Advertisement recommendation method and system based on big data

Abstract

The invention relates to the technical field of advertisement putting, in particular to an advertisement recommendation method and system based on big data, and the method comprises the steps: obtaining user behavior data of a user on an e-commerce platform; determining interest distribution data of the user and first feature distribution data of the user based on the user interest behavior model and the user behavior data; determining similar users similar to the interest preference of the current user according to the interest distribution data of the user; determining a first advertisement recommendation list in which similar users are interested; determining interests and preferences of the current user, and determining a second advertisement recommendation list interested by the current user; the intersection of the first advertisement recommendation list and the second advertisement recommendation list is solved, a third advertisement recommendation list is obtained, and the third advertisement recommendation list comprises overlapped advertisements in the first advertisement recommendation list and the second advertisement recommendation list; and recommending advertisements to the current user according to the third advertisement recommendation list. According to the invention, the advertisement recommendation accuracy, reliability and flexibility are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement placement, and in particular to an advertisement recommendation method and system based on big data. Background Art

[0002] Big data has the characteristics of huge data volume, fast data generation and processing speed, and diverse data types. With the rapid development of mobile Internet and network terminals, the user penetration rate of mobile Internet in my country has been increasing year by year. The Internet generates a large amount of data every day, which makes Internet publicity very attractive to product merchants and advertising agencies. Therefore, big data technology can be used to mine users' interests and potential needs from the Internet and recommend online advertisements to corresponding users.

[0003] In some scenarios, personalized advertising recommendation algorithms mainly include content-based algorithms and association rule-based algorithms. Among them, the recommendation based on association rules is relatively less personalized for users, and it is difficult to accurately recommend advertisements based on the personal interests of individual users, and the recommendation accuracy is low. The content-based recommendation algorithm mainly promotes text advertisements for users, and the reliability and flexibility of advertisement recommendations are low. Therefore, how to improve the accuracy, reliability and flexibility of advertisement recommendations is a technical problem that technicians in this field need to solve. Summary of the invention

[0004] In order to solve the technical problems of low accuracy, reliability and flexibility of advertisement recommendation, the purpose of the present invention is to provide an advertisement recommendation method and system based on big data. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides an advertising recommendation method based on big data, including: obtaining user behavior data of users on an e-commerce platform; determining the user's interest distribution data and the user's first feature distribution data based on the user's interest behavior model and the user behavior data; determining similar users with similar interest preferences as the current user based on the user's interest distribution data; determining a first advertising recommendation list of interest to similar users based on the interest distribution data and the first feature distribution data of the similar users; determining the current user's interest preferences based on search keywords in the current user's user behavior data, and determining a second advertising recommendation list of interest to the current user based on the current user's interest preferences; finding the intersection of the first advertising recommendation list and the second advertising recommendation list to obtain a third advertising recommendation list for the current user, the third advertising recommendation list including overlapping advertisements in the first advertising recommendation list and the second advertising recommendation list; and recommending advertisements to the current user based on the third advertising recommendation list.

[0006] Optionally, determining a user's interest distribution data and a user's first characteristic distribution data based on a user interest behavior model and user behavior data includes: determining a first hyperparameter in the user interest behavior model, a second hyperparameter, a first number of interests appearing in the user's user behavior data, and a second number of interests appearing in the user behavior data of all users; determining the user's interest distribution data based on the first hyperparameter and the first number; and determining the user's first characteristic distribution data based on the second hyperparameter and the second number.

[0007] Optionally, determining similar users with similar interest preferences to the current user based on the user's interest distribution data includes: determining a first probability density function of the current user's interest distribution data, and a second probability density function of other users' interest distribution data; determining the similarity between the current user and other users based on the first probability density function and the second probability density function; and selecting other users with a similarity greater than a first threshold as similar users.

[0008] Optionally, determining a first advertisement recommendation list of interest to similar users based on the interest distribution data of similar users and the first characteristic distribution data includes: selecting a target interest from the interest distribution data of similar users; obtaining the second characteristic distribution data of all advertisements to be recommended; determining a first correlation based on the first characteristic distribution data of the target interest of similar users and the second characteristic distribution data of the advertisements, the first correlation characterizing the degree of correlation between the target interest and the advertisements to be recommended; and selecting advertisements whose first correlation is greater than a second threshold to obtain a first advertisement recommendation list.

