Personalized stock pushing method and system based on behavior-driven model

Abstract

The invention relates to the technical field of stock systems, in particular to a personalized stock pushing method and system based on a behavior-driven model, and the method comprises the steps: obtaining basic information and asset information of a user, integrating the basic information and asset information into static reference information, and dividing the user to analyze a demand index of the user; operation behavior information and trading behavior information of the user are obtained and integrated into dynamic reference information, and the investment emotion tendency of the user is analyzed; industry trend information of each stock and historical behavior data sent by corresponding other users are obtained, and a basic value corresponding to each stock is generated; training a behavior-driven model based on the demand index and the investment emotion tendency, and inputting the stock information and the basic value of each stock into the behavior-driven model to output a recommendation motivation and a tendency assignment of each stock; and generating a push index of each stock according to the recommendation motivation and the tendency assignment so as to carry out personalized push. The stock pushing method and the stock pushing device have the effect of providing personalized stock pushing based on user behaviors.

Description

Technical Field

[0001] The present application relates to the technical field of stock systems, and in particular to a personalized stock push method and system based on a behavior-driven model. Background Art

[0002] With the continuous development of science and technology and informatization, more and more stock investors choose to buy and sell stocks, obtain stock information, and view stock trends through online stock software.

[0003] Currently, various stock software will recommend and push certain stocks, but the current technology pushes stocks based on the popularity of stock trading, a sharp rise in trends, yields and other logic. On the one hand, the push logic is relatively thin and the accuracy is poor. On the other hand, it is unable to achieve personalized push for different users and does not fully consider user needs and trading habits. Summary of the invention

[0004] In order to provide personalized stock push based on user behavior, the present application provides a personalized stock push method and system based on a behavior-driven model.

[0005] In the first aspect, the present application provides a personalized stock push method based on a behavior-driven model, which adopts the following technical solution: A personalized stock push method based on a behavior-driven model includes the following steps: Obtain the user's basic information and asset information and integrate them into static reference information; Dividing the users based on the static reference information to analyze demand indicators corresponding to the users; Obtain user operation behavior information and trading behavior information and integrate them into dynamic reference information; Analyzing the user's investment sentiment tendency based on the dynamic reference information, wherein the investment sentiment tendency includes independence, conformity and balance; Obtain industry trend information of each stock and corresponding historical behavior data sent by other users, and generate the basic value corresponding to each stock based on the trend information and historical behavior data; Training a behavior-driven model based on the demand index and the investment sentiment tendency, and inputting the stock information and basic value of each stock into the behavior-driven model to output the recommendation motivation and tendency assignment corresponding to each stock; A push index corresponding to each of the stocks is generated according to the recommendation motivation and the tendency assignment, and personalized push is performed on the corresponding stocks based on the push index.

[0006] In some of the embodiments, the basic information includes the user's age, occupation, time of using similar software, investment time, and usage tendency; the asset information includes balance, number of purchases, purchased shares, number of sales, and sold shares.

[0007] In some embodiments, dividing the users based on the static reference information to analyze the demand indicators corresponding to the users includes the following steps: Classifying the user based on the basic information to enter a corresponding classification group; Based on the comparison between the asset information and the asset range stored in the classification group, the corresponding demand cluster is determined, wherein the demand cluster includes a basic demand cluster, a security demand cluster, an attribution demand cluster and an implementation demand cluster; A demand motivation value that meets the triggering requirement is generated based on the achievement value corresponding to the demand cluster, and the demand motivation value is defined as the demand indicator.

[0008] In some of the embodiments, the operation behavior information includes click target stocks, number of clicks, browse target stocks, number of views, browse time, stay target stocks, number of stays, and stay time; the buying and selling behavior information includes buy target stocks, number of buy actions, sell target stocks, number of sell actions, buy maintenance time, and sell maintenance time.

[0009] In some embodiments, analyzing the user's investment sentiment tendency based on the dynamic reference information includes the following steps: Generate several different types of confrontation sets according to the operation behavior information and the buying and selling behavior information, each of the confrontation sets is composed of a pre-position and a post-position, and the confrontation sets include an object confrontation set, a time confrontation set, and a conversion confrontation set; In the object confrontation set, the click object stock, the browse object stock and the stay object stock are added in the front position, and the purchase object stock and the sale object stock are added in the back position, and a first result is generated based on the relationship confrontation result in the object confrontation set; In the time confrontation set, the browsing time and the dwell time are added to the leading position, and the buying duration and the selling duration are added to the trailing position, and a second result is generated based on the relationship confrontation result in the time confrontation set; In the conversion confrontation set, click counts, page views and stay counts are added to the leading position, and buying behavior counts and selling behavior counts are added to the trailing position, and a third result is generated based on the relationship confrontation result in the conversion confrontation set; The confrontation wins and confrontation win rates of the preceding position and the succeeding position are determined based on the first result, the second result, and the third result, and the investment sentiment tendency is obtained based on the confrontation wins and the confrontation win rates.

