A method and apparatus for analyzing and predicting user network behavior

Abstract

The application discloses a user network behavior analysis and prediction method, comprising the following steps: obtaining a user network behavior information set; extracting a consumption behavior set and a pre-consumption behavior set from the user network behavior information set; performing correlation analysis on the consumption behavior set and the pre-consumption behavior set to obtain a user payment prediction model; and processing the collected user network behavior information by using the user payment prediction model to obtain network content prediction information and a prediction value of user payment possibility. The application performs deep extraction and mining on user network behavior data, accurately establishes a quantitative relationship between user behavior and payment intention, and thus optimizes network content such as a video matrix and improves network traffic conversion rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of big data and artificial intelligence, and in particular to a user network behavior analysis and prediction method and device. BACKGROUND

[0002] Currently, with the continuous improvement of commercial informatization level, the knowledge payment mode of video stream marketing is gradually becoming the mainstream business model. For the MCN platform, it needs to screen out high-value users who really have payment intention and knowledge demand from a large number of users. At the same time, in the video matrix marketing, the network behavior of users is an important basis for analyzing and optimizing marketing strategies. At the same time, with the diversification of Internet platforms, user behavior data also presents the characteristics of diversification and time-varying. It is difficult to accurately predict the consumption characteristics and consumption intention of users with a single mode of data analysis model.

[0003] How to deeply extract and mine the diversity and complexity of user network behavior data, accurately establish the quantitative relationship between user behavior and payment intention, and thus optimize the video matrix and other network content and improve the network traffic conversion rate is a problem that needs to be solved at present. SUMMARY

[0004] In order to solve the problem of how to deeply extract and mine user network behavior, establish the quantitative relationship between user behavior and payment intention, and thus optimize the video matrix content and improve the network traffic conversion rate, the present application discloses a user network behavior analysis and prediction method.

[0005] In a first aspect, the present application discloses a user network behavior analysis and prediction method, comprising:

[0006] S1, obtaining a user network behavior information set;

[0007] S2, extracting a consumption behavior set and a pre-consumption behavior set from the user network behavior information set;

[0008] S3, performing correlation analysis on the consumption behavior set and the pre-consumption behavior set to obtain a user payment prediction model;

[0009] S4, using the user payment prediction model to process the collected user network behavior information to obtain network content prediction information and a prediction value of user payment possibility.

[0010] The consumption behavior set and the pre-consumption behavior set extracted from the user network behavior information set comprise:

[0011] From the user network behavior information set, a consumption behavior set is extracted; the consumption behavior set includes a plurality of consumption behaviors and corresponding occurrence times; the consumption behaviors include purchase behaviors, subscription behaviors, and reward behaviors;

[0012] From the user network behavior information set, all non-consumption behaviors occurring before the occurrence time of the consumption behavior are extracted according to the occurrence time of the consumption behavior, and a pre-consumption behavior set is constructed using the extracted all non-consumption behaviors; the pre-consumption behavior set includes pre-consumption behaviors and corresponding occurrence times.

[0013] The correlation analysis on the consumption behavior set and the pre-consumption behavior set obtains a user payment prediction model, including:

[0014] S31, the consumption behavior set is classified to obtain a consumption target subset, a consumption attribute subset, and a network operation subset;

[0015] S32, the network operation subset is subjected to path association processing to obtain a behavior sequence subset;

[0016] S33, the consumption behavior set, the consumption target subset, the consumption attribute subset, and the network operation subset are subjected to consumption behavior target association processing to obtain an initial content prediction model; the output of the initial content prediction model is initial network content prediction information;

[0017] S34, the consumption behavior set, the consumption target subset, the consumption attribute subset, and the behavior sequence subset are subjected to behavior sequence target association processing to obtain a marketing content sequence prediction model; the output of the marketing content sequence prediction model is network content prediction sequence information;

[0018] S35, the consumption behavior set, the consumption target subset, and the consumption attribute subset are subjected to consumption prediction processing to obtain a user consumption prediction model; the output of the user consumption prediction model is a prediction value of user payment possibility;

[0019] S36, the initial content prediction model, the marketing content sequence prediction model, and the user consumption prediction model are subjected to fusion processing to obtain a user payment prediction model.

[0020] The consumption target subset includes keywords and occurrence times in search behaviors and click behaviors of the user;

[0021] The consumption attribute subset includes device usage information, space activity information, user attribute information, traffic source information, and sentiment analysis information;

[0022] The network operation sub-set includes viewing behavior information, interaction behavior information, subscription and attention information, social interaction information, exit behavior information, sharing channel analysis information, and advertisement interaction information.

[0023] Each information included in the network operation sub-set includes an operation type, an operation object, and an occurrence time; the operation object includes video information, webpage information, tag information, comment information, and advertisement information to which the network operation is directed. All operation objects constitute an operation object set.

[0024] The consumption behavior target correlation processing on the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set, and the network operation sub-set obtains an initial content prediction model, including:

[0025] S331, taking the operation object set in the consumption behavior set and the network operation sub-set as a target variable, taking the consumption target sub-set and the consumption attribute sub-set as an input variable, performing correlation analysis on the target variable and the input variable to obtain a corresponding relationship model of the input variable and the target variable;

[0026] S332, determining the corresponding relationship model as an initial content prediction model.