[0009] Optionally, selecting the target interest from the interest distribution data of similar users includes: selecting the first T interests in the interest distribution data of similar users, where T is a natural number greater than 1; and removing repeated interests from the first T interests of all similar users to obtain the target interest.

[0010] Optionally, determining the first correlation based on the first characteristic distribution data of the target interest of similar users and the second characteristic distribution data of the advertisement includes: determining a first mean and a first standard deviation of the first characteristic distribution data of the target interest, and determining a second mean and a second standard deviation of the second characteristic distribution data; calculating a first difference between the first characteristic distribution data of the target interest and the first mean, and a second difference between the second characteristic distribution data and the second mean; determining a first product of the first difference and the second difference, and a second product of the first standard deviation and the second standard deviation; and determining a ratio of the mathematical expectation of the first product to the second product as the first correlation.

[0011] Optionally, determining the interest preferences of the current user based on the search keywords in the user behavior data of the current user includes: counting the third number of all search keywords and the fourth number of each search keyword in the user behavior data of the current user; calculating the ratio of the fourth number of search keywords to the third number to obtain the frequency of occurrence of the search keywords; sorting the frequency of occurrence of each search keyword in descending order, and selecting the top N search keywords in the sort as the interest preferences of the current user, where N is a natural number greater than 1.

[0012] Optionally, determining a second advertisement recommendation list of interest to the current user based on the current user's interest preferences includes: obtaining second characteristic distribution data of all advertisements to be recommended; determining a second correlation based on the first characteristic distribution data of the current user's interest preferences and the second characteristic distribution data of the advertisements, the second correlation characterizing the degree of correlation between the current user's interest preferences and the advertisements to be recommended; and selecting advertisements whose second correlation is greater than a third threshold to obtain a second advertisement recommendation list.

[0013] Optionally, recommending advertisements to the current user based on the third advertisement recommendation list includes: determining from the third advertisement recommendation list the viewed advertisements clicked by the current user; deleting the viewed advertisements from the third advertisement recommendation list to obtain a fourth advertisement recommendation list; and recommending the advertisements in the fourth advertisement recommendation list to the current user.

[0014] In a second aspect, an embodiment of the present application provides an advertising recommendation system based on big data, which includes: a processor and a memory; wherein the memory is used to store computer programs that can be run on the processor; the processor is used to execute the program stored in the memory to implement the steps of the advertising recommendation method based on big data mentioned in the first aspect.

[0015] The present invention has the following beneficial effects: first, the user behavior data of the user on the e-commerce platform is obtained; then, based on the user interest behavior model and the user behavior data, the user's interest distribution data and the user's first characteristic distribution data are determined; then, similar users with similar interest preferences to the current user are determined according to the user's interest distribution data; secondly, a first advertisement recommendation list of interest to the similar users is determined according to the interest distribution data and the first characteristic distribution data of the similar users; and the current user's interest preferences are determined according to the search keywords in the user behavior data of the current user; then, a second advertisement recommendation list of interest to the current user is determined according to the current user's interest preferences; finally, the intersection of the first advertisement recommendation list and the second advertisement recommendation list is obtained to obtain a third advertisement recommendation list for the current user, the third advertisement recommendation list including overlapping advertisements in the first advertisement recommendation list and the second advertisement recommendation list; and advertisements are recommended to the current user according to the third advertisement recommendation list.