[0010] In some embodiments, judging the confrontation wins and confrontation win rates of the preceding position and the succeeding position based on the first result, the second result, and the third result, and obtaining the investment sentiment tendency based on the confrontation wins and the confrontation win rates, comprises the following steps: Calculate the first total winning rate of the front position and the second total winning rate of the rear position respectively; Generate a first win weight and a second win weight based on the confrontation wins of the front position and the back position respectively; The first total winning rate is multiplied by the first win weight to obtain the first tendency value of the front position, and the second total winning rate is multiplied by the second win weight to obtain the second tendency value of the back position; When the difference between the first tendency value and the second tendency value is greater than a preset positive upper limit value, the investment sentiment tendency is the herd mentality; when the difference is less than a preset negative upper limit value, the investment sentiment tendency is the independence; when the difference is between the positive upper limit value and the negative upper limit value, the investment sentiment tendency is the balance.

[0011] In some embodiments, obtaining industry trend information of each stock and corresponding historical behavior data sent by other users, and generating the basic value corresponding to each stock based on the trend information and historical behavior data, includes the following steps: generating a value pool for each of the stocks; Obtaining the historical profit data, growth potential and industry overview of each of the stocks to generate the industry trend information, the historical profit data including price-earnings ratio, price-to-book ratio, earnings per share, the growth potential including operating income growth rate and profit growth rate, and the industry overview including market share and industry ranking; Obtain the historical click count, historical browsing count, historical stay count, historical purchase count, historical sale count, and the concentrated time nodes of the behavior of each of the stocks to generate the historical behavior data; The value is increased or decreased based on the industry trend information and the historical behavior data, and the calculated value is added to the value pool to obtain the basic value.

[0012] In some embodiments, the stock information and basic value of each stock are input into the behavior-driven model to output the recommendation motivation and tendency assignment corresponding to each stock, including the following steps: Calculating the matching degree of the stock information corresponding to different demand clusters based on the stock information, and calculating the matching degree and the value proportion of the trend information in the basic value to obtain the recommendation motivation; A market heat ratio is generated based on the historical behavior data corresponding to the stock information, and the market heat ratio is calculated with the value proportion of the historical behavior data in the basic value to obtain the tendency assignment.

[0013] In some of the embodiments, generating a push index corresponding to each of the stocks according to the recommendation motivation and the tendency assignment, and performing personalized push for the corresponding stocks based on the push index, includes the following steps: generating a first weight and a second weight based on the push response data; Multiplying the recommendation motivation by the first weight, multiplying the tendency value by the second weight, and adding the results to obtain the push index; The stocks are sorted and pushed based on the push index in descending order.

[0014] In the second aspect, the present application provides a personalized stock push system based on a behavior-driven model, which adopts the following technical solutions: A personalized stock push system based on a behavior-driven model is used to implement the above method.

[0015] The technical solution provided by the embodiments of the present application has the following technical effects: Create static profiles of users and generate dynamic reference data through their dynamic behavior data; analyze users' investment needs and investment sentiments based on static and dynamic data; collect trend and historical data for each stock on the platform to calculate the value of each stock; input the stock information and value into a model trained by investment needs and investment sentiments to output results, which can indicate the degree of motivation for each stock to solve investment needs and the degree of match with investment sentiments; calculate the push index of each stock based on the two results, and use the push index to push stock information to different users in a personalized manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of the steps of a personalized stock push method based on a behavior-driven model provided in this embodiment. DETAILED DESCRIPTION

[0017] To more clearly understand the purpose, technical solutions and advantages of the present application, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. However, it should be understood by those of ordinary skill in the art that the present application can be implemented without these details. In some cases, in order to avoid unnecessary descriptions that make various aspects of the present application obscure, well-known methods, processes, systems, components and / or circuits that have been described at a higher level will not be described in detail. For those of ordinary skill in the art, it is obvious that various changes can be made to the embodiments disclosed in the present application, and without departing from the principles and scope of the present application, the general principles defined in the present application can be applied to other embodiments and application scenarios. Therefore, the present application is not limited to the embodiments shown, but conforms to the broadest scope consistent with the scope claimed for protection of the present application.

[0018] It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as there is no conflict between them.

[0019] For software implementation, the techniques described herein may be implemented using modules (e.g., procedures, functions, etc.) that perform the functions described herein. The software codes may be stored in a memory unit and executed by a processor. The memory unit may be implemented within the processor or external to the processor, in which case the memory unit may be communicatively coupled to the processor via various means known in the art.

[0020] In the description of this application, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used to distinguish the technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0021] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples.

[0022] like Figure 1As shown, the embodiment of the present application discloses a personalized stock push method based on a behavior-driven model, comprising the following steps: S100, obtaining the user's basic information and asset information and integrating them into static reference information.

