. 1 is a schematic view of an application scenario of an internet service mining system based on big data prediction according to an embodiment of the present disclosure;
FIG. 2 is a schematic flow chart of a big data prediction-based Internet service mining method according to an embodiment of the present disclosure;
fig. 3 is a functional module schematic diagram of an internet service mining device based on big data prediction according to an embodiment of the present disclosure;
fig. 4 is a block diagram illustrating a structure of a big data prediction system for implementing the above-described internet service mining method based on big data prediction according to an embodiment of the present disclosure.
Detailed Description
The following describes in detail aspects of embodiments of the present disclosure with reference to the drawings attached hereto. In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular system structures, interfaces, techniques, etc. in order to provide a thorough understanding of the particular embodiments of the disclosure.
Fig. 1 is a schematic view of a scenario of an internet service mining system 10 based on big data prediction according to an embodiment of the present disclosure. The big data prediction-based internet service mining system 10 may include a big data prediction system 100 and an internet service terminal 200 communicatively connected to the big data prediction system 100. The big data forecast based internet service mining system 10 shown in fig. 1 is only one possible example, and in other possible embodiments, the big data forecast based internet service mining system 10 may also include only at least some of the components shown in fig. 1 or may also include other components.
In this embodiment, the big data prediction system 100 and the internet service terminal 200 in the big data prediction-based internet service mining system 10 may cooperatively perform the big data prediction-based internet service mining method described in the following method embodiment, and the detailed description of the following method embodiment may be referred to for the execution steps of the big data prediction system 100 and the internet service terminal 200.
In order to solve the technical problem in the foregoing background, fig. 2 is a schematic flowchart of a big data prediction-based internet service mining method provided in an embodiment of the present disclosure, which can be executed by the big data prediction system 100 shown in fig. 1, and the details of the big data prediction-based internet service mining method are described below.
And step S101, obtaining Internet service big data of the Internet service terminal.
In this embodiment, the internet service big data may refer to operation behavior big data generated by a user of the internet service terminal in a process of using an internet service (such as an intelligent medical service, an intelligent office service, an intelligent education service, and the like), such as behavior data of information viewing, forwarding, sharing, subscribing, downloading, and the like.
And S102, performing feature extraction on the Internet service big data based on a network extraction node of the interest business behavior prediction network to obtain operation behavior feature distribution corresponding to the Internet service big data.
In this embodiment, the interest business behavior prediction network may be a neural network based on deep learning, the network extraction node may be configured to extract feature vector information of the internet service big data, and further obtain operation behavior feature distribution corresponding to the internet service big data, where the operation behavior feature distribution may refer to feature distribution information of each operation behavior feature under a reference feature index of a time sequence (time class dimension) or a null sequence (business class dimension).
Step S103, based on the interest prediction node of the interest business behavior prediction network, performing interest prediction on the operation behavior feature distribution to obtain target interest business behaviors corresponding to each key Internet data in the Internet service big data.
In this embodiment, the target interest business behavior may be used to represent business behaviors in which each piece of key internet data in the internet service big data may represent a tendency of a user to be interested in any business object.
Step S104, acquiring mining value indexes corresponding to all key Internet data in the Internet service big data based on target interest business behaviors corresponding to all key Internet data in the Internet service big data and interest estimation indexes corresponding to all the target interest business behaviors.
In this embodiment, each target interest service behavior may correspond to an interest prediction index in advance, and the interest prediction index may be used to represent a mining tendency reference degree of the target interest service behavior, for example, if an index value corresponding to the interest prediction index is larger, it indicates that the probability of mining the data content related to the target interest service behavior is larger.
Step S105, acquiring a mining index set corresponding to the internet service big data based on the mining value index corresponding to each key internet data in the internet service big data.
The interest business behavior prediction network is an artificial intelligent learning network obtained by training by taking reference internet service big data in a reference data set as training data and taking reference interest business behaviors of all key internet data of the reference internet service big data as training attributes; the interest prediction index corresponding to each target interest business behavior is determined based on a mining index set of each key internet data of the reference data set corresponding to the interest business behavior prediction network and referring to the internet service big data.
In summary, the present disclosure trains an interest business behavior prediction network by referring to internet service big data and reference interest business behaviors of each key internet data in the reference internet service big data to obtain a trained interest business behavior prediction network, determines target interest business behaviors of each reference key internet service data of the input internet service big data and interest prediction indexes corresponding to each target interest business behavior through the interest business behavior prediction network, and determines a mining index set of the internet service big data. By the scheme, the interest prediction index is determined by referring to the mining index set of each key internet data of the internet service big data in the reference data set corresponding to the interest service behavior prediction network, and the mining reference interval of the interest prediction index can be determined according to the mining index set corresponding to each key internet data, so that the interest prediction index corresponding to each target interest service behavior is more matched with the real interest tendency characteristics of the internet service big data corresponding to the target interest service behavior, the noise error generated when the mining index is determined for each key internet data through the interest prediction index is reduced, and the precision of the estimation of the mining index set in the internet service big data is improved.
The following continues to describe a flow of a big data prediction-based internet service mining method according to an exemplary embodiment of the present disclosure, where the flow of the big data prediction-based internet service mining method may include the following steps.
Step S201, acquiring a reference data set; the reference data set comprises reference internet service big data and reference interest business behaviors corresponding to all key internet data in the reference internet service big data respectively.
Step S202, based on the network extraction node of the interest business behavior prediction network, extracting the characteristics of the reference Internet service big data to obtain the reference operation behavior characteristic distribution corresponding to the reference Internet service big data.
Step S203, based on the reference operation behavior feature distribution, performing interest prediction by using an interest prediction node of the interest service behavior prediction network, and obtaining prediction results corresponding to each key internet data in the reference internet service big data.
Step S204, based on the prediction results corresponding to the key Internet data in the reference Internet service big data and the reference interest business behaviors corresponding to the reference key Internet service data in the reference Internet service big data, training the interest business behavior prediction network.
In this embodiment, the trained interest business behavior prediction network is configured to obtain a target interest business behavior of each key internet data in the internet service big data according to the input internet service big data, and determine a mining index set corresponding to the internet service big data according to an interest prediction index corresponding to each target interest business behavior; the interest prediction index corresponding to each target interest business behavior is determined based on a mining index set of each key internet data of the reference data set of the interest business behavior prediction network and the internet service big data.
Therefore, the method trains the interest business behavior prediction network by referring to the internet service big data and the reference interest business behaviors of all key internet data in the reference internet service big data to obtain the trained interest business behavior prediction network, determines the target interest business behaviors of all the reference key internet service data of the input internet service big data and the interest prediction indexes corresponding to all the target interest business behaviors through the interest business behavior prediction network, and determines the mining index set of the internet service big data. By the scheme, the interest prediction index is determined by referring to the mining index set of each key internet data of the internet service big data in the reference data set corresponding to the interest service behavior prediction network, and the mining reference interval of the interest prediction index can be determined according to the mining index set corresponding to each key internet data, so that the interest prediction index corresponding to each target interest service behavior is more matched with the real interest tendency characteristics of the internet service big data corresponding to the target interest service behavior, the noise error generated when the mining index is determined for each key internet data through the interest prediction index is reduced, and the precision of the estimation of the mining index set in the internet service big data is improved.
The following continues to describe a flow of a big data prediction-based internet service mining method according to an exemplary embodiment of the present disclosure, where the flow of the big data prediction-based internet service mining method may include the following steps.
Step S301, obtaining reference Internet service big data and reference interest business behaviors corresponding to each key Internet data in the reference Internet service big data.
In one embodiment, a reference data set is obtained; the reference data set comprises reference internet service big data and a key mining object corresponding to the reference internet service big data; the key mining object is used for indicating mining data characteristics of the key mining object in the reference internet service big data; acquiring a mining index set of each key internet data in the reference internet service big data based on the key mining object corresponding to the reference internet service big data; and acquiring the reference interest business behavior of each key internet data in the reference internet service big data based on the mining index set of each key internet data in the reference internet service big data.