[0016] In this way, the embodiment of the present invention can determine the first advertisement recommendation list based on the interest distribution and feature distribution of similar users who have similar interest preferences as the current user. Because the similar users have similar behavioral characteristics and interest orientations as the current user, the advertisements of interest can be recommended to the current user based on the first advertisement recommendation list. Furthermore, the embodiment of the present invention determines the real interest preferences of the current user based on the search keywords of the current user, forms a second advertisement recommendation list, and uses the advertisements that overlap with each other in the first advertisement recommendation list and the second advertisement recommendation list as advertisements recommended to the current user. In this way, the embodiment of the present invention combines the interest preferences of similar users with the interest preferences of the current user to generate an advertisement recommendation list for the personal interests of a single user. Because the similar users have similar behavioral characteristics and interest orientations as the current user, the advertisements of interest can be recommended to the current user based on the first advertisement recommendation list, and combined with the current user's own interest preferences, advertisements of interest can be accurately recommended to the current user, thereby improving the accuracy and reliability of advertisement recommendations. In addition, the embodiment of the present invention is not limited to recommending broad-text advertisements, thereby improving the flexibility of advertisement recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flowchart of an advertisement recommendation method based on big data provided by an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of the structure of a user interest behavior model provided by an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of the structure of an advertisement recommendation system based on big data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a method and system for advertising recommendation based on big data proposed by the present invention, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0022] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0023] The following is a detailed description of a method and system for advertising recommendation based on big data provided by the present invention in conjunction with the accompanying drawings.

[0024] Embodiment 1:

[0025] See also Figure 1 , which shows a flow chart of an advertisement recommendation method based on big data provided by an embodiment of the present invention, including:

[0026] S101, obtaining user behavior data of users on the e-commerce platform.

[0027] Specifically, user behavior data refers to the data generated by users’ search, click, purchase and interaction behaviors on e-commerce platforms. It includes but is not limited to search keywords, clicked ads, purchased products and published posts. User behavior data can reflect users’ consumption habits and interest preferences.

[0028] S102, determining the user's interest distribution data and the user's first feature distribution data based on the user's interest behavior model and the user's behavior data.

[0029] Specifically, the user interest behavior model can extract user interests from a large amount of user behavior data and display them in the form of probability distribution. Among them, the user interest behavior model can be a probabilistic topic model (Latent Dirichlet Allocation, LDA), which is a probabilistic topic model widely used in text mining and natural language processing. Figure 2 As shown, Figure 2 A schematic diagram of a user interest behavior model improved by an embodiment of the present invention, wherein symbol U is the user set of all users, and I is the set of all interests of all users. u is the interest distribution of user u, is the feature distribution of the user's interest. i represents the i-th interest, and t represents the t-th feature of the user's interest. α is the first hyperparameter of the user's interest behavior model, and β is the second hyperparameter of the user's interest behavior model.

[0030] Furthermore, the user's interest distribution data refers to the distribution of all interests of each user, that is, the probability distribution of all interests of each user, and the user's first characteristic distribution data refers to the characteristic distribution of each interest of each user. When determining the user's interest distribution data and the user's first characteristic distribution data, as an optional embodiment of the present invention, first determine the first hyperparameter, the second hyperparameter, the first number of interests appearing in the user's user behavior data, and the second number of the user's interests appearing in the user behavior data of all users in the user behavior data; then determine the user's interest distribution data based on the first hyperparameter and the first number; finally determine the user's first characteristic distribution data based on the second hyperparameter and the second number.

[0031] Specifically, the first hyperparameter and the second hyperparameter can be set by the user in advance for the user interest behavior model, which can be set according to the empirical value or the recommended value in the reference literature. The specific value is determined according to the actual application scenario, and the embodiment of the present application is not limited here. Further, the embodiment of the present invention can express the user's interest distribution data using the following formula:

[0032]

[0033] In the above formula, Represents the interest distribution data of user u, that is, the posterior probability of the interest distribution of user u. Represents the first number of interest i appearing in the user behavior data of user u. i The first hyperparameter representing interest i. |I| represents the number of all interests of all users. It indicates the proportion of the number of interests of user u to the number of interests of all users.

[0034] Furthermore, in the embodiment of the present invention, the first characteristic distribution data of the user is expressed by the following formula:

[0035]

[0036] In the above formula, Represents the user's first feature distribution data, that is, the posterior probability of the feature distribution of the user's interest i. Represents the second number of the t-th feature of interest i that appears in the user behavior data of all users. β t The second hyperparameter representing the tth feature. |V| represents the number of features of a certain interest of all users. The ratio of the number of features that represent a certain interest of a user to the total number of features of all users.

[0037] S103, determining similar users having similar interest preferences to the current user according to the user's interest distribution data.

[0038] Specifically, users with similar interests and behaviors have similar purchasing intentions and consumption behaviors. Therefore, the interest preferences of the current user can be predicted based on the interest preferences of similar users. When determining similar users, the embodiments of the present invention can determine based on the probability density function of the interest distribution or based on the Euclidean distance between the interest distribution data of the users.