[0027] The consumption behavior target correlation processing on the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set, and the network operation sub-set obtains an initial content prediction model, including:

[0028] S3311, extracting each consumption behavior information in the consumption behavior set and the corresponding operation object in the network operation sub-set; using all extracted operation objects to construct an operation object information set;

[0029] S3312, establishing a corresponding relationship between each consumption target information and consumption attribute information and the operation object in the operation object information set according to the corresponding consumption target information and consumption attribute information of each consumption behavior information in the consumption behavior set;

[0030] S3313, respectively performing digital quantization coding processing on the consumption target sub-set, the consumption attribute sub-set, and the operation object information set to obtain a consumption target data set, a consumption attribute data set, and an operation object data set;

[0031] S3314, extracting corresponding consumption target data in the consumption target data set and corresponding consumption attribute data in the consumption attribute data set from each operation object data in the operation object data set;

[0032] S3315, for each operation object data corresponding to all the consumption target data and consumption attribute data, the corresponding two-dimensional data set is constructed; the two-dimensional data in the two-dimensional data set is the consumption target data and the consumption attribute data;

[0033] S3316, for each operation object data corresponding to the two-dimensional data set, data dimension reduction processing is performed to obtain a corresponding reduced two-dimensional data set; the reduced two-dimensional data set includes consumption target data and consumption attribute data;

[0034] S3317, based on all operation object data corresponding to the reduced two-dimensional data set, the corresponding relationship between each reduced two-dimensional data and operation object data is established;

[0035] S3318, according to the corresponding relationship between the reduced two-dimensional data and the operation object data, the corresponding relationship between the operation object and the consumption target and the consumption attribute is established;

[0036] S3319, the corresponding relationship between the operation object and the consumption target and the consumption attribute is determined as the corresponding relationship model of the input variable and the target variable.

[0037] The behavior sequence target association processing on the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set, the behavior sequence sub-set, the marketing content sequence prediction model is obtained, including:

[0038] S341, the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set are clustered and analyzed to obtain a consumption category model and consumption category information; the consumption category model is used for processing the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set to obtain the consumption category information;

[0039] S342, the behavior sequence sub-set corresponding to each consumption category information is modeled to obtain a behavior sequence transition model;

[0040] S343, the consumption category model and the behavior sequence transition model are used to construct a marketing content sequence prediction model.

[0041] The second aspect of the embodiment of the application discloses a user network behavior analysis and prediction device, the device comprises:

[0042] The memory stores executable program code;

[0043] The processor is coupled with the memory;

[0044] The processor calls the executable program code stored in the memory to execute the user network behavior analysis and prediction method.

[0045] In a third aspect, the application discloses a computer storage medium, which stores computer instructions, and the computer instructions are used to execute the user network behavior analysis and prediction method.

[0046] In a fourth aspect, the application discloses an information data processing terminal, which is used to implement the user network behavior analysis and prediction method.

[0047] The application has the following advantages:

[0048] 1. The application classifies and processes network behaviors related to user payment behaviors, maximally mines various user network behaviors, firstly performs consumption prediction processing on the consumption behavior set, the consumption target subset and the consumption attribute subset, obtains a user consumption prediction model, realizes prediction of user payment possibility, improves the accuracy of content delivery, establishes an initial content prediction model, realizes prediction of network behaviors attracting user attention, generates network content extremely attractive to users by using the model, establishes a marketing content sequence prediction model on the basis, generates a content sequence continuously attractive, finally fuses the initial content prediction model, the marketing content sequence prediction model and the user consumption prediction model, and obtains a user payment prediction model, realizing prediction of user final payment behavior.

[0049] 2. The application is used in a network marketing platform, and the advantages include:

[0050] Optimizing content strategy: adjusting video content and publishing strategy according to user preferences and interaction;

[0051] Precise marketing: using user behavior data for personalized recommendation and targeted advertising delivery;

[0052] Improving conversion rate: identifying key factors affecting user purchase decisions and optimizing marketing funnel;

[0053] Enhancing user stickiness: improving user participation and loyalty through interaction and community building. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The flowchart of the method of the application;

[0055] Figure 2 The composition diagram of the user network behavior analysis and prediction device of the application;

[0056] Figure 3A logical relationship diagram of each prediction model of the present application. DETAILED DESCRIPTION

[0057] For better understanding of the present application, an embodiment is given here.

[0058] Figure 1 An implementation flowchart of the method of the present application; Figure 2 A composition diagram of the user network behavior analysis and prediction device of the present application; Figure 3 A logical relationship diagram of each prediction model of the present application.