In one embodiment, the key mining object may be generated based on interest service nodes of each interest feature information on each reference internet service big data, that is, the key mining object determines mining index positions and mining index information on the reference internet service big data according to the interest service node information of each interest feature information on each reference internet service big data.
In one embodiment, a reference mining index distribution graph corresponding to the reference internet service big data is obtained based on a key mining object corresponding to the reference internet service big data; the reference mining index distribution graph is used for indicating data service nodes where key mining objects in the reference internet service big data are located; based on the reference mining index distribution graph, performing data processing through a Gaussian convolution kernel to obtain mining confidence of a reference mining index corresponding to the reference internet service big data; and respectively determining mining indexes of each key internet data in the reference internet service big data based on the mining confidence of the reference mining indexes to obtain a mining index set of each key internet data of the reference internet service big data.
In an embodiment, when the reference internet service big data and the key mining object corresponding to the reference internet service big data are obtained, a position corresponding to the key mining object on the reference internet service big data may be labeled according to the key mining object corresponding to the reference internet service big data, so as to obtain a reference mining index distribution map corresponding to the reference internet service big data. For example, in the reference mining index distribution map, N interest feature information points x1 to xn in the map are considered. For each interest feature information point xi, a two-dimensional distribution map Hi may be generated, where the distribution value of only the interest feature information point positions is 1 and the rest positions are 0, then Hi corresponding to all interest feature information points are added to obtain a distribution map H (i.e., a reference mining index distribution map) of all interest feature information points corresponding to the reference internet service big data, and the mining index distribution set of the distribution map is a mining index set.
When internet service big data fragments are split for reference internet service big data, when any key internet data in the reference internet service big data contains an interest feature information point of some interest feature information, but all internet service big data parts of the interest feature information are not completely located in the key internet data, but the interest feature information is considered to be completely located in the key internet data because the interest feature information point of the interest feature information is located in the key internet data, so that mining index information of each key internet data in the reference internet service big data is represented inaccurately by generating a reference mining index distribution map corresponding to the reference internet service big data. At this time, the mining confidence of the reference mining index corresponding to the reference internet service big data can be obtained by performing convolution on a normalized gaussian convolution kernel distribution map, the mining confidence of the reference mining index is a gaussian distribution map formed based on each interest feature information point in the reference internet service big data, the distribution value of each point in the mining confidence of the reference mining index is used for indicating the mining index density of each point in the mining confidence of the reference mining index, so the mining confidence of the reference mining index can be used for indicating the mining index density of each standardized data area point on the reference internet service big data, and since the gaussian kernel is normalized, the mining confidence of the reference mining index obtained after data processing is performed through the gaussian convolution kernel, and the value obtained after determining the mining index is still the mining index set in the reference internet service big data, similarly, mining indexes of each key internet data in the reference internet service big data are determined, and a mining index set corresponding to each key internet data in the reference internet service big data can be obtained.
In one embodiment, an interest service behavior dimension corresponding to the interest prediction node is obtained; the interest business behavior dimension comprises at least two interest business sub-behavior dimensions; classifying through the interest business behavior dimension based on the mining index set of each key internet data in the reference internet service big data to obtain a reference interest business behavior corresponding to each key internet data in the reference internet service big data; the reference interest business behavior is used for indicating the interest business sub-behavior dimension corresponding to each key internet data in the reference internet service big data in the interest business behavior dimension.
For example, a reference mining index distribution graph corresponding to the reference internet service big data may be generated according to the reference internet service big data and the key mining object corresponding to the reference internet service big data, and the mining confidence P1 of the reference mining index corresponding to the reference internet service big data may be obtained by convolving the distribution graph by the normalized gaussian convolution kernel, and the distribution value size of each point in the mining confidence P1 of the reference mining index may indicate the mining index density of each point in the mining confidence of the reference mining index. Determining mining indexes based on each key internet data in the mining confidence of the reference mining indexes to obtain a mining index set PW corresponding to each key internet data of the reference internet service big data, and classifying mining index sets PW corresponding to each key Internet data of the reference Internet service big data through the interest service behavior dimension R in the interest prediction node to obtain interest service behavior dimensions corresponding to each reference key Internet service data in the reference Internet service big data respectively, wherein the interest business behavior dimension R comprises each interest business sub-behavior dimension of [ A0, A1], [ A1, A2], [ A2, A3], [ A3, A4], [ A4, A5], the reference interest business behavior corresponding to the interest business sub-behavior dimension [ A0, A1] in the interest business behavior dimension is S1; the reference interest business behavior corresponding to the interest business sub-behavior dimension [ A1, A2] in the interest business behavior dimension is S2; the reference interest business behavior corresponding to the interest business sub-behavior dimension [ A2, A3] in the interest business behavior dimension is S3; the reference interest business behavior corresponding to the interest business sub-behavior dimension [ A3, A4] in the interest business behavior dimension is S4; the reference interest business behavior corresponding to the interest business sub-behavior dimension [ a4, a5] in the interest business behavior dimension is S5. For example, for "a 1.a 2" of the mining index set PW corresponding to each key internet data in the reference internet service big data, the interest business behavior dimension R may be classified as an [ a0, a1] interest business sub-behavior dimension, and the corresponding interest business behavior is labeled as S1; for the 'A4. A2' of the mining index set PW corresponding to each key Internet data in the reference Internet service big data, the interest business behavior dimension R can be classified into an interest business sub-behavior dimension of [ A4, A5], and the corresponding interest business behavior is labeled as S4.
In one embodiment, a mining interest node sequence is obtained based on mining interest nodes of a mining index set of each key internet data of each reference internet service big data in the reference data set; the mining interest node sequence is used for indicating a mining interest node interval of the interest service behavior dimension; determining the splitting mining interest node interval based on the mining interest node interval of the interest service behavior dimension; the splitting mining interest node interval is used for indicating an interval distinguishing point of the interest business behavior dimension; the interval distinguishing point is used for dividing the interest business behavior dimension into each interest business sub-behavior dimension; and acquiring the interest service behavior dimension corresponding to the interest prediction node based on the mining interest node sequence and the split mining interest node interval.
The mining interest node interval of the interest business behavior dimension can be determined according to mining interest nodes of a mining index set of each key internet data in each reference internet service big data. The interest business behavior dimension corresponding to the classification layer is used for classifying the mining index set of the key internet data of each reference internet service big data in the reference data set, so that the interest business behavior dimension corresponding to the classification layer comprises mining interest nodes of the mining index set of each key internet data in each reference internet service big data in the reference data set.
In one embodiment, the edge mining index of the mining index set of each key internet data in each reference internet service big data is an edge mining index which is not an empty set in the mining index set of each key internet data in each reference internet service big data.
In one embodiment, the mining interest nodes of the mining index set of each key internet data in each piece of reference internet service big data are acquired as the mining interest node sequence.
The edge mining index of the mining index set of each key internet data in each reference internet service big data is obtained as a first interval point in the mining interest node sequence, the limit mining index of the mining index set of each key internet data in each reference internet service big data is obtained as a second interval point in the mining interest node sequence, and the left second interval point is a mining interest node interval of the interest business behavior dimension corresponding to the classification layer.
When the interest business behavior dimension is ensured to include the mining index set of each key internet data in all the reference internet service big data in the reference data set, the more the interest business behavior dimension is, the more accurate the classification is, so that the mining interest node of the mining index set of each key internet data in each reference internet service big data in the reference data set can be directly determined as the mining interest node interval of the interest business behavior dimension corresponding to the classification layer.
After the mining interest node sequence is determined, that is, after the mining interest node section of the interest business behavior dimension corresponding to the classification layer is determined, the section distinguishing point of the interest business behavior dimension corresponding to the classification layer can be determined according to the mining interest node section of the interest business behavior dimension corresponding to the classification layer.
In one embodiment, the classification number corresponding to the classification layer is obtained; and determining interval distinguishing points of the interest business behavior dimensionality corresponding to the classification layer based on the classification number corresponding to the classification layer.