[0039] Furthermore, as an optional embodiment of the present invention, determining similar users with similar interest preferences to the current user based on the user's interest distribution data includes: first determining a first probability density function of the current user's interest distribution data, and a second probability density function of other users' interest distribution data; then determining the similarity between the current user and other users based on the first probability density function and the second probability density function; finally, selecting other users with a similarity greater than a first threshold as similar users.

[0040] Specifically, taking user i and user j as an example, the similarity between the current user and other users can be calculated using the following formula:

[0041]

[0042] In the above formula, sim(i,j) represents the similarity between user i and user j. p(x) represents the first probability density function of the interest distribution data of user i. q(x) represents the second probability density function of the interest distribution data of user j. exp() represents an exponential function with a natural constant as the base. log() represents a logarithmic function with a base of 10. It means traversing all the interest distribution data x, and using q(x) to approximate p(x), it can be found that it is non-negative, that is, And when p(x)=q(x), so The larger it is, the greater the difference between p(x) and q(x).

[0043] It is worth noting that the calculation results of using q(x) to approximate p(x) and using p(x) to approximate q(x) are different. Therefore, we need to calculate the reverse again. in, It means traversing all the interest distribution data x, and using p(x) to approximate q(x), it can be found that it is non-negative, that is, And when q(x)=p(x), so The larger the value, the greater the difference between p(x) and q(x). and The smaller the average value, the smaller the difference between user i and user j, and the smaller the similarity between user i and user j.

[0044] Furthermore, after determining the similarity sim(i,j) between user i and user j, the larger the sim(i,j), the smaller the difference between the interest distribution data of user i and user j. In the embodiment of the present invention, the first threshold is denoted as k, and the N users with sim(i,j)>k are identified as the TOP-N nearest neighbors of the current user i, that is, similar users. It should be noted that the size of the first threshold k represents the tolerance for the similarity of user interest distribution, and its value can be selected according to the actual usage scenario and needs. In the embodiment of the present invention, k=0.75.

[0045] S104: Determine a first advertisement recommendation list that the similar users are interested in based on the interest distribution data of the similar users and the first characteristic distribution data.

[0046] Specifically, the first advertisement recommendation list includes at least one advertisement, and the advertisements in the first advertisement recommendation list are closer to the interest preferences of similar users and the current user. When determining the first advertisement recommendation list, as an optional embodiment of the present invention, firstly, a target interest is selected from the interest distribution data of similar users; then, the second characteristic distribution data of all advertisements to be recommended is obtained; then, a first correlation is determined based on the first characteristic distribution data of the target interest of similar users and the second characteristic distribution data of the advertisement, and the first correlation represents the degree of correlation between the target interest and the advertisement to be recommended; finally, advertisements whose first correlation is greater than the second threshold are selected to obtain the first advertisement recommendation list.

[0047] Specifically, after obtaining similar users with similar interest preferences to the current user, the interest distribution data of each similar user can be obtained. When selecting the target interest from the interest distribution data of similar users, first select the first T interests in the interest distribution data of similar users, where T is a natural number greater than 1; then eliminate the repeated interests in the first T interests of all similar users to obtain the target interest.

[0048] Assuming that the number of similar users is N, we select the first T interests in the interest distribution data of each similar user, and we can get N×T interests. It should be noted that there is a high probability that there are repeated interests in these N×T interests, so the repeated interests in N×T interests are eliminated. The repeated interests in N×T interests refer to the interests in which the same interest appears more than once in the N×T interest set.

[0049] Through the above-mentioned embodiment of the present invention, the target interests of similar users are obtained, and then similar users and advertisements that may be of interest to the current user are selected according to the target interests. When selecting advertisements that may be of interest to the current user, the embodiment of the present invention can select according to the correlation between the current user and the advertisements. When determining the first correlation, as an optional embodiment of the present invention, first determine the first mean and first standard deviation of the first characteristic distribution data of the target interest, and determine the second mean and second standard deviation of the second characteristic distribution data; then calculate the first difference between the first characteristic distribution data of the target interest and the first mean, and the second difference between the second characteristic distribution data and the second mean; then determine the first product of the first difference and the second difference, and the second product of the first standard deviation and the second standard deviation; finally determine the ratio of the mathematical expectation of the first product to the second product as the first correlation.