[0059] In a first aspect, the embodiment of the present application discloses a user network behavior analysis and prediction method, comprising:

[0060] S1, obtaining a user network behavior information set;

[0061] S2, extracting a consumption behavior set and a pre-consumption behavior set from the user network behavior information set;

[0062] S3, performing correlation analysis on the consumption behavior set and the pre-consumption behavior set to obtain a user payment prediction model;

[0063] S4, processing the collected user network behavior information by using the user payment prediction model to obtain network content prediction information and a prediction value of user payment possibility;

[0064] The extraction of the consumption behavior set and the pre-consumption behavior set from the user network behavior information set comprises:

[0065] From the user network behavior information set, a consumption behavior set is extracted; the consumption behavior set comprises a plurality of consumption behaviors and corresponding occurrence times; the consumption behaviors comprise purchase behaviors, subscription behaviors and reward behaviors;

[0066] From the user network behavior information set, all non-consumption behaviors occurring before the occurrence time of the consumption behaviors are extracted according to the occurrence time of the consumption behaviors, and a pre-consumption behavior set is constructed by using all the extracted non-consumption behaviors; the pre-consumption behavior set comprises pre-consumption behaviors and corresponding occurrence times;

[0067] The correlation analysis on the consumption behavior set and the pre-consumption behavior set to obtain a user payment prediction model comprises:

[0068] S31, performing classification processing on the consumption behavior set to obtain a consumption target sub-set, a consumption attribute sub-set and a network operation sub-set;

[0069] S32, performing path association processing on the network operation sub-set to obtain a behavior sequence sub-set;

[0070] S33, performing consumption behavior target association processing on the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set, and the network operation sub-set to obtain an initial content prediction model; the output of the initial content prediction model is initial network content prediction information;

[0071] S34, performing behavior sequence target association processing on the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set, and the behavior sequence sub-set to obtain a marketing content sequence prediction model; the output of the marketing content sequence prediction model is network content prediction sequence information;

[0072] S35, performing consumption prediction processing on the consumption behavior set, the consumption target sub-set, and the consumption attribute sub-set to obtain a user consumption prediction model; the output of the user consumption prediction model is a predicted value of user payment possibility;

[0073] S36, performing fusion processing on the initial content prediction model, the marketing content sequence prediction model, and the user consumption prediction model to obtain a user payment prediction model;

[0074] The path association processing on the network operation sub-set to obtain a behavior sequence sub-set is based on each network operation information in the network operation sub-set, and each network operation contained in the network operation information is sorted according to the occurrence time of the network operation to obtain a behavior sequence corresponding to the network operation information;

[0075] The behavior sequence sub-set is constructed by using the behavior sequences corresponding to all network operation information;

[0076] The consumption target sub-set includes but is not limited to keywords and occurrence times in search behavior and click behavior of users;

[0077] The consumption attribute sub-set includes but is not limited to device usage information, space activity information, user attribute information, traffic source information, and sentiment analysis information;

[0078] The network operation sub-set includes but is not limited to viewing behavior information, interactive behavior information, subscription and attention information, social interaction information, exit behavior information, sharing channel analysis information, and advertisement interaction information;

[0079] Each information contained in the network operation sub-set includes but is not limited to operation type, operation object, and occurrence time; the operation object includes video information, webpage information, tag information, comment information, and advertisement information to which the network operation is directed.

[0080] keywords in the search behavior and click behavior of the user, including keywords in the search input by the user and keywords clicked by the user;

[0081] The consumption behavior target association processing on the consumption behavior set, the consumption target subset, the consumption attribute subset and the network operation subset obtains an initial content prediction model, and the method comprises the following steps:

[0082] S331, taking the operation object set in the consumption behavior set and the network operation subset as a target variable, taking the consumption target subset and the consumption attribute subset as input variables, performing correlation analysis on the target variable and the input variable to obtain a corresponding relationship model of the input variable and the target variable;

[0083] S332, determining the corresponding relationship model as an initial content prediction model;

[0084] The consumption behavior target association processing on the consumption behavior set, the consumption target subset, the consumption attribute subset and the network operation subset obtains an initial content prediction model, and the method comprises the following steps:

[0085] S3311, extracting each consumption behavior information in the consumption behavior set to obtain a corresponding operation object in the network operation subset; using all the extracted operation objects, an operation object information set is constructed;

[0086] S3312, according to the consumption target information and the consumption attribute information corresponding to each consumption behavior information in the consumption behavior set, a corresponding relationship between each consumption target information and consumption attribute information and the operation object in the operation object information set is established;

[0087] S3313, respectively performing digital quantization coding processing on the consumption target subset, the consumption attribute subset and the operation object information set to obtain a consumption target data set, a consumption attribute data set and an operation object data set;

[0088] S3314, for each operation object data in the operation object data set, corresponding consumption target data in the consumption target data set and corresponding consumption attribute data in the consumption attribute data set are extracted;

[0089] S3315, for all the corresponding consumption target data and consumption attribute data of each operation object data, a corresponding two-dimensional data set is constructed; the two-dimensional data in the two-dimensional data set is the consumption target data and the consumption attribute data;

[0090] S3316, performing data dimension reduction processing on the corresponding two-dimensional data set of each operation object data to obtain a corresponding reduced two-dimensional data set; the reduced two-dimensional data set includes consumption target data and consumption attribute data;

[0091] S3317, establishing a corresponding relationship between each reduced two-dimensional data and operation object data based on the corresponding reduced two-dimensional data set of all operation object data;

[0092] S3318, establishing a corresponding relationship between the operation object and the consumption target and the consumption attribute according to the corresponding relationship between the reduced two-dimensional data and the operation object data;

[0093] S3319, determining the corresponding relationship between the operation object and the consumption target and the consumption attribute as the corresponding relationship model of the input variable and the target variable.