The classification number corresponding to the classification layer is used for indicating the number of types which can be obtained by the classification layer after classifying the input reference internet service big data. For example, when the classification number corresponding to the classification layer is N (N is greater than or equal to 2, and N is a positive integer), that is, after data is classified by the classification layer, N types of confidence degrees can be obtained for the data, at this time, the interval discrimination point of the interest business behavior dimension corresponding to the classification layer may be N-1, and the interest business behavior dimension corresponding to the classification layer is segmented by N-1 interval discrimination points, so that N interest business sub-behavior dimensions corresponding to the classification layer can be obtained.
In one embodiment, based on the interest business behavior dimension corresponding to the classification layer, the interest business behavior dimension is split averagely through the classification number corresponding to the classification layer, and the mining interest node interval of the interest business behavior dimension corresponding to the classification layer is obtained.
In another embodiment, the mining interest node interval of the interest business behavior dimension corresponding to the classification layer may be e ^ K (log (b) -log (a))/K + log (a)) }, where, it is assumed that, except for an area where the mining index is 0, the smallest mining index set is a, the largest mining index set is b, and the number of the interest business behavior sub-dimensions to be divided is K. At this time, the interval sizes of the sub-behavior dimensions of each interest service are in nonlinear distribution, that is, the dimensions of the sub-behavior of the interest service for classifying smaller mining indexes are distributed more densely, and the dimensions of the sub-behavior of the interest service for classifying larger mining indexes are distributed more dispersedly, so that a better classification effect is achieved for mining index sets with different densities.
Step S302, based on the network extraction node of the interest business behavior prediction network, performing feature extraction on the reference Internet service big data to obtain reference operation behavior feature distribution corresponding to the reference Internet service big data.
The network extraction node in the interest business behavior prediction network is used for extracting the characteristics of the reference internet service big data in the reference data set to obtain the internet service big data characteristics corresponding to the reference internet service big data, wherein the internet service big data characteristics obtained by extracting the characteristics through the network extraction node are used for indicating mining index information in the reference internet service big data, so that the reference operation behavior characteristic distribution corresponding to the reference internet service big data can be used for indicating a mining index set and a mining index density corresponding to the reference internet service big data.
In one embodiment, the characteristic distribution range of the reference operation behavior characteristic distribution is the same as the characteristic distribution range of the reference internet service big data.
The reference operation behavior feature distribution obtained by feature extraction of the reference internet service big data through the interest business behavior prediction network is the same as the size of the standardized data area of the input reference internet service big data.
Step S303, based on the reference operation behavior feature distribution, performing interest prediction through an interest prediction node of the interest service behavior prediction network to obtain prediction results corresponding to each key Internet data in the reference Internet service big data.
In one embodiment, the prediction result is used to indicate a prediction confidence set corresponding to each key internet data in the reference internet service big data and the interest prediction node.
In one embodiment, based on the reference operation behavior feature distribution, performing interest prediction through an interest prediction node in the interest service behavior prediction network to obtain a prediction confidence set corresponding to each key internet data in the reference internet service big data and the interest prediction node; the prediction confidence set is used for indicating the confidence of each type corresponding to each interest prediction node of each key internet data in the reference internet service big data; and acquiring target interest business behaviors corresponding to the key internet data and the interest prediction nodes in the reference internet service big data based on the prediction confidence sets corresponding to the key internet data in the reference internet service big data.
For example, when the reference operation behavior feature distribution is processed by the interest prediction node, a prediction confidence set corresponding to each key internet data in the reference operation behavior feature distribution and the interest prediction node may be obtained, where the prediction confidence set is used to indicate a confidence that each key internet data of the reference internet service big data belongs to each type of the interest prediction node (that is, a confidence that each key internet data belongs to each interest service sub-behavior dimension of the interest service behavior dimension, respectively).
After the prediction confidence set is obtained, the determination of the maximum confidence in the prediction confidence set corresponding to each piece of key internet data can be determined as the interested business behavior of each piece of key internet data in the reference internet service big data.
Step S304, training the interest business behavior prediction network based on the prediction results respectively corresponding to each key Internet data in the reference Internet service big data and the reference interest business behaviors respectively corresponding to each reference key Internet service data in the reference Internet service big data.
In one embodiment, target interest business behaviors corresponding to all key internet data in the reference internet service big data are obtained; and training the interest business behavior prediction network according to the target interest business behaviors respectively corresponding to all key internet data in the reference internet service big data and the reference interest business behaviors respectively corresponding to all key internet data in the reference internet service big data.
In another embodiment, the response and the prediction result are used to indicate confidence distributions corresponding to each key internet data in the reference internet service big data and each interest business sub-behavior dimension of the interest business behavior dimension, and the interest business behavior prediction network is trained based on the reference interest business behavior corresponding to each reference key internet service data in the reference internet service big data and the confidence distributions.
The loss function value corresponding to the key internet data can be obtained by referring to the reference interest business behavior corresponding to each key internet data in the internet service big data and the confidence coefficient distribution corresponding to the key internet data, and the training can be performed according to the loss function value, or the loss functions corresponding to the key internet data can be obtained by obtaining the reference interest business behaviors corresponding to the key internet data in the reference internet service big data and the confidence coefficient distributions corresponding to the key internet data respectively, and the training can be performed according to the loss functions corresponding to the 35764Betre region.
Step S305, obtaining the Internet service big data of the Internet service terminal.
And step S306, based on the network extraction node of the interest business behavior prediction network, extracting the characteristics of the Internet service big data, and obtaining the operation behavior characteristic distribution corresponding to the Internet service big data.
In an embodiment, a network extraction node in the interest business behavior prediction network is configured to perform feature extraction on the internet service big data to obtain internet service big data features corresponding to the internet service big data, where the internet service big data features obtained by performing feature extraction through the network extraction node are used to indicate mining index information in the internet service big data, and therefore, operation behavior feature distribution corresponding to the internet service big data may be used to indicate a mining index set and a mining index density corresponding to the internet service big data.
In one embodiment, the characteristic distribution range of the operation behavior characteristic distribution is the same as the characteristic distribution range of the internet service big data.
The operation behavior feature distribution obtained by feature extraction of the internet service big data through the interest business behavior prediction network is the same as the size of the standardized data area of the input internet service big data.
Step S307, based on the operation behavior feature distribution, performing interest prediction through an interest prediction node of an interest business behavior prediction network to obtain target interest business behaviors corresponding to each key Internet data of the Internet service big data.
In one embodiment, based on the operation behavior feature distribution, interest prediction is performed through an interest prediction node of an interest service behavior prediction network to obtain a prediction result corresponding to each key internet data of the internet service big data, and based on the prediction result corresponding to each key internet data of the internet service big data, the prediction result corresponding to each key internet data of the internet service big data is obtained.
In one embodiment, based on the operation behavior feature distribution, performing interest prediction through an interest prediction node in the interest service behavior prediction network to obtain a prediction confidence set corresponding to each key internet data in the internet service big data and the interest prediction node; the prediction confidence set is used for indicating the confidence of each type corresponding to each interest prediction node of each key internet data in the internet service big data; and acquiring target interest business behaviors corresponding to the key internet data and the interest prediction nodes in the internet service big data based on the prediction confidence sets corresponding to the key internet data in the internet service big data.
For example, when the operation behavior feature distribution is processed by the interest prediction node, a prediction confidence set corresponding to each key internet data in the operation behavior feature distribution and the interest prediction node may be obtained, where the prediction confidence set is used to indicate a confidence that each key internet data of the internet service big data belongs to each type of the interest prediction node (that is, a confidence that each key internet data belongs to each interest business sub-behavior dimension of the interest business behavior dimension).
After the prediction confidence set is obtained, the determination that the confidence in the prediction confidence set corresponding to each piece of key internet data is the maximum can be determined as the interested business behavior of each piece of key internet data in the internet service big data.
Step S308, acquiring mining value indexes corresponding to all key Internet data of the Internet service big data based on target interest business behaviors corresponding to all key Internet data of the Internet service big data and interest estimation indexes corresponding to all target interest business behaviors.