[0023] Basic information includes the user's age, occupation, time spent using similar software, investment time, and usage tendencies; asset information includes balance, number of purchases, shares purchased, number of sales, and shares sold.

[0024] This information is relatively fixed for users and can indicate the user's usage profile. By integrating the above data, we can learn the overall classification of users. Different categories of people have different purposes and needs for using stock software.

[0025] S200, dividing users based on static reference information to analyze demand indicators corresponding to the users.

[0026] Through static reference information, users can be specifically classified to obtain specific user population portraits, and based on the classified population they belong to, we can analyze what the user's target needs are through stock software and how to meet the user's needs.

[0027] S300, obtaining the user's operation behavior information and trading behavior information and integrating them into dynamic reference information.

[0028] The operation behavior information includes click target stocks, number of clicks, browse target stocks, number of views, browse time, stay target stocks, number of stays, and stay time. The buying and selling behavior information includes buy target stocks, number of buy actions, sell target stocks, number of sell actions, buy maintenance time, and sell maintenance time.

[0029] The above data are all collected based on the data when users use stock software in real time to perform certain behaviors and actions. This type of data is real-time and variable. The user's behavioral actions can be used to analyze the user's overall investment tendencies and investment mentality.

[0030] S400, analyzing the user's investment sentiment tendency based on the dynamic reference information, where the investment sentiment tendency includes independence, conformity and balance.

[0031] Through comprehensive analysis of dynamic reference information, we can break down each user's behavior-driven investment sentiment. Investment sentiment is represented by the user's expected mentality when making an investment. That is, we can infer what kind of mood the user will be in when investing in stocks based on the user's behavior.

[0032] Independence is characterized by the fact that after analyzing the user's behavior, it is believed that when users use investment software to buy and sell stocks, they will not or generally will not invest based on market trends, other investors' advice, or follow trends, but will make investment decisions based on their own ideas.

[0033] Herd mentality is characterized by the fact that after analyzing the user's behavior, it is believed that when users use investment software to buy and sell stocks, they will or generally will follow the market trends, the advice of other investors, and invest in line with the trend, showing a certain degree of investment compliance.

[0034] Balance is characterized by the fact that after analyzing the user's behavior, it is determined that the user exhibits independence and conformity on average or close to average.

[0035] For different overall investment sentiment tendencies, there are corresponding different stock push plans to match the user's potential needs.

[0036] S500, obtaining industry trend information of each stock and corresponding historical behavior data sent by other users, and generating a basic value corresponding to each stock based on the trend information and historical behavior data.

[0037] Obtain the trend data of each stock connected to the investment software and the historical behavior data of other users on the platform on each stock, and use these data to assign and adjust the value of the stock itself. The better the stock trend and the more past user behaviors on these stocks, the greater the popularity, potential, and attractiveness of these stocks, and the greater their corresponding basic value.

[0038] S600, training a behavior-driven model based on demand indicators and investment sentiment tendencies, and inputting stock information and basic values ​​of each stock into the behavior-driven model to output the recommendation motivation and tendency assignment corresponding to each stock.

[0039] By analyzing the relevant data of users and taking the analyzed demand indicator emotional tendencies as training sets and summarizing them into the behavior-driven model for training, the trained behavior-driven model corresponds to each independent user, on which a stock push logic provision model based on user behavior is constructed through training. By inputting each stock on the platform together with stock information and basic value into the model, the model can infer whether each stock is suitable for push to its corresponding user based on the specific information of the stock and its own value, so as to meet the user's personalized stock information needs and meet the user's stock investment emotional habits.

[0040] Among them, in the embodiment of the present application, the behavior-driven model is a hybrid model composed of LSTM and BERT. LSTM can analyze the user's historical behavior and the stock's historical data to analyze future behavior motivations, and BERT is used to analyze the correlation between the investment sentiment corresponding to the user's investment behavior and subsequent investment decisions.

[0041] Therefore, the recommendation motivation and tendency assignment output by the behavior-driven model are respectively characterized by the size of demand that can be solved when a stock is pushed to the user. The size of the demand serves as the motivation for whether the stock can be pushed, and the tendency assignment is characterized by the degree to which a stock matches the user's investment sentiment habits. The higher the degree of match, the greater the tendency assignment of the stock.

[0042] S700, generating a push index corresponding to each stock according to the recommendation motivation and tendency assignment, and performing personalized push for the corresponding stock based on the push index.

[0043] The push index corresponding to each stock is calculated through the recommendation motivation and tendency assignment output by the behavior-driven model. The higher the push index, the higher the priority of the stock in being pushed.

[0044] After calculating the push index of all stocks, personalized stock push can be provided to each user based on the push index.