[0094] According to the corresponding relationship model of the input variable and the target variable, after obtaining the consumption target information and the consumption attribute information of the user, the corresponding operation object can be obtained, which is the operation object that can most make the user generate the consumption behavior. The operation object corresponds to the content displayed to the user on the network. According to the operation object, the corresponding content is displayed to the user, which can greatly improve the probability of the occurrence of the consumption behavior of the user.

[0095] According to the corresponding relationship between the reduced two-dimensional data and the operation object data, the corresponding relationship between the operation object and the consumption target and the consumption attribute is established according to digital quantization coding processing, each data is corresponded to the corresponding operation object, consumption target or consumption attribute, and then the data is replaced with text according to the corresponding relationship between the reduced two-dimensional data and the operation object data to obtain the corresponding relationship between the operation object and the consumption target and the consumption attribute.

[0096] The data dimension reduction algorithm used in the data dimension reduction processing of the corresponding two-dimensional data set of each operation object data can be a principal component analysis algorithm or a clustering algorithm. The data dimension reduction processing is to delete the data having a correlation relationship in the two-dimensional data set from the data set, thereby reducing the number of data.

[0097] The operation object is a link, a picture or a video watched by the user, etc.

[0098] The establishment of the corresponding relationship between each consumption target information and consumption attribute information and the operation object in the operation object information set is to determine that the consumption target information and the consumption attribute information corresponding to the same consumption behavior information have a corresponding relationship with the operation object.

[0099] The behavior sequence target association processing on the consumption behavior set, the consumption target subset, the consumption attribute subset, and the behavior sequence subset obtains a marketing content sequence prediction model, and the method comprises the following steps of:

[0100] S341, the consumption behavior set, the consumption target subset, the consumption attribute subset are clustered and analyzed to obtain a consumption category model and consumption category information; the consumption category model is used for processing the consumption behavior set, the consumption target subset, and the consumption attribute subset to obtain the consumption category information;

[0101] S342, obtaining a consumption behavior set corresponding to each consumption category information; determining a behavior sequence subset corresponding to a network operation subset in the consumption behavior set corresponding to the consumption category information as the behavior sequence subset of the consumption category information;

[0102] S343, the behavior sequence subset of each consumption category information is subjected to state space behavior transition process modeling to obtain a behavior sequence transition model of the consumption category information; the behavior sequence transition model is used for processing input consumption category information to obtain a marketing content prediction sequence;

[0103] S344, the consumption category model and the behavior sequence transition model are used to construct a marketing content sequence prediction model;

[0104] The behavior sequence subset of each consumption category information is subjected to state space behavior transition process modeling to obtain a behavior sequence transition model of the consumption category information, and the method comprises the following steps of:

[0105] The behavior sequence subset of each consumption category information is modeled by using a Markov chain to obtain a transition probability of a network content corresponding to each behavior in the behavior sequence subset; all transition probabilities are used to construct a transition probability matrix of all network contents;

[0106] The transition probability matrix is used to construct the behavior sequence transition model of the consumption category information; in the behavior sequence transition model, at each behavior occurrence time, the network content with the maximum transition probability value is the prediction content at this occurrence time, and all prediction contents at all occurrence times are arranged according to time to obtain a marketing content prediction sequence;

[0107] The marketing content sequence prediction model comprises:

[0108] The consumption behavior set, the consumption target subset, the consumption attribute subset are processed by using the consumption category model to obtain the consumption category information;

[0109] The behavior sequence transfer model is used to process the consumption category information to obtain a marketing content prediction sequence;

[0110] The marketing content prediction sequence is a sequence of marketing content that is displayed to the user and that, after being displayed, most promotes the user to generate a consumption behavior. The marketing content sequence is a sequence of network content arranged in chronological order. The network content includes videos, pictures, text, music, and the like.

[0111] The clustering analysis of the consumption behavior set, the consumption target subset, and the consumption attribute subset obtains consumption category information and a consumption category model, including:

[0112] The digital quantization coding of the consumption behavior set, the consumption target subset, and the consumption attribute subset obtains consumption behavior data, consumption target data, and consumption attribute data.

[0113] Each consumption behavior data and corresponding consumption target data and consumption attribute data are represented as a consumption data sequence. The consumption data sequence can be a three-dimensional data [x, y, z].

[0114] All consumption data sequences are used to construct a consumption data set.

[0115] The clustering of the consumption data set obtains data subclass information and consumption data sequences contained in each data subclass.

[0116] Each data subclass information is numbered to obtain a data category value. The mean value of all data in each data subclass is obtained to obtain a consumption feature data sequence of the data subclass.

[0117] A classification optimization model is constructed and solved to obtain a classification feature quantity.

[0118] The classification feature quantity is used to construct a consumption category model.

[0119] The consumption category model is used to process the consumption data sequence corresponding to each consumption behavior data and corresponding consumption target data and consumption attribute data to obtain consumption category information corresponding to the consumption data sequence.

[0120] The expression of the consumption category model is:

[0121]

[0122] wherein z is a classification feature quantity, x i is the i-th consumption data sequence, is the data category value of the i-th consumption data sequence.