[0050] Specifically, the embodiment of the present invention takes an interest X among N×T interests and an advertisement Y owned by an advertising agency in an e-commerce platform as an example, and the first correlation between the target interest X and the advertisement Y can be expressed by the following formula:

[0051]

[0052] In the above formula, ρ(X,Y) represents the first correlation between target interest X and advertisement Y. E() represents mathematical expectation. Represents the first mean of the first feature distribution data of the target interest X. Represents the second mean of the second characteristic distribution data of advertisement Y. X Represents the first standard deviation of the first feature distribution data of target interest X. σ Y Represents the second standard deviation of the second characteristic distribution data of advertisement Y. The value range of ρ(X,Y) is (-1,1). The closer ρ(X,Y) is to 1, the closer the characteristic distribution of target interest X is to the characteristic distribution of advertisement Y. On the contrary, the closer ρ(X,Y) is to -1, the closer the characteristic distribution of target interest X is to the characteristic distribution of advertisement Y.

[0053] Furthermore, the embodiment of the present invention sets a second threshold, and determines the advertisements with the first correlation greater than the second threshold as the advertisements recommended to the current user. If the first correlation is greater than the second threshold, it means that the greater the correlation between the target interest and the advertisement, the closer the advertisement is to the interest preference of the current user. The value of the second threshold can be determined according to the actual scenario, and the embodiment of the present invention takes the value of 0.8.

[0054] S105, determining the interest preference of the current user according to the search keywords in the user behavior data of the current user, and determining a second advertisement recommendation list that the current user is interested in according to the interest preference of the current user.

[0055] Specifically, the frequency of the search keyword appearing is the behavior probability of the user. The embodiment of the present invention can regard similar search keywords as the same behavior. For example, "jeans" and "overalls" should be regarded as the same search behavior. This can solve the sparsity problem of user behavior data and the cold start user problem, and improve the accuracy of the recommendation algorithm. Therefore, the embodiment of the present invention clusters the search keywords in the user behavior data of the user to obtain n clusters. The clustering method can adopt the K-means clustering algorithm. Then for a cluster composed of i search keywords, the embodiment of the present invention regards the i search keywords in its cluster as the same search behavior, then the value of the nth cluster is:

[0056]

[0057] In the above formula, U n Indicates the value of the nth cluster in the clustered search behavior set. It indicates the behavior probability of the search behavior of the nth cluster. The larger its value, the higher the probability of the user's search behavior corresponding to the nth cluster. i indicates the number of search keywords contained in a cluster in the cluster. represents the set of search behaviors. θ is the user's interest distribution, is the characteristic distribution of user’s interests.

[0058] Furthermore, in order to more conveniently represent the probability of the search behavior of the nth cluster, the frequency of the search keyword is used to represent it in the embodiment of the present invention. And the frequency of the search keyword is used to determine the interest preference of the current user. When determining the interest preference of the current user, as an optional embodiment of the present invention, firstly, the third number of all search keywords and the fourth number of each search keyword in the user behavior data of the current user are counted; then, the ratio of the fourth number of search keywords to the third number is calculated to obtain the frequency of the search keyword; finally, the frequency of each search keyword is sorted in descending order, and the top N search keywords are selected as the interest preference of the current user, where N is a natural number greater than 1.

[0059] Specifically, in the embodiment of the present invention, taking the search keyword n ​​as an example, the frequency of occurrence of the search keyword can be expressed by the following formula:

[0060]

[0061] In the above formula, P(n) represents the frequency of occurrence of search keyword n. i(n) represents the fourth number of search keyword n. Q represents the third number of all search keywords.

[0062] Through the above embodiment of the present invention, the frequencies of all search keywords of the current user are obtained. The frequencies of search keywords are regarded as probabilities, and the overall behavior probability distribution α of the current user can be obtained. Then, the frequencies of each search keyword are sorted in descending order, and the top N search keywords are selected as the interest preferences of the current user.