[0045] Through the above steps, a static profile of the user is created, and dynamic reference data is generated through the user's dynamic behavior data. The user's investment needs and investment sentiment are analyzed based on the static and dynamic data. The trend and historical data of each stock in the platform are collected to calculate the value of each stock. The stock information and value are input into the model trained by the investment needs and investment sentiment to output the result. The result can represent the degree of motivation for each stock to solve the investment needs and the degree of matching with the investment sentiment. Based on the two results, the push index of each stock is comprehensively calculated, so as to push stock information to different users in a personalized manner.

[0046] In some other embodiments, dividing users based on static reference information to analyze the demand indicators corresponding to the users includes the following steps: S210: Classify the user based on the basic information to enter a corresponding classification group.

[0047] Each user is preliminarily classified according to his / her age, occupation, time spent using similar software, investment time, usage tendency, etc. The classification groups include the elderly, young people, amateurs, professionals, etc. People in different groups have different abilities to use the software, their understanding of investment, and their needs for using investment software.

[0048] S220, comparing the asset information with the asset range stored in the classification group to determine the corresponding demand cluster, the demand cluster including a basic demand cluster, a security demand cluster, an attribution demand cluster and an implementation demand cluster.

[0049] Subsequently, a detailed classification is carried out in the group based on the asset information corresponding to the user. Different assets represent the user's own investment volume on the one hand, and the user's investment desire can be analyzed through the number of assets bought and sold on the other hand. Through the analysis of investment volume and investment desire, each classification group can be further divided into several demand clusters, and the degree of demand in each demand cluster is different.

[0050] The basic demand cluster is characterized by the user's investment demand as the basic usage demand, which is manifested in that the user's balance on the investment software is 0 or less, and the number of stocks bought and sold is 0 or less. The general needs of this type of users are to obtain dynamic stock information and view information through the stock software, and they do not require or have no demand for hot stocks, stock trends, high-yield stocks and other content.

[0051] The security demand cluster represents the user's investment demand for secure investment asset management. This is manifested in the user's balance on the investment software being 0 or less, and the number of stocks they buy and sell is large. This type of user generally needs to use the stock software to manage the stocks they hold in a relatively safe and stable manner, and has less demand for subsequent reinvestment. They will use the stock software to dynamically obtain stock information and view information, and the content obtained and viewed is mostly related to the stocks they currently hold.

[0052] The attribution demand cluster represents the user's investment demand to obtain attribution investment in stocks of related or similar industries through the stocks they hold. It is manifested in that the user has a certain balance on the investment software but the number of stocks they hold and buy and sell is relatively small. Such users often use a large number of stock software to obtain dynamic stock information and check information to learn about the content related to the stocks they hold, and have a certain balance to invest according to the stock trend.

[0053] The realization demand cluster is characterized by the user's investment demand to achieve professional, large-scale stock investment income and work through professional, long-term use of investment software. This type of user has a large balance and a large number of stocks held. This type of user often requires a large amount of stock information data.

[0054] S230, generating a demand motivation value that meets the triggering requirement based on the achievement value corresponding to the demand cluster, and defining the demand motivation value as a demand indicator.

[0055] The achievement value required by the user in the demand cluster is obtained through the classification group into which each user is classified and the specific demand cluster in the classification group. The achievement value indicates the demand motivation value that the user needs in the classification to meet his or her investment needs, so the demand motivation value is defined as the demand indicator.

[0056] In some other embodiments, analyzing the user's investment sentiment tendency based on dynamic reference information includes the following steps: S410, generating several different types of confrontation sets according to the operation behavior information and the trading behavior information, each confrontation set is composed of a pre-position and a post-position, and the confrontation sets include object confrontation sets, time confrontation sets, and conversion confrontation sets.

[0057] Generate several adversarial sets, which are composed of data at two different positions for cross-data performance confrontation, including the front position and the back position for placing data.

[0058] There are different types of adversarial sets. The object adversarial set is characterized by the adversarial relationships between the objects on which different behaviors are applied. The time adversarial set is characterized by the adversarial relationships between the time parameters corresponding to different behaviors. The transformation adversarial set is characterized by the acquisition of corresponding relationships through equivalent transformation between different types of data corresponding to different behaviors, and the adversarial relationships are then used.

[0059] The adversarial calculation of the adversarial set can be performed through the generative adversarial network, which contains a generative model and a discriminative model. The generative model is responsible for capturing the distribution of sample data, while the discriminative model is generally a binary classifier that determines whether the input is real data or a generated sample. The optimization process of this model is a "binary minimax game" problem. During training, one of the parties (the discriminative network or the generative network) is fixed, and the parameters of the other model are updated, and the iterations are repeated. Finally, the generative model can estimate the distribution and tendency of the sample data.

[0060] S420, in the object confrontation set, add click target stocks, browse target stocks and stay target stocks in the front position, add buy target stocks and sell target stocks in the back position, and generate a first result based on the relationship confrontation result in the object confrontation set.

[0061] First, in the object confrontation set, the front position is the click target stock, the browse target stock and the stay target stock, and the back position is added to the buy target stock and the sell target stock.