The interest prediction index corresponding to each target interest business behavior is determined based on a mining index set of each key internet data of the reference internet service big data in the reference data set corresponding to the interest business behavior prediction network.
The interest prediction index corresponding to the target interest business behavior can be used for indicating a mining value index of key internet data corresponding to the target interest business behavior. That is, when the key internet data corresponds to the target interest business behavior, the interest prediction index corresponding to the target interest business behavior may be regarded as the mining value index corresponding to the key internet data.
In one embodiment, class reference key internet data is obtained; the type of reference key internet data is key internet data corresponding to the type of reference interest business behavior in the key internet data of each piece of reference internet service big data in the reference data set; the reference interest business behavior is any one of the reference interest business behaviors; and determining interest estimation indexes corresponding to the reference interest business behaviors based on mining index sets corresponding to all key internet data in the reference key internet data.
In one embodiment, mining index data values corresponding to the reference interest business behaviors are determined based on mining indexes in a preset sorting range of a mining index set corresponding to each key internet data in the reference key internet data.
In the embodiment of the present disclosure, the interest prediction index corresponding to each reference interest business behavior may be determined according to a mining index set corresponding to the key internet data corresponding to the reference interest business behavior in each key internet data of each reference internet service big data in the reference data set. The interest prediction indexes corresponding to the reference interest business behaviors may be determined by mining indexes in a preset sorting range of a mining index set corresponding to the key internet data corresponding to the reference interest business behaviors in the key internet data of the reference internet service big data in the reference data set, at this time, the sum of deep prediction errors between the real mining index of the key internet data corresponding to each reference interest business behavior and the interest prediction indexes is smaller, and at this time, when the real internet service big data is predicted through the interest prediction indexes, the generated deep prediction errors should be smaller.
Step S309, acquiring a mining index set corresponding to the internet service big data based on the mining value index corresponding to each key internet data in the internet service big data.
In one embodiment, mining index information corresponding to each piece of reference key internet service data in the internet service big data is collected to obtain a mining index set corresponding to the internet service big data.
In one embodiment, mining index information meeting specified conditions in mining index information corresponding to each piece of reference key internet service data in the internet service big data is collected to obtain a mining index set corresponding to the internet service big data.
The specified condition may be mining index information excluding a limit mining index and an edge mining index from all mining index information corresponding to each piece of reference key internet service data in the internet service big data.
In summary, the interest business behavior prediction network is trained by referring to the internet service big data and the reference interest business behavior of each key internet data in the reference internet service big data to obtain the trained interest business behavior prediction network, the target interest business behavior of each reference key internet service data of the input internet service big data and the interest prediction index corresponding to each target interest business behavior are determined by the interest business behavior prediction network, and the mining index set of the internet service big data is determined. By the scheme, the interest prediction index is determined by referring to the mining index set of each key internet data of the internet service big data in the reference data set corresponding to the interest service behavior prediction network, and the mining reference interval of the interest prediction index can be determined according to the mining index set corresponding to each key internet data, so that the interest prediction index corresponding to each target interest service behavior is more matched with the real interest tendency characteristics of the internet service big data corresponding to the target interest service behavior, the noise error generated when the mining index is determined for each key internet data through the interest prediction index is reduced, and the precision of the estimation of the mining index set in the internet service big data is improved.
In one embodiment, an embodiment of the present disclosure further provides an information push updating method based on big data prediction, including the following steps.
Step A201, based on a mining index set corresponding to the Internet service big data, acquiring an interactive session process data sequence containing a plurality of interactive session process data and a target online session process data sequence containing a plurality of target online session process data, which correspond to the mining index set, from the Internet service big data.
In the embodiment of the disclosure, the interactive session flow data sequence is a data sequence composed of interactive session flow data, the interactive session flow data sequence is a data sequence including a preset key recommendation object, the target online session flow data sequence is a data sequence composed of target online session flow data, and the target online session flow data sequence is a data sequence to be interacted, the preset key recommendation object is extracted from the interactive session flow data sequence, and the extracted preset key recommendation object is interacted into the target online session flow data sequence according to service requirements.
The interactive session flow data sequence and the target online session flow data sequence may have similar or identical subscription information of the interactive service, for example, both may be associated interactive session activity event characteristics or business object session activity event tags, and the like.
Step A202, collecting target frequent item conversation process data containing a preset key recommendation object from the interactive conversation process data containing the preset key recommendation object respectively according to the conversation activity event information corresponding to each interactive conversation process data in the interactive conversation process data sequence, and obtaining the target frequent item conversation process data sequence.
In the embodiment of the present disclosure, after the session activity event activation is performed on each piece of interacted session process data in the interacted session process data sequence, each piece of interacted session process data has corresponding session activity event information, the session activity event information is generally stored in the enabling log information of the session activity event, in the specific implementation, a plurality of pieces of interacted session process data in the interacted session process data sequence may correspond to the enabling log information of one session activity event, for example, the whole interacted session process data sequence may correspond to the enabling log information of one session activity event, that is, the session activity event information of all interacted session process data is stored in the enabling log information of one session activity event, when the session activity event information of a certain interacted session process data is needed, it may be queried from the log-enabled information for the session activity event. Or, it may be that one piece of interacted session flow data corresponds to the enabling log information of one session activity event, so that when the session activity event information of a certain piece of interacted session flow data is needed, the enabling log information of the session activity event of the interacted session flow data can be obtained from the query, and further the session activity event information is obtained. The session activity event information may include characteristics of session flow node data setting the session activity event and session log information in the interacted session flow data.
In the embodiment of the disclosure, whether the interactive session flow data includes the preset key recommendation object may be determined according to the session activity event information of the interactive session flow data, and then the interactive session flow data including the preset key recommendation object may be extracted when the preset key recommendation object is extracted. For each interactive session flow data containing a preset key recommendation object, the process of extracting the preset key recommendation object is similar, so that the process is described below by taking the interactive session flow data STE as an example, and the step of collecting specific content of the target frequent item session flow data from the interactive session flow data STE comprises the following steps.
Step a2021, obtains session activity event information of the interacted session flow data STE from the log information of the session activity event of the interacted session flow data STE.
Specifically, for the interacted session process data STE, all session activity event information of the interacted session process data STE is stored in the enabling log information of the session activity event, and then the corresponding session activity event information can be obtained according to the enabling log information of the session activity event of the interacted session process data STE.
Illustratively, the manner of collecting the target frequent conversation process data from the interacted conversation process data STE is as follows: the interacted session flow data STE is office session flow data, session activity event activation is performed on a service object session activity event label in the interacted session flow data STE, taking a preset key recommendation object as a service object session activity event label as an example, the session activity event information may specifically include session flow node data characteristic information and session log information in a session activity event marking frame, for example, "WR" represents a service object session activity event label corresponding to the session flow node data.
Step A2022, determining whether the interacted session flow data STE contains a preset key recommendation object according to the session activity event information.
The session activity event information indicates the session process node data characteristics of the interactive session process data STE activated by the session activity event, so that whether the interactive session process data STE contains the preset key recommendation object can be determined according to the session activity event information, and when the session activity event information indicates that the session process node data characteristics of the interactive session process data STE do not contain the preset key recommendation object, the process is ended.
Step a2023, if the determination result in step a2022 is yes, determining the target frequent item session flow node data feature in the interacted session flow data STE according to the session log information of the preset key recommendation object indicated by the session activity event information.
When the session activity event information indicates that the session process node data feature of the interacted session process data STE includes the preset key recommendation object, the target frequent item session process node data feature may be determined from the interacted session process data STE according to the session log information of the preset key recommendation object indicated by the session activity event information.
Step A2024, collecting target frequent item conversation process data corresponding to the key point recommendation object preset in the interactive conversation process data STE.