[0123] The consumption behavior data, the corresponding consumption target data and the consumption attribute data of each are expressed as a consumption data sequence, and specifically, the consumption behavior data, the corresponding consumption target data and the consumption attribute data are expressed as a three-dimensional data [x, y, z];

[0124] The classification optimization model has an expression as follows:

[0125]

[0126] z|=1,

[0127] wherein z is a classification characteristic quantity to be solved, is a column vector, x ij represents the jth consumption characteristic data sequence of the ith data subclass, x i0 represents the data category value of the ith data subclass, N is the total number of data subclasses, and M is the number of consumption characteristic data sequences contained in a data subclass.

[0128] The classification optimization model has an expression as follows:

[0129] P=W(S0A-R0)

[0130] subject toAA T =I A ,

[0131] wherein I A represents a unit matrix with the row dimension of the matrix A as the dimension, A is a classification characteristic quantity to be solved, P is a first difference matrix, P ij represents an element in the ith row and the jth column of the first difference matrix, W is a weighted transformation matrix, which can adopt the form of a two-dimensional angle discrete matrix, and an element in the ith row and the jth column of the weighted transformation matrix is expressed as W ij =cos(2πi / Nr+θ1)sin(2πj / Ns+θ2), Nr and Ns are both segmentation variables, wherein θ1 and θ2 are starting angles of the weighted transformation matrix; the dimension of the matrix (S0A-R0) is Ns×Nr; S0 is a matrix composed of consumption characteristic data sequences, and R0 is a matrix composed of data category values of data subclasses;

[0132] To solve the classification optimization model and obtain the classification characteristic quantity A, the following steps are included:

[0133] S101, an initialization coding matrix A0 is used as an initial solution to determine an increment matrix ΔA;

[0134] S102, the objective function is expressed as a function f(A) of the matrix A;

[0135] S103, obtaining a first-order partial derivative matrix of f(A) with respect to the elements of matrix A at A0

[0136] S104, constructing a first solving equation of the iterative increment matrix ΔA0:

[0137]

[0138] S105, solving the first solving equation to obtain the value of the iterative increment matrix ΔA0; determining whether |ΔA0| is less than a set threshold value, if yes, determining A0+ΔA0 as the calculation result A; otherwise, replacing A0 with A0+ΔA0 and executing S103;

[0139] The expression of the consumption category model is:

[0140]

[0141] Wherein, A is a classification feature quantity, X i is the i-th consumption data matrix, is the data category vector value of the i-th consumption data matrix.

[0142] The solving of the classification optimization model can adopt an ant colony algorithm or a genetic algorithm.

[0143] The clustering processing can adopt an EM clustering method;

[0144] The clustering analysis can adopt a k-means method or a k-center method;

[0145] The consumption behavior data, the corresponding consumption target data and the consumption attribute data are expressed as a consumption data sequence in a numerical coding manner.

[0146] The consumption prediction processing of the consumption behavior set, the consumption target sub-set and the consumption attribute sub-set to obtain a user consumption prediction model includes:

[0147] S351, respectively performing digital quantization coding processing on the consumption behavior set, the consumption target sub-set and the consumption attribute sub-set to obtain a consumption behavior data set, a consumption target data set and a consumption attribute data set;

[0148] S352, performing function modeling processing on the consumption target data set and the consumption attribute data set as independent variables and the consumption behavior data set as a dependent variable to obtain a user consumption prediction model; the input of the user consumption prediction model is the consumption target data and the consumption attribute data;

[0149] The user network behavior information collected is processed by using the user payment prediction model to obtain network content prediction information and a prediction value of user payment possibility, comprising:

[0150] S41, consumption target information, consumption attribute information and network operation information in the user network behavior information are extracted;

[0151] S42, the network operation information is processed by path association to obtain behavior sequence information;

[0152] S43, the consumption target information, the consumption attribute information and the network operation information are processed by using an initial content prediction model to obtain initial network content prediction information;

[0153] S44, the consumption target information, the consumption attribute information and the behavior sequence information are processed by using a marketing content sequence prediction model to obtain network content prediction sequence information;

[0154] S45, the initial network content prediction information and the network content prediction sequence information are used to constitute network content prediction information;

[0155] S46, the consumption target information and the consumption attribute information are processed by using a user consumption prediction model to obtain a prediction value of user payment possibility;

[0156] The user payment prediction model obtained by analyzing the correlation of the consumption behavior set and the pre-consumption behavior set further comprises:

[0157] The consumption behavior set, the consumption target sub-set and the consumption attribute sub-set are processed by user knowledge graph construction to obtain a consumption knowledge graph;

[0158] The consumption knowledge graph is added to the user payment prediction model;

[0159] The consumption knowledge graph is obtained by processing the consumption behavior set, the consumption target sub-set and the consumption attribute sub-set by user knowledge graph construction, comprising:

[0160] The consumption behavior set, the consumption target sub-set and the consumption attribute sub-set are processed by entity recognition to obtain a consumption entity set; the consumption entity set comprises consumption behavior entities, consumption target entities and consumption attribute entities;

[0161] The consumption entity set is processed by relationship extraction to obtain consumption triples;

[0162] The consumption knowledge graph is constructed by using the consumption entity set and the consumption triples;

[0163] The digital quantization encoding process can adopt UTF-8 encoding or GBK encoding.