[0062] The confrontation results between the front position and the back position can be used to determine the proportional distribution results between users' browsing behavior, clicking behavior, staying behavior and users' stock buying and selling behavior. If one side wins, it means that the user is more inclined to perform one type of behavior and less inclined to perform another type of behavior.

[0063] Clicking, browsing, staying and other behaviors are generally prerequisites for buying and selling behaviors, that is, it is believed that users will facilitate the prediction of stock buying and selling through clicking, browsing, staying and other behaviors.

[0064] S430, in the time confrontation set, add browsing time and dwell time in the leading position, add buying duration and selling duration in the trailing position, and generate a second result based on the relationship confrontation result in the time confrontation set.

[0065] In the time confrontation set, the front position is the browsing time and the dwell time, and the back position is the buying duration and the selling duration.

[0066] The confrontation results of the front position and the back position can be used to determine the proportional distribution of the time users spend on dynamically acquiring stock information, studying stock trends and values, and the time they spend on continuous management of bought and sold stocks.

[0067] However, behaviors such as browsing and staying are generally prerequisites for buying and selling behaviors, that is, it is believed that the time users spend on acquiring and researching stock dynamics contributes to their prediction of stock management time.

[0068] S440, in the conversion confrontation set, the number of clicks, the number of views and the number of stays are added to the leading position, and the number of buying behaviors and the number of selling behaviors are added to the trailing position, and a third result is generated based on the relationship confrontation result in the conversion confrontation set.

[0069] In the conversion confrontation set, the leading positions are clicks, views and stays, while the trailing positions are used to add buying behaviors and selling behaviors.

[0070] Through the confrontation results of the front position and the back position, we can judge the conversion ratio distribution of the user's efforts in obtaining dynamic information about stocks and studying the trend and value of stocks, which is ultimately converted into the number of buying and selling behaviors of stocks.

[0071] Browsing and staying behaviors are generally the prerequisites for buying and selling behaviors, that is, it is believed that the user's acquisition and research of stock dynamics will eventually be converted into the prediction of the conversion rate of specific stock buying and selling behaviors. S450, judging the confrontation wins and confrontation win rates of the front position and the back position based on the first result, the second result and the third result, and obtaining the investment sentiment tendency based on the confrontation wins and confrontation win rates.

[0072] The specific victory results of each confrontation set are obtained according to the confrontation results of the three different confrontation sets. The confrontation wins are represented by the number of confrontation wins of the front position and the back position in all confrontation sets. The confrontation win rate is represented by the specific tendency value of the front position and the back position in all confrontation sets.

[0073] The user's investment sentiment tendency is comprehensively judged by the number of wins and the winning rate.

[0074] In some other embodiments, judging the confrontation wins and confrontation win rates of the front position and the back position based on the first result, the second result and the third result, and obtaining the investment sentiment tendency based on the confrontation wins and confrontation win rates includes the following steps: S451, respectively calculating the first total winning rate of the front position and the second total winning rate of the back position.

[0075] The first win rate represents the degree to which the data tendency and distribution are biased towards the front position after the confrontation set is confronted, and the first total win rate represents the sum of the first win rates of the front positions in all confrontation sets.

[0076] The second win rate represents the degree to which the data tendency and distribution are biased towards the post-position after the confrontation set is confronted, and the second total win rate represents the sum of the second win rates of the post-position in all confrontation sets.

[0077] When the first winning rate is greater than the second winning rate, the front position obtains a confrontation win; conversely, when the second winning rate is greater than the first winning rate, the back position obtains a confrontation win.

[0078] S452: Generate a first win weight and a second win weight based on the confrontation wins of the leading position and the trailing position, respectively.

[0079] The first win weight and the second win weight are generated according to the respective confrontation wins of the front position and the rear position, and the more confrontation wins, the greater the corresponding weight. The values ​​of the first win weight and the second win weight are greater than 1 and less than 2.

[0080] The more victories a front or back position has, the more data distribution and tendency in the confrontation set are closer to the front or back position, so the weighting is greater.

[0081] S453, multiplying the first total winning rate by the first win weight to obtain a first tendency value of the front value position, and multiplying the second total winning rate by the second win weight to obtain a second tendency value of the back value position.

[0082] Multiply the total winning rate of the front position and the back position with the corresponding winning field weight to obtain the first tendency value and the second tendency value. This step is to reduce the contradictory error between the two different data.

[0083] In some cases, the front position may obtain all the wins, but the first total win rate is small (the data distribution difference in each confrontation set is not large); in other cases, the front position only obtains one win, but the first total win rate is greater than the second total win rate (the data distribution difference in the confrontation set where the front position wins is extremely large, and the data distribution difference in the confrontation set where the front position does not win is small). In order to avoid the above-mentioned contradictions, it is necessary to generate a weighted value through the wins to make up for the above-mentioned error value.