Specifically, a fragment splitting manner may be adopted, that is, after the data characteristics of the target frequent item session process node are determined, other frequent items in the interacted session process data STE are split, and only the data characteristics of the target frequent item session process node are retained, so that the target frequent item session process data is obtained. Or, because the session flow data is stored in the form of migratable session flow distribution information when stored in the computer device, the data of the target frequent item session flow node data feature can be read from the interacted session flow data STE according to the association relationship between the session log information and the migratable session flow information in the session flow data, and then the corresponding target frequent item session flow data can be generated according to the data.
Further, it can be determined that the interacted session flow data STE includes the service object session activity event tag according to the session activity event information of the interacted session flow data STE, so that the preset key recommendation object can be extracted from the interacted session flow data STE, and then the service behavior feature information of the service object session activity event tag in the interacted session flow data STE can be determined according to the session log information, and then the migratable session flow information can be read from the migratable session flow distribution information of the interacted session flow data STE, so as to generate the target frequent item session flow data including the service object session activity event tag.
After the above-mentioned method is adopted to operate on each interactive session flow data in the interactive session flow data sequence, the target frequent item session flow data corresponding to each interactive session flow data can be obtained, so that the target frequent item session flow data sequence is formed by summarizing the online session flow data. Specifically, reference may be made to the following description content of a target frequent item session flow data sequence, which takes a preset key recommendation object as a service object session activity event tag as an example, where the target frequent item session flow data sequence is composed of target frequent item session flow data of a plurality of preset key recommendation objects extracted from already-interacted session flow data, and the target frequent item session flow data sequence is composed of extracted session flow data of each service object session activity event tag.
Step A203, based on the obtained target frequent item conversation process data sequence, performing promotion intention matching on at least one target conversation process data in the target online conversation process data sequence aiming at a preset key recommendation object, and obtaining at least one target online conversation process data containing the preset key recommendation object.
Step A204, carrying out big data mining on at least one target online conversation process data containing the preset key recommendation object to obtain a recommendation effect portrait corresponding to the preset key recommendation object.
For example, the intention tendency feature information of at least one target online conversation process data containing the preset key recommendation object may be extracted, and the intention tendency feature information may be classified based on a pre-trained recommendation effect sketch classification model to obtain a recommendation effect sketch obtained corresponding to the classification.
And S205, updating an information pushing strategy of the preset key recommendation object based on the recommendation effect portrait corresponding to the preset key recommendation object.
For example, a preset update rule set of a recommendation effect portrait corresponding to the preset key recommendation object may be obtained, a target update rule set matched with the preset key recommendation object may be obtained from the preset update rule set, and an information push policy of the preset key recommendation object may be updated based on the target update rule set.
In the embodiment of the present disclosure, after the target frequent item session flow data set is obtained, the obtained target frequent item session flow data set may be used to perform promotion intention matching with the target online session flow data sequence, so as to obtain at least one optimized target online session flow data, that is, at least one target online session flow data including a preset key recommendation object. In practical application, the optimized at least one target online conversation process data and other target online conversation process data which are not matched with the promotion intention in the target frequent item conversation process data sequence are summarized to form a new online conversation process data sequence, and the online conversation process data sequence is an online conversation process data sequence obtained by optimizing the conversation process node data and can be used as a training data sequence in practical strategy configuration.
Based on the above, the promotion intention matching is performed by using the target frequent item conversation process data set and the target online conversation process data sequence, and the following contents may be specifically included.
Step A2031, at least one target online session flow data is determined from the target online session flow data sequence.
In the embodiment of the present disclosure, at least one target online conversation process data that needs to be matched with a promotion intention may be selected according to a specific situation of the target online conversation process data sequence.
Specifically, when the target online conversation process data sequence is a data sequence which does not include the preset key recommendation object at all, for example, all conversation process data in the entire target online conversation process data sequence may be matched with the promotion intention, or at least one target online conversation process data may be selected according to a certain proportion, for example, 60% or 75% of the target online conversation process data in the target online conversation process data sequence may be matched with the promotion intention.
Specifically, when the target online session flow data sequence locally includes the preset key recommendation object but the session flow node data session interaction evaluation parameter has a serious deviation, for example, all session flow data in the entire target online session flow data sequence may be matched with the promotion intention, or at least one target online session flow data that does not include the preset key recommendation object may be screened.
In specific implementation, in order to ensure the comprehensiveness of the session interaction evaluation parameters of various session process node data in the target online session process data sequence, the session interaction evaluation parameters of the pre-set key recommendation object may be determined according to the session interaction evaluation parameters of other feature session process node data in the target online session process data sequence, which have been activated by the session activity event. Specifically, the session interaction evaluation parameters of other feature session process node data activated by the session activity event in the target online session process data sequence can be obtained through the session activity event information, so that the session interaction evaluation parameters of the preset key recommendation object activated by the session activity event in the target online session process data sequence and the session interaction evaluation parameters of other feature session process node data can be determined according to the session activity event information of each target online session process data in the target online session process data sequence, and the session interaction evaluation parameters of the preset key recommendation object to be optimized in the target online session process data sequence can be determined according to the session interaction evaluation parameters of the preset key recommendation object and the session interaction evaluation parameters of other feature session process node data.
And then, determining the session interaction evaluation parameters of the target online session flow data to be selected according to the session interaction evaluation parameters of the preset key recommendation object to be optimized, namely determining at least one target online session flow data from the target online session flow data sequence, so that the fusion parameters of the session interaction evaluation parameters of the preset key recommendation object corresponding to each target online session flow data in the determined at least one target online session flow data are matched with the session interaction evaluation parameters of the preset key recommendation object to be optimized.
In specific implementation, the target online conversation process data with few conversation interaction evaluation parameters including the preset key recommendation object can be screened as at least one target online conversation process data. Specifically, the session interaction evaluation parameters of the preset key recommendation object corresponding to each target online session flow data sequence may be respectively determined according to the session activity event information of each target online session flow data in the target online session flow data sequence, and then, the target online session flow data in which the session interaction evaluation parameters of the preset key recommendation object are matched with the set session interaction evaluation parameters is determined as at least one target online session flow data.
Step A2032, for each target online conversation process data, screening at least one target frequent item conversation process data from the target frequent item conversation process data sequence.
After the at least one target online session flow data is selected, the at least one target frequent item session flow data may be filtered from the target frequent item session flow data sequence for each of the selected at least one target online session flow data.
Specifically, the target frequent item session flow data of the corresponding session interaction evaluation parameter may be screened for each target online session flow data according to the session interaction evaluation parameter of the preset key recommendation object to be optimized, and the fusion parameter of the session interaction evaluation parameter of the preset key recommendation object to be optimized of each target online session flow data in the at least one target online session flow data is matched with the session interaction evaluation parameter of the preset key recommendation object to be optimized. .
Or, the session interaction evaluation parameter of each target online session flow data may be preset, and then the target frequent item session flow data of the session interaction evaluation parameter may be screened and set for each target online session flow data.
It should be noted that, the screening of the target online session flow data that needs to be subjected to the matching of the promotion intention and the screening of the target frequent item session flow data corresponding to each target online session flow data may be performed simultaneously, that is, step a2031 and step a2032 may be performed simultaneously, for example, after determining the session interaction evaluation parameter of the preset key recommendation object that needs to be added, at least one session interaction evaluation parameter of the target online session flow data and the session interaction evaluation parameter of the target frequent item session flow data corresponding to each target online session flow data may be selected simultaneously.
Step A2033, carrying out promotion intention matching on the target online conversation process data and the corresponding at least one target frequent item conversation process data to obtain the target online conversation process data containing the preset key recommendation object.
Since the process of matching the promotion intention of each target online conversation process data and each target frequent item conversation process data is similar, the process of matching the promotion intention of a target frequent item conversation process data of a target online conversation process data is specifically described below, for example, the target frequent item conversation process data src3 is interacted with the target frequent item conversation process data src 2.
Specifically, when the target frequent item conversation process data src3 is interacted with the target frequent item conversation process data src2, firstly, the service behavior characteristics of the conversation process data of a preset key recommendation object in one target online conversation process data need to be determined.