[0063] Furthermore, after obtaining the interest preference of the current user, a second advertisement recommendation list close to the interest preference is determined. When determining the second advertisement recommendation list, as an optional embodiment of the present invention, the second characteristic distribution data of all advertisements to be recommended is first obtained; then the second correlation is determined based on the first characteristic distribution data of the interest preference of the current user and the second characteristic distribution data of the advertisement, and the second correlation represents the degree of correlation between the interest preference of the current user and the advertisement to be recommended; finally, the advertisement whose second correlation is greater than the third threshold is selected to obtain the second advertisement recommendation list.

[0064] Specifically, the calculation method of the second correlation can refer to the calculation method of the first correlation in the above-mentioned embodiment of the present invention, and the embodiment of the present invention will not be repeated here. Furthermore, the embodiment of the present invention sets a third threshold value, and determines the advertisements whose second correlation is greater than the third threshold value as the advertisements recommended to the current user. If the second correlation is greater than the third threshold value, it means that the greater the correlation between the interest preference of the current user and the advertisement, the closer the advertisement is to the interest preference of the current user. Among them, the value of the third threshold value can be determined according to the actual scenario, and the embodiment of the present invention takes the value of 0.8 here.

[0065] S106, finding the intersection of the first advertisement recommendation list and the second advertisement recommendation list to obtain a third advertisement recommendation list for the current user, the third advertisement recommendation list including advertisements that overlap in the first advertisement recommendation list and the second advertisement recommendation list, and recommending advertisements to the current user according to the third advertisement recommendation list.

[0066] Specifically, after obtaining the first advertisement recommendation list and the second advertisement recommendation list, there may be overlapping advertisements in the first advertisement recommendation list and the second advertisement recommendation list, and these overlapping advertisements can better reflect the user's interest preferences. Therefore, the intersection of the first advertisement recommendation list and the second advertisement recommendation list can be calculated, and the overlapping advertisements in the first advertisement recommendation list and the second advertisement recommendation list can be selected to form a third advertisement recommendation list.

[0067] Furthermore, the third advertisement recommendation list may contain advertisements that the user has already clicked to view, so in order to avoid repeated advertisement recommendations to the user and improve the efficiency of advertisement recommendations, when recommending advertisements to the current user based on the third advertisement recommendation list, the embodiment of the present invention first determines the viewed advertisements that the current user has clicked on from the third advertisement recommendation list.

[0068] Then, the viewed advertisements are deleted from the third advertisement recommendation list to obtain a fourth advertisement recommendation list; and finally, the advertisements in the fourth advertisement recommendation list are recommended to the current user.

[0069] When recommending advertisements in the fourth advertisement recommendation list to the current user, the advertisements in the fourth advertisement recommendation list may be sorted in descending order according to the second relevance, and recommended to the user in descending order of the sorting order.

[0070] The embodiment of the present invention can determine the first advertisement recommendation list based on the interest distribution and feature distribution of similar users who have similar interest preferences as the current user. Since the similar users have similar behavioral characteristics and interest orientations as the current user, advertisements of interest can be recommended to the current user based on the first advertisement recommendation list. Furthermore, the embodiment of the present invention determines the real interest preferences of the current user based on the search keywords of the current user, forms a second advertisement recommendation list, and uses the advertisements that overlap with each other in the first advertisement recommendation list and the second advertisement recommendation list as advertisements recommended to the current user. In this way, the embodiment of the present invention combines the interest preferences of similar users with the interest preferences of the current user to generate an advertisement recommendation list for the personal interests of a single user. Since the similar users have similar behavioral characteristics and interest orientations as the current user, advertisements of interest can be recommended to the current user based on the first advertisement recommendation list, and combined with the interest preferences of the current user, advertisements of interest can be accurately recommended to the current user, thereby improving the accuracy and reliability of advertisement recommendations. In addition, the embodiment of the present invention is not limited to recommending broad-text advertisements, thereby improving the flexibility of advertisement recommendations.