[0084] S454, when the difference between the first tendency value and the second tendency value is greater than the preset positive upper limit value, the investment sentiment tendency is herd mentality; when the difference is less than the preset negative upper limit value, the investment sentiment tendency is independence; when the difference is between the positive upper limit value and the negative upper limit value, the investment sentiment tendency is balance.

[0085] The difference between the first tendency value and the second tendency value is calculated. When the first tendency value is greater than the second tendency value, and the difference is greater than a preset positive upper limit, it is considered that the user's behavior tends to buy and sell stocks after a certain amount of insight and analysis of stock information, stock market trends, etc. This behavior considers that the user's investment sentiment tends to be herd mentality, because a large amount of the behavior spent is used to obtain existing stock data and make investment behaviors that follow market data; When the second tendency value is greater than the first tendency value and the difference is less than the negative upper limit value, it is considered that the user makes buying and selling behaviors without or with the help of obtaining and analyzing the stock information and stock market trends in the software. In this case, the user's investment sentiment tendency is considered to be independent, because a large number of his buying and selling behaviors are made without the acquisition and assistance of existing stock data.

[0086] When the difference between the first tendency value and the second tendency value is between the positive upper limit and the negative upper limit, it is believed that the cost invested by the user in assisting, acquiring, browsing, etc. the existing stock information is the same or similar to the cost invested in the investment behavior of buying and selling stocks. This behavioral tendency believes that the user's investment sentiment tendency is balanced.

[0087] In other embodiments, obtaining industry trend information of each stock and corresponding historical behavior data sent by other users, and generating a basic value corresponding to each stock based on the trend information and historical behavior data, includes the following steps: S510, generating a value pool for each stock.

[0088] S520, obtain the historical profit data, growth potential and industry overview of each stock to generate industry trend information. The historical profit data includes price-earnings ratio, price-to-book ratio, and earnings per share. The growth potential includes operating income growth rate and profit growth rate. The industry overview includes market share and industry ranking.

[0089] The potential value of a stock is composed of the sum of historical earnings, growth potential and industry profile. Among them, in the historical earnings data, the price-earnings ratio (PE) is represented by the product of the stock price and earnings per share, and the price-to-book ratio (PB) is represented by the product of the stock price and net assets per share.

[0090] The partial value of the stock is calculated through multiple dimensions such as the stock's historical trends, the industry profile of the company corresponding to the stock, and the company's growth potential.

[0091] S530, obtaining the historical number of clicks, historical number of views, historical number of stays, historical number of purchases, historical number of sales, and the concentrated time nodes of the behaviors of each stock to generate historical behavior data.

[0092] The historical number of behaviors of each stock on this software or other software that can access the data, such as historical clicks, historical views and historical stays, can be used to represent the historical visit popularity of this stock. The historical purchase and sales numbers of this stock can represent the current high and low status of the stock and whether it is a hot stock. The concentrated time nodes of the corresponding behaviors of this stock can be used to predict and analyze the suitable buying and selling time stages of the stock.

[0093] Historical behavior data can also be used to calculate part of the stock's value.

[0094] S540, based on the industry trend information and the historical behavior data, the value is increased or decreased respectively, and the calculated value is added to the value pool to obtain the basic value.

[0095] The value corresponding to the stock is calculated by combining the value shown by the industry trend information and the value shown by the historical behavior data. If the stock is a hot stock, has high popularity and discussion volume, and its recent momentum is high, then its corresponding value will be high.

[0096] In other embodiments, the stock information and basic value of each stock are input into the behavior-driven model to output the recommendation motivation and tendency assignment corresponding to each stock, including the following steps: S610, based on the stock information, the matching degree corresponding to different demand clusters is calculated, and the matching degree and the value ratio of the trend information in the basic value are calculated to obtain the recommendation motivation.

[0097] First, the keyword information in the stock information is obtained through the natural language processing algorithm, and the corresponding name, industry, price, increase or decrease, market value and other content of the stock are obtained, and then matched with each demand promotion line to determine the degree of match.

[0098] The matching degree is characterized by the degree to which the key features of stock information are related to the needs of each demand cluster and play a positive role in satisfying the needs. For example, for the belonging demand cluster, the needs expressed by this type of users are that they need a large amount of stock information related to the stocks they hold, so the stocks of other companies in the same or similar industries as the stocks they hold have a higher matching degree with this cluster; and for the security demand cluster, this type of users often use software to manage their held stocks, and their demand for reinvestment needs to be carried out under the condition of safe investment. Therefore, for this type of cluster, stocks with high market value and high growth rate have a higher matching degree.

[0099] S620, generating a market heat ratio based on the historical behavior data corresponding to the stock information, and calculating the market heat ratio and the value proportion of the historical behavior data in the basic value to obtain a tendency assignment.

[0100] The market popularity of each stock is judged through historical behavioral data. The higher the market popularity ratio, the more people click, browse and perform other related actions on the stock, which means the stock is a hot stock.