Generally speaking, in the session flow data, each session flow node data corresponds to respective possibly activated service behavior feature information, for example, when determining the associated interactive session activity event feature, the session activity state of the associated session activity is generally constant, and the session activity state of the service object session activity event label is generally stable, so that a relatively stable interactive environment in the session flow data is generally activated, and therefore, the service behavior feature information in which the session activity state is set in the target online session flow data can be used as the service behavior feature information corresponding to the target frequent session flow data, which can perform service demand interaction. Because the target online conversation process data is stored in the mode of migratable conversation process distribution information, the business behavior characteristic information in the set conversation process data sequence in the migratable conversation process distribution information of the target online conversation process data can be determined as the business behavior characteristic of the conversation process data.
In some examples, when the pre-emphasis recommended object is a service object session activity event tag, the more stable service behavior feature information in the target online session flow data may be set as the service behavior feature of the session flow data, for example, a part of hot resource features in the target online session flow data may be used as the service behavior feature of the session flow data.
In specific implementation, a model for identifying (interacting) that the session flow node data of each feature may have an activated state may also be trained, so that service behavior feature information of which the initiating quantity is greater than a preset quantity in the target online session flow data of a preset (interacting) key recommendation object is identified through the model, and the service behavior feature information of which the initiating quantity is greater than the preset quantity is determined as the service behavior feature of the session flow data.
In the embodiment of the present disclosure, after determining the service behavior feature of the session flow data, the target frequent item session flow data may be interacted into the service behavior feature of the session flow data according to the service requirement, where the target frequent item session flow data may be interacted into the random state of the service behavior feature of the session flow data according to the service requirement, but the session flow node data frequent item that has been subjected to the session activity event marking in the target online session flow data needs to be considered.
In the embodiment of the disclosure, after the promotion intention matching is performed on at least one target online conversation process data, a preset key recommendation object may be adjusted in the target online conversation process data, so as to obtain at least one target online conversation process data after the promotion intention matching is performed.
Step A206, updating the starting log information of the session activity event of the target online session flow data according to the session flow node data characteristic information corresponding to each target frequent item session flow data in at least one target frequent item session flow data corresponding to the target online session flow data and the session log information of each target frequent item session flow data in one target online session flow data.
In the embodiment of the present disclosure, since each target online session flow data is to be used for subsequent policy configuration, session activity event information needs to be adjusted for each preset key recommendation object, where the session activity event information includes session flow node data feature information corresponding to each target frequent item session flow data in at least one target frequent item session flow data and session log information of each target frequent item session flow data in one target online session flow data, and thus, each corresponding session activity event information may be updated in the activation log information of the session activity event of each target online session flow data, so as to be used for subsequent policy configuration.
In some optional embodiments, the determining the key recommendation evaluation index of the internet service terminal according to the feedback behavior data for the session information pushing result specifically includes the following sub-step a:
sub-step a1, determining a sequence of push response valid objects through the feedback behavior data for the session information push result, where the sequence of push response valid objects includes x consecutive groups of push response valid objects, and x is an integer greater than 1;
a substep a2, obtaining a key response valid object sequence according to the push response valid object sequence, wherein the key response valid object sequence includes x consecutive groups of key response valid objects;
substep a3, obtaining an effective recommendation evaluation index sequence based on a first recommendation evaluation template included in a session information push strategy by using the push response effective object sequence, wherein the effective recommendation evaluation index sequence includes x effective recommendation evaluation indexes;
substep a4, obtaining a key recommendation evaluation index sequence based on a second recommendation evaluation template included in the session information pushing policy by using the key response valid object sequence, where the key recommendation evaluation index sequence includes x key recommendation evaluation indexes;
substep a5, obtaining, by using the effective recommendation evaluation index sequence and the key recommendation evaluation index sequence, key recommendation evaluation information corresponding to the push response effective object sequence based on a key recommendation evaluation template included in the session information push policy;
further, the obtaining of the key recommendation evaluation information corresponding to the push response effective object sequence based on the key recommendation evaluation template included in the session information push policy by using the effective recommendation evaluation index sequence and the key recommendation evaluation index sequence may specifically include performing the following embodiments.
Acquiring x first partition recommendation indexes by using the effective recommendation evaluation index sequence based on a first splitting template included in the session information pushing strategy, wherein each first partition recommendation index corresponds to one effective recommendation evaluation index; acquiring x second partition recommendation indexes by using the key recommendation evaluation index sequence based on a second splitting template included in the session information pushing strategy, wherein each second partition recommendation index corresponds to one key recommendation evaluation index; performing index mapping fusion on the x first partition recommendation indexes and the x second partition recommendation indexes to obtain x target fusion recommendation indexes, wherein each target fusion recommendation index comprises a first partition recommendation index and a second partition recommendation index; acquiring global partition recommendation indexes by using the x target fusion recommendation indexes based on a recommended application service scene unit included in the session information pushing strategy, wherein the global partition recommendation indexes are determined according to the x target fusion recommendation indexes and x application service scene IDs, and each target fusion recommendation index corresponds to one application service scene ID; and acquiring the key recommendation evaluation information corresponding to the push response effective object sequence based on a key recommendation evaluation template included in the session information push strategy by using the global partition recommendation index.
And a substep A6, determining a key recommendation evaluation index of the push response effective object sequence according to the key recommendation evaluation information.
By performing the above-described sub-step a 1-sub-step a6, the following advantages can be achieved:
in order to obtain a key response effective object sequence, firstly, a push response effective object sequence is determined according to feedback behavior data aiming at a session information push result, and then the key response effective object sequence is obtained in the push response effective object sequence, so that the accuracy of the obtained key response effective object sequence can be ensured, secondly, the push response effective object sequence is utilized, an effective recommendation evaluation index sequence and a key recommendation evaluation index sequence are respectively obtained based on a first recommendation evaluation template and a second recommendation evaluation template which are included in a session information push strategy, then, key recommendation evaluation information is obtained based on a key recommendation evaluation template which is included in the session information push strategy, and then, key recommendation evaluation indexes of the push response effective object sequence are determined according to the key recommendation evaluation information.
Therefore, on the basis of accurately obtaining the key response effective object sequence, the identification unit and the evaluation unit which are included in the session information pushing strategy are used for identifying and evaluating the pushed response effective object sequence one by one, so that the integrity of the obtained key recommendation evaluation information can be ensured, the key recommendation evaluation information can be accurately used as a reliable basis for determining the key recommendation evaluation index of the pushed response effective object sequence, and the availability of the obtained key recommendation evaluation index can be ensured.
In summary, through the data processing at the data level of the session flow nodes, a new data sequence is generated, session flow node data of existing session activity event data sequences may be interacted with by business requirements, a data sequence containing more session flow node data categories is obtained without the need to manually re-perform session activity event tagging, and the problem of deviation of data sequences in the interactive processing process can be solved by controlling the session interactive evaluation parameters of the interactive session process node data with various service requirements, and furthermore, limited data sequences can be utilized, data sequences containing more types of session process node data can be generated in a relatively efficient mode, and incomplete technical problems of different types of session process node data session interaction evaluation parameters in the existing data sequences can be solved through session process node data session interaction evaluation parameters.
By the design, for the partially lost conversation process node data or the biased conversation process node data in the target online conversation process data sequence, the targeted online session flow data can be specifically adjusted through the above-mentioned process, thereby optimizing the target online conversation process data sequence, enabling the process interaction state between each conversation process node data in the target online conversation process data sequence to be matched with the actual business interaction requirement, further, when strategy configuration is carried out by utilizing the target online conversation process data sequence, the configured strategy can comprehensively learn and obtain the characteristics of the interaction level of each conversation process node data without focusing on the identification of individual types of conversation process node data, therefore, the configured strategy can realize efficient interaction on the session process node data.
Fig. 3 is a schematic functional module diagram of an internet service mining apparatus 300 based on big data prediction according to an embodiment of the present disclosure, and the functions of the functional modules of the internet service mining apparatus 300 based on big data prediction are explained in detail below.