[0164] The function modeling process can be realized by using a curve fitting or data interpolation method.

[0165] The user network behavior information set includes a plurality of user network behavior information, and the user network behavior information includes but is not limited to:

[0166] 1. Viewing behavior information (Viewing Behavior):

[0167] Number of views: the total number of times a user watches a video;

[0168] Viewing duration: the duration of a single video viewing and the total viewing duration;

[0169] Completion rate: the proportion of users who watch a video from beginning to end;

[0170] Exit point: the time point at which the user leaves the video;

[0171] 2. Interaction behavior information (Engagement Behavior):

[0172] Like: the number of times a user expresses like for video content;

[0173] Comment: the messages and feedbacks posted by users under the video;

[0174] Share: the number of times a user shares a video to other platforms or social media;

[0175] Favorite: the number of times a user adds a video to a favorite list;

[0176] 3. Subscription and follow information (Subscription and Follow):

[0177] Subscribe to channel: the behavior of a user subscribing to a channel or account of a video publisher;

[0178] Unsubscribe: the behavior of a user unsubscribing, reflecting user loyalty;

[0179] 4. Social interaction information (Social Interaction):

[0180] Tagging: a user tags friends or a specific group when sharing;

[0181] Forward comment: a user adds personal comments or insights when sharing;

[0182] Engagement with topics related to the video content;

[0183] 5. Search Behavior Information (Search Behavior):

[0184] Keyword Search: Keywords and phrases searched by users on the platform;

[0185] Search Frequency: Number and frequency of searches performed by users;

[0186] Search Path: Behavioral path from search to watch;

[0187] 6. Click-through Behavior Information (Click-through Behavior):

[0188] Link Clicks: Number of times users click on links in the video or description;

[0189] Ad Clicks: Number of times users click on pre-roll, mid-roll, or post-roll ads;

[0190] Button and CTA Clicks: Number of times users click on call-to-action buttons like "Learn More" or "Buy";

[0191] 7. Device and Platform Usage Information (Device and Platform Usage):

[0192] Device Type: Devices used by users, such as mobile phones, tablets, computers;

[0193] Operating System and Browser: Operating systems (iOS, Android) and browser types used by users;

[0194] Access Channel: Platforms or applications through which users access video content, such as Douyin, Kuaishou, YouTube;

[0195] 8. Geographical Location Information (Geographical Location):

[0196] Regional Distribution: Cities, regions, or countries where users are located;

[0197] Time Zone and Active Time: Time periods when users are active, related to geographical location;

[0198] 9. User Attribute Information (User Attributes):

[0199] Age Group: Age classification of users;

[0200] Gender: Male, female, or other;

[0201] Interests: Themes or categories of interest to users;

[0202] 10. Conversion Behavior:

[0203] Purchase Behavior: User's purchase behavior after watching the video and the time of occurrence;

[0204] Subscription Behavior: User's registration, subscription to specific content, and the time of occurrence;

[0205] Tip Behavior: User's tip amount, frequency, and time of occurrence for specific content;

[0206] 11. Sentiment Analysis:

[0207] Sentiment Tendency: Analyzing user's sentiment tendency (positive, neutral, negative) through AI analysis of comments and interactive content;

[0208] Emotional Fluctuation: User's emotional change pattern under different content;

[0209] 12. Ad Interaction:

[0210] Ad Viewing Completion Rate: The proportion of users who complete watching the ad;

[0211] Ad Skipping Behavior: The number of times and time points users choose to skip the ad;

[0212] Ad Feedback: User's comments or feedback on the ad content;

[0213] 13. Share Channel Analysis:

[0214] Sharing Platform: Specific platforms where users share videos, such as WeChat, Weibo, Facebook;

[0215] Sharing Effect: Secondary dissemination effect and new user acquisition through sharing;

[0216] 14. Traffic Source:

[0217] Direct Access: User directly enters the website or accesses through favorites;

[0218] Search Engine: Access through Baidu, Google, etc;

[0219] External Link: Access through links from other websites or platforms;

[0220] 15. Bounce Rate and Exit Behavior:

[0221] Churn rate: the proportion of users who leave after entering a page without any interaction;

[0222] Exit page: the last page visited by a user before leaving a website or application;

[0223] 16. Behavioral Pathway:

[0224] Viewing sequence: the order in which users watch videos, and which videos are watched consecutively;

[0225] Jumping behavior: the behavior of users jumping from one video to another;

[0226] ·Dwell time: the time a user spends on a particular page or content;

[0227] 17. Behavioral Sequencing:

[0228] ·User path: the access path and behavior sequence of a user within a website or application.

[0229] ·Key behavior nodes: nodes where users complete key behaviors in the conversion path, such as making a purchase after watching a specific video.

[0230] In a second aspect of the embodiments of the present application, a device for analyzing and predicting user network behavior is disclosed, which comprises:

[0231] a memory storing executable program code;

[0232] a processor coupled to the memory;

[0233] the processor invokes the executable program code stored in the memory to execute the method for analyzing and predicting user network behavior.

[0234] In a third aspect of the embodiments of the present application, a computer storage medium is disclosed, which stores computer instructions, and when the computer instructions are invoked, the method for analyzing and predicting user network behavior is executed.