[0101] By comparing the market heat ratio with the value corresponding to the historical behavior data, we can determine whether the value of the behavior of the historical population is positively correlated with the market heat ratio of the stock. If it is positively correlated, it means that the value corresponding to the historical behavior of the stock needs to be obtained when more people gather or pay attention to the stock together. If it is negatively correlated, it means that the value corresponding to the historical behavior of the stock does not require too many people to gather or pay attention to the stock together.

[0102] The tendency assignment is obtained based on the specific proportion of positive correlation or negative correlation. The higher the tendency assignment, the greater the degree of positive correlation. If the tendency assignment is lower, the degree of negative correlation is greater.

[0103] Among them, if the user's investment sentiment tends to be conformist, he needs to follow the focus or hot stock trend that a large number of investors pay attention to and take action on to make a follow-up investment. Users with this tendency hope that the value of the behavior made by the historical group is positively correlated with the market popularity of the stock. Therefore, the higher the tendency value is, the more suitable it is to push to this type of people; on the contrary, if the user's investment sentiment tends to be independent, he does not want to follow the focus or hot stock trend that a large number of investors pay attention to and take action on to make a follow-up investment. He hopes to make independent investment behavior. Then he hopes that the value of the behavior made by the historical group is negatively correlated with the market popularity of the stock. Therefore, the lower the tendency value is, the more suitable it is to push to this type of people.

[0104] In other embodiments, a push index corresponding to each stock is generated according to the recommendation motivation and the tendency assignment, and personalized push is performed on the corresponding stock based on the push index, including the following steps: S710: Generate a first weight and a second weight based on the push response data.

[0105] The first weight corresponding to the recommendation motivation and the second weight for the tendency assignment are generated by the user's feedback response degree to the stock data pushed by the system in real time. The feedback response degree is characterized by the degree to which the user responds to the stocks pushed by the system, such as clicking, browsing, staying, buying, selling, etc.

[0106] If the user responds to the pushed stocks, it means that the user is interested in the data pushed by the system, and the user pays attention to the stock content pushed after the system intelligently analyzes it, then the corresponding first weight and second weight will be greater.

[0107] S720, multiply the recommendation motivation by the first weight, multiply the tendency value by the second weight, and add the results to obtain a push index.

[0108] The recommendation motivation and tendency assignment output by the model are calculated with the first weight and the second weight respectively, and the two calculation results are added together to obtain the push index.

[0109] S730, sorting and pushing a number of stocks based on the push index in descending order.

[0110] According to the push index, several stocks are sorted before being pushed, and these stocks are pushed to the user end based on the sorting.

[0111] The embodiment of the present application also discloses a personalized stock push system based on a behavior-driven model, which is used to implement the above method.

[0112] The implementation principle is: Create static profiles of users and generate dynamic reference data through their dynamic behavior data; analyze users' investment needs and investment sentiments based on static and dynamic data; collect trend and historical data for each stock on the platform to calculate the value of each stock; input the stock information and value into a model trained by investment needs and investment sentiments to output results, which can indicate the degree of motivation for each stock to solve investment needs and the degree of match with investment sentiments; calculate the push index of each stock based on the two results, and use the push index to push stock information to different users in a personalized manner.

[0113] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the instructions of the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise clearly stated in this document, the execution of these steps is not strictly limited in order and can be performed in other orders.

[0114] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A personalized stock push method based on a behavior-driven model, characterized in that: The following steps are involved: Obtain the user's basic information and asset information and integrate them into static reference information; Dividing the users based on the static reference information to analyze demand indicators corresponding to the users; Obtain user operation behavior information and trading behavior information and integrate them into dynamic reference information; Analyzing the user's investment sentiment tendency based on the dynamic reference information, wherein the investment sentiment tendency includes independence, conformity and balance; Obtain industry trend information of each stock and corresponding historical behavior data sent by other users, and generate the basic value corresponding to each stock based on the trend information and historical behavior data; Training a behavior-driven model based on the demand index and the investment sentiment tendency, and inputting the stock information and basic value of each stock into the behavior-driven model to output the recommendation motivation and tendency assignment corresponding to each stock; A push index corresponding to each of the stocks is generated according to the recommendation motivation and the tendency assignment, and personalized push is performed on the corresponding stocks based on the push index.

2. The personalized stock push method based on the behavior-driven model according to claim 1 is characterized in that: The basic information includes the user's age, occupation, time of using similar software, investment time, and usage tendency; the asset information includes balance, number of purchases, purchased shares, number of sales, and sold shares.

3. The personalized stock push method based on the behavior-driven model according to claim 2 is characterized in that: The user is divided based on the static reference information to analyze the demand indicators corresponding to the users, including the following steps: Classifying the user based on the basic information to enter a corresponding classification group; Based on the comparison between the asset information and the asset range stored in the classification group, the corresponding demand cluster is determined, wherein the demand cluster includes a basic demand cluster, a security demand cluster, an attribution demand cluster and an implementation demand cluster; A demand motivation value that meets the triggering requirement is generated based on the achievement value corresponding to the demand cluster, and the demand motivation value is defined as the demand indicator.