The first obtaining module 310 is configured to obtain internet service big data of an internet service terminal.
The first extraction module 320 is configured to extract features of the internet service big data based on a network extraction node of the interest business behavior prediction network, so as to obtain an operation behavior feature distribution corresponding to the internet service big data.
The second extraction module 330 is configured to predict interest of the operation behavior feature distribution based on an interest prediction node of the interest service behavior prediction network, so as to obtain a target interest service behavior corresponding to each key internet data in the internet service big data.
The second obtaining module 340 is configured to obtain, based on the target interest business behavior corresponding to each key internet data in the internet service big data and the interest prediction index corresponding to each target interest business behavior, a mining value index corresponding to each key internet data in the internet service big data of the internet service terminal.
The third obtaining module 350 is configured to obtain a mining index set corresponding to the internet service big data of the internet service terminal based on the mining value index corresponding to each key internet data in the internet service big data.
Fig. 4 illustrates a hardware structure diagram of a big data prediction system 100 for implementing the big data prediction-based internet service mining method, according to an embodiment of the present disclosure, and as shown in fig. 4, the big data prediction system 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a communication unit 140.
In a specific implementation process, at least one processor 110 executes computer-executable instructions stored in the machine-readable storage medium 120, so that the processor 110 may execute the internet service mining method based on big data prediction according to the above method embodiment, the processor 110, the machine-readable storage medium 120, and the communication unit 140 are connected through the bus 130, and the processor 110 may be configured to control transceiving actions of the communication unit 140, so as to perform data transceiving with the aforementioned internet service terminal 200.
For a specific implementation process of the processor 110, reference may be made to the above-mentioned various method embodiments executed by the big data prediction system 100, which implement principles and technical effects similar to each other, and this embodiment is not described herein again.
In addition, the embodiment of the disclosure also provides a readable storage medium, in which a computer execution instruction is preset, and when a processor executes the computer execution instruction, the internet service mining method based on big data prediction is implemented.
Finally, it should be understood that the examples in this specification are only intended to illustrate the principles of the examples in this specification. Other variations are also possible within the scope of this description. Accordingly, by way of example, and not limitation, alternative configurations of the embodiments of the specification can be seen as matching the teachings of the specification. Accordingly, the embodiments of the present description are not limited to only those embodiments explicitly described and depicted herein.
Claims (10)
1. An internet service mining method based on big data prediction is applied to a big data prediction system which is in communication connection with a plurality of internet service terminals, and the method comprises the following steps:
acquiring internet service big data of the internet service terminal;
performing feature extraction on the Internet service big data based on a network extraction node of an interest service behavior prediction network to obtain operation behavior feature distribution corresponding to the Internet service big data;
based on an interest prediction node of the interest business behavior prediction network, performing interest prediction on the operation behavior feature distribution to obtain target interest business behaviors corresponding to all key internet data in the internet service big data;
acquiring mining value indexes corresponding to all key internet data in the internet service big data based on target interest business behaviors corresponding to all key internet data in the internet service big data and interest prediction indexes corresponding to all the target interest business behaviors;
acquiring a mining index set corresponding to the internet service big data based on mining value indexes corresponding to all key internet data in the internet service big data;
the interest business behavior prediction network is an artificial intelligence learning network obtained by training by taking reference internet service big data in a reference data set as training data and taking reference interest business behaviors of all key internet data of the reference internet service big data as training attributes; and the interest pre-estimation indexes corresponding to the target interest business behaviors are determined based on the mining index set of each key internet data of the reference internet service big data.
2. The big data prediction-based internet service mining method according to claim 1, wherein the method further comprises:
acquiring reference internet service big data and reference interest business behaviors corresponding to all key internet data in the reference internet service big data respectively;
performing feature extraction on the reference internet service big data based on a network extraction node of an interest service behavior prediction network to obtain reference operation behavior feature distribution corresponding to the reference internet service big data;
based on the reference operation behavior feature distribution, performing interest prediction through an interest prediction node of an interest service behavior prediction network to obtain prediction results corresponding to each key internet data in the reference internet service big data;
and training the interest business behavior prediction network based on the prediction results respectively corresponding to each key internet data in the reference internet service big data and the reference interest business behaviors respectively corresponding to each reference key internet service data in the reference internet service big data.
3. The internet service mining method based on big data prediction as claimed in claim 2, wherein the obtaining of the reference internet service big data and the reference interest business behavior corresponding to each key internet data in the reference internet service big data respectively comprises:
acquiring the reference data set; the reference data set comprises reference internet service big data and a key mining object corresponding to the reference internet service big data; the key mining object is used for indicating mining data characteristics of the key mining object in the reference internet service big data;
acquiring a mining index set of each key internet data in the reference internet service big data based on the key mining object corresponding to the reference internet service big data;
and acquiring the reference interest business behavior of each key internet data in the reference internet service big data based on the mining index set of each key internet data in the reference internet service big data.
4. The internet service mining method based on big data prediction as claimed in claim 3, wherein the obtaining of the reference interest business behavior of each key internet data in the reference internet service big data based on the mining index set of each key internet data in the reference internet service big data comprises:
obtaining an interest service behavior dimension corresponding to the interest prediction node; the interest business behavior dimension comprises at least two interest business sub-behavior dimensions;
classifying through the interest business behavior dimension based on the mining index set of each key internet data in the reference internet service big data to obtain a reference interest business behavior corresponding to each key internet data in the reference internet service big data;
the obtaining of the interest service behavior dimension corresponding to the interest prediction node includes:
acquiring a mining interest node sequence based on mining interest nodes of a mining index set of each key internet data of each reference internet service big data in the reference data set; the mining interest node sequence is used for indicating a mining interest node interval of the interest service behavior dimension;
determining a split mining interest node interval based on the mining interest node interval of the interest service behavior dimension; the splitting mining interest node interval is used for indicating an interval distinguishing point of the interest business behavior dimension; the interval distinguishing point is used for dividing the interest business behavior dimension into each interest business sub-behavior dimension;
and acquiring the interest service behavior dimension corresponding to the interest prediction node based on the mining interest node sequence and the split mining interest node interval.
5. The big data prediction-based internet service mining method according to claim 4, wherein the method further comprises:
acquiring class reference key internet data; the class reference key internet data is key internet data corresponding to the class reference interest business behavior in the key internet data of each reference internet service big data in the reference data set; the class reference interest business behavior is any one of the reference interest business behaviors;
and determining interest estimation indexes corresponding to the class reference interest business behaviors based on preset sorting range mining indexes of a mining index set corresponding to each key internet data in the class reference key internet data.
6. The internet service mining method based on big data prediction according to claim 3, wherein the obtaining of the mining index set of each key internet data in the reference internet service big data based on the key mining object corresponding to the reference internet service big data comprises:
obtaining a reference mining index distribution map corresponding to the reference internet service big data based on the reference internet service big data and a key mining object corresponding to the reference internet service big data; the reference mining index distribution graph is used for indicating data service nodes where mining indexes in the reference internet service big data are located;
based on the reference mining index distribution graph, performing data processing through a Gaussian convolution kernel to obtain mining confidence of a reference mining index corresponding to the reference internet service big data;
and respectively determining mining indexes of all key internet data in the reference internet service big data based on the mining confidence degrees of the reference mining indexes to obtain a mining index set of all key internet data of the reference internet service big data.