[0235] In a fourth aspect of the embodiments of the present application, an information data processing terminal is disclosed, which is used to implement the method for analyzing and predicting user network behavior.

[0236] In a fifth aspect of the embodiments of the present application, a device for analyzing and predicting user network behavior is disclosed, which comprises:

[0237] a user network behavior acquisition module, a network content prediction module, and a user payment prediction module;

[0238] The user network behavior acquisition module is connected with the network content prediction module and the user payment prediction module respectively, and is configured to collect user network behavior information;

[0239] The network content prediction module is configured to process the collected user network behavior information to obtain network content prediction information;

[0240] The user payment prediction module is configured to process the collected user network behavior information to obtain a prediction value of user payment possibility;

[0241] The network content prediction module processes the collected user network behavior information to obtain network content prediction information, including:

[0242] Consumption target information, consumption attribute information and network operation information in the user network behavior information are extracted;

[0243] The network operation information is subjected to path association processing to obtain behavior sequence information;

[0244] The consumption target information, the consumption attribute information and the network operation information are processed by using an initial content prediction model to obtain initial network content prediction information;

[0245] The consumption target information, the consumption attribute information and the behavior sequence information are processed by using a marketing content sequence prediction model to obtain network content prediction sequence information;

[0246] The initial network content prediction information and the network content prediction sequence information are used to constitute the network content prediction information;

[0247] The user payment prediction module is configured to process the collected user network behavior information to obtain a prediction value of user payment possibility, including:

[0248] The consumption target information and the consumption attribute information are subjected to consumption prediction processing by using a user consumption prediction model to obtain the prediction value of user payment possibility;

[0249] Figure 3This is a schematic entity of the prediction logic of the present invention. The present invention predicts user network behavior from two dimensions: user and merchant. For the user dimension, the user payment prediction model is used to predict the user's network behavior on the network platform and obtain a payment prediction value. Merchants can use this prediction value to carry out special marketing activities with high conversion rates for specific users with high willingness to pay. For the merchant dimension, the initial content prediction model is used to obtain highly attractive network content to attract users to pay attention to the content displayed by the merchant. In order to continuously attract users, the Markov chain method is used to establish network content prediction sequence information. By continuously establishing user action interaction, the user consumption behavior is promoted.

[0250] It should be noted that since the method of the embodiment of the present application is executed in a computer device, the processing objects of each computer device exist in the form of data or information. For example, time is actually time information. It can be understood that if size, quantity, position, etc. are mentioned in subsequent embodiments, the corresponding data exist for the computer device to process. The details will not be repeated here.

[0251] It should be noted that the artificial intelligence related technologies that may be involved in this application are briefly described. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0252] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0253] The image processing process involved in the embodiments of the present application can be realized by computer vision technology. Computer vision technology (Computer Vision, CV) is a technology that studies how to make machines "see". Further, it refers to using cameras and computers to replace human eyes to identify and measure targets, and further to process images so that the computer processing becomes more suitable for human eye observation or image transmission to instruments for detection. As a scientific discipline, computer vision researches related theories and technologies, and attempts to establish artificial intelligence systems that can obtain information from images or multidimensional data. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, and other technologies. It also includes common face recognition, fingerprint recognition and other biometric identification technologies.

[0254] Single-modal information is only one type of data, such as text, image, audio, video, electromagnetic signal, etc. Multi-modal information is data information including at least two single-modal information. Further, multi-modal information is suitable for complex tasks that require the integration of multiple information sources, such as sentiment analysis, robot interaction, autonomous driving, etc. By integrating information from multiple modalities, higher performance and accuracy can usually be achieved in tasks.

[0255] A large model refers to an artificial neural network model with a very large number of parameters. In the field of artificial intelligence, a large model usually refers to a model with hundreds of millions to trillions of parameters. The model usually needs to be trained on large-scale data sets and requires a large amount of computing resources for optimization and adjustment. Large models are usually used to solve complex natural language processing, computer vision, and speech recognition tasks. Generative AI is an AI that can create new content and ideas, including conversations, stories, images, videos, and music. In the embodiments of the present application, the large model can be a large language model such as ChatGPT, BERT, XLNet, Zhibu model, Claude, Moonshot AI model, ChatGLM model, Qianyitong question model, MiniMax model, Xinghuo model, Llama model, 360GPT model, Qwen model, Baichuan model, Yunque model, vivoLM model, and Wenxin Yiyang, etc. The embodiments of the present application are not limited.

[0256] The above only describes the embodiments of the present application and is not intended to limit the present application. Those skilled in the art can make various modifications and changes to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of the claims of the present application.