4. The personalized stock push method based on the behavior-driven model according to claim 1 is characterized in that: The operation behavior information includes click target stocks, number of clicks, browse target stocks, number of views, browse time, stay target stocks, number of stays, and stay time; the buying and selling behavior information includes buy target stocks, number of buy behaviors, sell target stocks, number of sell behaviors, buy maintenance time, and sell maintenance time.

5. The personalized stock push method based on the behavior-driven model according to claim 4 is characterized in that: Analyzing the user's investment sentiment tendency based on the dynamic reference information includes the following steps: Generate several different types of confrontation sets according to the operation behavior information and the buying and selling behavior information, each of the confrontation sets is composed of a pre-position and a post-position, and the confrontation sets include an object confrontation set, a time confrontation set, and a conversion confrontation set; In the object confrontation set, the click object stock, the browse object stock and the stay object stock are added in the front position, and the purchase object stock and the sale object stock are added in the back position, and a first result is generated based on the relationship confrontation result in the object confrontation set; In the time confrontation set, the browsing time and the dwell time are added to the leading position, and the buying duration and the selling duration are added to the trailing position, and a second result is generated based on the relationship confrontation result in the time confrontation set; In the conversion confrontation set, click counts, page views and stay counts are added to the leading position, and buying behavior counts and selling behavior counts are added to the trailing position, and a third result is generated based on the relationship confrontation result in the conversion confrontation set; The confrontation wins and confrontation win rates of the preceding position and the succeeding position are determined based on the first result, the second result, and the third result, and the investment sentiment tendency is obtained based on the confrontation wins and the confrontation win rate.

6. The personalized stock push method based on the behavior-driven model according to claim 5 is characterized in that: The method of determining the confrontation wins and confrontation win rates of the preceding position and the following position based on the first result, the second result, and the third result, and obtaining the investment sentiment tendency based on the confrontation wins and the confrontation win rates, comprises the following steps: Calculate the first total winning rate of the front position and the second total winning rate of the rear position respectively; Generate a first win weight and a second win weight based on the confrontation wins of the front position and the back position respectively; The first total winning rate is multiplied by the first win weight to obtain the first tendency value of the front position, and the second total winning rate is multiplied by the second win weight to obtain the second tendency value of the back position; When the difference between the first tendency value and the second tendency value is greater than a preset positive upper limit value, the investment sentiment tendency is the herd mentality; when the difference is less than a preset negative upper limit value, the investment sentiment tendency is the independence; when the difference is between the positive upper limit value and the negative upper limit value, the investment sentiment tendency is the balance.

7. The personalized stock push method based on behavior-driven model according to claim 1, characterized in that: Obtaining industry trend information of each stock and corresponding historical behavior data sent by other users, and generating the basic value corresponding to each stock based on the trend information and historical behavior data, including the following steps: generating a value pool for each of the stocks; Obtaining the historical profit data, growth potential and industry overview of each of the stocks to generate the industry trend information, the historical profit data including price-earnings ratio, price-to-book ratio, earnings per share, the growth potential including operating income growth rate and profit growth rate, and the industry overview including market share and industry ranking; Obtain the historical click count, historical browsing count, historical stay count, historical purchase count, historical sale count, and the concentrated time nodes of the behavior of each of the stocks to generate the historical behavior data; The value is increased or decreased based on the industry trend information and the historical behavior data, and the calculated value is added to the value pool to obtain the basic value.

8. The personalized stock push method based on the behavior-driven model according to claim 7 is characterized in that: Inputting the stock information and basic value of each stock into the behavior-driven model to output the recommendation motivation and tendency assignment corresponding to each stock includes the following steps: Calculating the matching degree of the stock information corresponding to different demand clusters based on the stock information, and calculating the matching degree and the value proportion of the trend information in the basic value to obtain the recommendation motivation; A market heat ratio is generated based on the historical behavior data corresponding to the stock information, and the market heat ratio is calculated with the value proportion of the historical behavior data in the basic value to obtain the tendency assignment.

9. The personalized stock push method based on behavior-driven model according to claim 1, characterized in that: Generating a push index corresponding to each of the stocks according to the recommendation motivation and the tendency assignment, and performing personalized push for the corresponding stocks based on the push index, including the following steps: generating a first weight and a second weight based on the push response data; Multiplying the recommendation motivation by the first weight, multiplying the tendency value by the second weight, and adding the results to obtain the push index; The stocks are sorted and pushed based on the push index in descending order.

10. A personalized stock push system based on a behavior-driven model, characterized in that: Used to implement the method according to any one of claims 1 to 9.

Images