7. The big data prediction-based internet service mining method according to any one of claims 1 to 6, wherein the method further comprises:
acquiring an interactive session process data sequence which comprises a plurality of interactive session process data and a target online session process data sequence which comprises a plurality of target online session process data, wherein the interactive session process data sequence corresponds to the mining index set; the interactive session process data and the target online session process data have the same interactive service subscription information, each interactive session process data and each target online session process data are summarized by adopting session process node data in multiple session modes, and at least one session process node data included in each interactive session process data and each target online session process data is configured with corresponding session activity event information;
respectively acquiring target frequent item conversation process data containing a preset key recommendation object from each interactive conversation process data according to the conversation activity event information corresponding to each interactive conversation process data in the interactive conversation process data sequence, and acquiring a target frequent item conversation process data sequence;
based on the obtained target frequent item conversation process data sequence, carrying out promotion intention matching on at least one target online conversation process data in the target online conversation process data sequence aiming at the preset key recommendation object, and obtaining at least one target online conversation process data containing the preset key recommendation object;
carrying out big data mining on at least one target online conversation process data containing the preset key recommendation object to obtain a recommendation effect portrait corresponding to the preset key recommendation object;
and updating the information pushing strategy of the preset key recommendation object based on the recommendation effect portrait corresponding to the preset key recommendation object.
8. The internet service mining method based on big data prediction according to claim 7, wherein the step of obtaining target frequent item conversation process data including a preset key recommendation object from each interactive conversation process data according to the conversation activity event information corresponding to each interactive conversation process data in the interactive conversation process data sequence to obtain the target frequent item conversation process data sequence comprises:
acquiring corresponding session activity event information aiming at the starting log information of the session activity event of each interactive session flow data;
for each obtained session activity event information, the following operations are respectively executed:
and when the session process node data characteristics indicated in the obtained session activity event information are determined to contain the preset key recommendation object, acquiring target frequent item session process data corresponding to the preset key recommendation object from corresponding interactive session process data based on a frequent model item algorithm according to the session log information of the preset key recommendation object indicated by the session activity event information.
9. The internet service mining method based on big data prediction according to claim 7, wherein the obtaining at least one target online conversation process data including the preset key recommendation object by performing promotion intention matching on at least one target online conversation process data in the target online conversation process data sequence for the preset key recommendation object based on the obtained target frequent item conversation process data sequence comprises:
determining session interaction evaluation parameters of the preset key recommendation object in a session interaction state in the target online session flow data sequence and session interaction evaluation parameters of other characteristic session flow node data according to session activity event information of each target online session flow data in the target online session flow data sequence;
determining the session interaction evaluation parameters of the preset key recommendation object to be optimized in the target online session flow data sequence according to the session interaction evaluation parameters of the preset key recommendation object and the session interaction evaluation parameters of the other characteristic session flow node data;
determining the at least one target online conversation process data from the target online conversation process data sequence according to the conversation interaction evaluation parameter of the preset key recommendation object to be optimized; the method comprises the steps that a fusion parameter of conversation interaction evaluation parameters of a preset key recommendation object corresponding to each target online conversation process data in at least one target online conversation process data is matched with the conversation interaction evaluation parameters of the preset key recommendation object to be optimized;
for each target online conversation process data in the at least one target online conversation process data, respectively executing the following operations:
screening target frequent item conversation process data of the corresponding conversation interaction evaluation parameter for the target online conversation process data according to the conversation interaction evaluation parameter of the preset key recommendation object to be optimized; the method comprises the steps that a fusion parameter of session interaction evaluation parameters of a preset key recommendation object to be optimized of each target online session flow data in at least one target online session flow data is matched with the session interaction evaluation parameters of the preset key recommendation object to be optimized; or, screening and setting target frequent item conversation process data of conversation interaction evaluation parameters for the target online conversation process data, and performing promotion intention matching on the target online conversation process data and the at least one target frequent item conversation process data to obtain one target online conversation process data containing the preset key recommendation object.
10. A big data prediction system, comprising a processor, a machine-readable storage medium, and a communication unit, wherein the machine-readable storage medium, the communication unit, and the processor are associated through a bus system, the communication unit is configured to be communicatively connected to at least one internet service terminal, the machine-readable storage medium is configured to store computer instructions, and the processor is configured to execute the computer instructions in the machine-readable storage medium to perform the big data prediction based internet service mining method of any one of claims 1 to 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110682187.8A CN113449012A (en) | 2021-06-20 | 2021-06-20 | Internet service mining method based on big data prediction and big data prediction system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110682187.8A CN113449012A (en) | 2021-06-20 | 2021-06-20 | Internet service mining method based on big data prediction and big data prediction system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113449012A true CN113449012A (en) | 2021-09-28 |
Family
ID=77811974
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110682187.8A Withdrawn CN113449012A (en) | 2021-06-20 | 2021-06-20 | Internet service mining method based on big data prediction and big data prediction system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113449012A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115329205A (en) * | 2022-09-15 | 2022-11-11 | 郭振东 | Big data mining method serving personalized push service and AI recommendation system |
CN115687792A (en) * | 2022-12-20 | 2023-02-03 | 邢台达喆网络科技有限公司 | Big data acquisition method and system for online internet service |
CN115828011A (en) * | 2022-10-13 | 2023-03-21 | 徐州海清信息科技有限公司 | Data analysis method and platform based on big data |
CN116956030A (en) * | 2023-07-21 | 2023-10-27 | 广州一号家政科技有限公司 | Household business processing method and system based on artificial intelligence |
-
2021
- 2021-06-20 CN CN202110682187.8A patent/CN113449012A/en not_active Withdrawn
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115329205A (en) * | 2022-09-15 | 2022-11-11 | 郭振东 | Big data mining method serving personalized push service and AI recommendation system |
CN115828011A (en) * | 2022-10-13 | 2023-03-21 | 徐州海清信息科技有限公司 | Data analysis method and platform based on big data |
CN115828011B (en) * | 2022-10-13 | 2023-11-10 | 四川宏智科信数字科技有限公司 | Data analysis method and platform based on big data |
CN115687792A (en) * | 2022-12-20 | 2023-02-03 | 邢台达喆网络科技有限公司 | Big data acquisition method and system for online internet service |
CN116956030A (en) * | 2023-07-21 | 2023-10-27 | 广州一号家政科技有限公司 | Household business processing method and system based on artificial intelligence |
CN116956030B (en) * | 2023-07-21 | 2024-02-02 | 广州一号家政科技有限公司 | Household business processing method and system based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11487941B2 (en) | Techniques for determining categorized text | |
CN111859960B (en) | Semantic matching method, device, computer equipment and medium based on knowledge distillation | |
CN111966914B (en) | Content recommendation method and device based on artificial intelligence and computer equipment | |
CN113449012A (en) | Internet service mining method based on big data prediction and big data prediction system | |
CN113449011A (en) | Big data prediction-based information push updating method and big data prediction system | |
Bastani et al. | Online nonparametric bayesian activity mining and analysis from surveillance video | |
CN110852881B (en) | Risk account identification method and device, electronic equipment and medium | |
JP2020109631A (en) | Agile video query using ensembles of deep neural networks | |
CN113535825A (en) | Cloud computing intelligence-based data information wind control processing method and system | |
CN109685104B (en) | Determination method and device for recognition model | |
CN107203772B (en) | User type identification method and device | |
CN114492601A (en) | Resource classification model training method and device, electronic equipment and storage medium | |
CN115130711A (en) | Data processing method and device, computer and readable storage medium | |
Lee et al. | Effective evolutionary multilabel feature selection under a budget constraint | |
Hassanat et al. | Magnetic force classifier: a Novel Method for Big Data classification | |
WO2022148108A1 (en) | Systems, devices and methods for distributed hierarchical video analysis | |
Gao et al. | Reinforcement learning based web crawler detection for diversity and dynamics | |
Liu et al. | [Retracted] Deep Learning and Collaborative Filtering‐Based Methods for Students’ Performance Prediction and Course Recommendation | |
Kumagai et al. | Learning non-linear dynamics of decision boundaries for maintaining classification performance | |
CN113159213A (en) | Service distribution method, device and equipment | |
Oneto et al. | Advances in artificial neural networks, machine learning and computational intelligence | |
Nasfi et al. | A novel feature selection method using generalized inverted Dirichlet-based HMMs for image categorization | |
CN116305103A (en) | Neural network model backdoor detection method based on confidence coefficient difference | |
CN116541592A (en) | Vector generation method, information recommendation method, device, equipment and medium | |
CN114818900A (en) | Semi-supervised feature extraction method and user credit risk assessment method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210928 |