Claims

1. A method of predicting analysis of user network behavior, characterized in that, Comprise: S1, obtain a user network behavior information set; S2, extract a consumption behavior set and a pre-consumption behavior set from the user network behavior information set; S3, perform correlation analysis on the consumption behavior set and the pre-consumption behavior set to obtain a user payment prediction model, comprising: S31, classify the consumption behavior set to obtain a consumption target subset, a consumption attribute subset, and a network operation subset; S32, perform path association processing on the network operation subset to obtain a behavior sequence subset; S33, perform consumption behavior target association processing on the consumption behavior set, the consumption target subset, the consumption attribute subset, and the network operation subset to obtain an initial content prediction model, comprising: S331, taking the operation object set in the consumption behavior set and the network operation subset as the target variable, and taking the consumption target subset and the consumption attribute subset as the input variable, performing correlation analysis on the target variable and the input variable to obtain a corresponding relationship model of the input variable and the target variable, comprising: S3311, extract each consumption behavior information in the consumption behavior set, and the corresponding operation object in the network operation subset; use all the extracted operation objects to construct an operation object information set; S3312, according to the corresponding consumption target information and consumption attribute information of each consumption behavior information in the consumption behavior set, establish the corresponding relationship between each consumption target information and consumption attribute information and the operation object in the operation object information set; S3313, respectively perform digital quantization coding processing on the consumption target subset, the consumption attribute subset, and the operation object information set to obtain a consumption target data set, a consumption attribute data set, and an operation object data set; S3314, for each operation object data in the operation object data set, extract the corresponding consumption target data in the consumption target data set and the corresponding consumption attribute data in the consumption attribute data set; S3315, for each operation object data, construct a corresponding two-dimensional data set of all the corresponding consumption target data and consumption attribute data; the two-dimensional data in the two-dimensional data set is the consumption target data and the consumption attribute data; S3316, perform data dimension reduction processing on the corresponding two-dimensional data set of each operation object data to obtain a corresponding reduced two-dimensional data set; the reduced two-dimensional data set comprises consumption target data and consumption attribute data; S3317, based on the corresponding reduced two-dimensional data set of all operation object data, establish the corresponding relationship between each reduced two-dimensional data and operation object data; S3318, according to the corresponding relationship between the reduced two-dimensional data and the operation object data, establish the corresponding relationship between the operation object and the consumption target and the consumption attribute; S3319, determine the corresponding relationship between the operation object and the consumption target and the consumption attribute as the corresponding relationship model of the input variable and the target variable; S332, determine the corresponding relationship model as the initial content prediction model; An output of the initial content prediction model is initial network content prediction information; S34, performing behavior sequence target association processing on the consumption behavior set, the consumption target sub-set, the consumption attribute sub-set, and the behavior sequence sub-set to obtain a marketing content sequence prediction model, including: S341, performing clustering analysis on the consumption behavior set, the consumption target sub-set, and the consumption attribute sub-set to obtain a consumption category model and consumption category information; the consumption category model is used to process the consumption behavior set, the consumption target sub-set, and the consumption attribute sub-set to obtain the consumption category information; S342, performing state space behavior transition process modeling on the behavior sequence sub-set corresponding to each consumption category information to obtain a behavior sequence transition model; S343, constructing the marketing content sequence prediction model by using the consumption category model and the behavior sequence transition model; An output of the marketing content sequence prediction model is network content prediction sequence information; S35, performing consumption prediction processing on the consumption behavior set, the consumption target sub-set, and the consumption attribute sub-set to obtain a user consumption prediction model; an output of the user consumption prediction model is a prediction value of user payment possibility; S36, performing fusion processing on the initial content prediction model, the marketing content sequence prediction model, and the user consumption prediction model to obtain a user payment prediction model; S4, processing the collected user network behavior information by using the user payment prediction model to obtain network content prediction information and a prediction value of user payment possibility.

2. The method of claim 1, wherein the user network behavior is analyzed and predicted based on the user network behavior information. The consumption behavior set and the pre-consumption behavior set are extracted from the user network behavior information set, including: The consumption behavior set is extracted from the user network behavior information set; the consumption behavior set includes a plurality of consumption behaviors and corresponding occurrence times; the consumption behaviors include purchase behaviors, subscription behaviors, and reward behaviors; The pre-consumption behavior set is constructed by using all the non-consumption behaviors extracted according to the occurrence times of the consumption behaviors; the pre-consumption behavior set includes pre-consumption behaviors and corresponding occurrence times.

3. The method of claim 2, wherein the user network behavior is predicted based on the user network behavior of the user network behavior prediction model. The consumption target sub-set includes keywords and occurrence times in search behaviors and click behaviors of users; The consumption attribute sub-set includes device usage information, space activity information, user attribute information, traffic source information, and emotional analysis information; The network operation sub-set includes viewing behavior information, interactive behavior information, subscription and attention information, social interaction information, exit behavior information, sharing channel analysis information, and advertisement interaction information; Each information in the network operation sub-set includes an operation type, an operation object, and an occurrence time; the operation object includes video information, webpage information, tag information, comment information, and advertisement information that are targeted by network operations; all the operation objects constitute an operation object set.

4. An apparatus for analyzing and predicting user network behavior, characterized by, The device includes: a memory storing executable program codes; a processor coupled to the memory; the processor invokes the executable program code stored in the memory to execute the method for analyzing and predicting user network behavior according to any one of claims 1 to 3.

5. A computer storable medium, characterized by The computer storage medium stores computer instructions, which are invoked to execute the method for analyzing and predicting user network behavior according to any one of claims 1 to 3.

6. An information data processing terminal, characterized by The information data processing terminal is used to implement the method for analyzing and predicting user network behavior according to any one of claims 1 to 3.

Images