[0071] Embodiment 2:

[0072] Corresponding to the advertisement recommendation method based on big data provided in the above embodiment, based on the same technical concept, the embodiment of the present application also provides an advertisement recommendation system based on big data, and the advertisement recommendation system based on big data is used to execute the advertisement recommendation method based on big data. Figure 3 A schematic diagram of a structure of an advertisement recommendation system based on big data to implement each embodiment of the present application is shown in FIG. Figure 3 The advertising recommendation system based on big data may have relatively large differences due to different configurations or performances, and may include one or more processors 301 and memory 302, the memory 302 is used to store computer programs that can be run on the processor 301, and the processor 301 is used to execute the program stored in the memory 302 to achieve the above Figure 1The steps in the method embodiment are as follows. The memory 302 may be a temporary storage or a permanent storage. The application stored in the memory 302 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the big data-based advertising recommendation system.

[0073] Furthermore, the processor 301 can be configured to communicate with the memory 302 to execute a series of computer executable instructions in the memory 302 on the big data-based advertising recommendation system. The big data-based advertising recommendation system can also include one or more power supplies 303, one or more wired or wireless network interfaces 304, one or more input and output interfaces 305, and one or more keyboards 306.

[0074] Specifically in this embodiment, the advertising recommendation system based on big data includes a processor, a communication interface, a memory and a communication bus; wherein the processor, the communication interface and the memory communicate with each other through the bus; the memory is used to store computer programs; the processor is used to execute the programs stored in the memory to achieve the above Figure 1 The various steps in the method embodiment are similar to those in the method embodiment, and have the beneficial effects of the above method embodiments. To avoid repetition, the embodiments of the present application will not be described again here.

[0075] It should be noted that the big data-based advertising recommendation system provided in the embodiment of the present application and the big data-based advertising recommendation method provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned big data-based advertising recommendation method, and has the same or similar beneficial effects, and the repetitions will not be repeated.

[0076] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. An advertisement recommendation method based on big data, characterized in that: The advertisement recommendation method based on big data includes: Obtain user behavior data on e-commerce platforms; Determine the interest distribution data of the user and the first feature distribution data of the user based on the user interest behavior model and the user behavior data; Determining similar users with similar interest preferences to the current user based on the interest distribution data of the user; Determine a first advertisement recommendation list that the similar user is interested in according to the interest distribution data of the similar user and the first characteristic distribution data; Determine the interest preference of the current user according to the search keywords in the user behavior data of the current user, and determine a second advertisement recommendation list of interest to the current user according to the interest preference of the current user; finding the intersection of the first advertisement recommendation list and the second advertisement recommendation list to obtain a third advertisement recommendation list of the current user, wherein the third advertisement recommendation list includes advertisements that overlap with those in the first advertisement recommendation list and the second advertisement recommendation list; An advertisement is recommended to the current user according to the third advertisement recommendation list.

2. The advertisement recommendation method based on big data according to claim 1, characterized in that: The determining the user's interest distribution data and the user's first feature distribution data based on the user interest behavior model and the user behavior data comprises: Determine a first hyperparameter, a second hyperparameter, a first number of interests appearing in the user behavior data of the user, and a second number of interests of the user appearing in the user behavior data of all users in the user interest behavior model; Determine the interest distribution data of the user according to the first hyperparameter and the first quantity; Determine first characteristic distribution data of the user according to the second hyperparameter and the second quantity.

3. The advertisement recommendation method based on big data according to claim 1, characterized in that: Determining similar users with similar interest preferences to the current user according to the interest distribution data of the user includes: Determine a first probability density function of the interest distribution data of the current user and a second probability density function of the interest distribution data of other users; Determine the similarity between the current user and other users according to the first probability density function and the second probability density function; Other users whose similarity is greater than a first threshold are selected as the similar users.

4. The advertisement recommendation method based on big data according to claim 1, characterized in that: The determining of the first advertisement recommendation list that the similar user is interested in according to the interest distribution data of the similar user and the first characteristic distribution data comprises: Selecting a target interest from the interest distribution data of the similar users; Obtain second feature distribution data of all advertisements to be recommended; Determine a first correlation according to the first characteristic distribution data of the target interest of the similar user and the second characteristic distribution data of the advertisement, wherein the first correlation represents the degree of correlation between the target interest and the advertisement to be recommended; Advertisements whose first relevance is greater than a second threshold are selected to obtain the first advertisement recommendation list.

5. The advertisement recommendation method based on big data according to claim 4, characterized in that: The selecting a target interest from the interest distribution data of the similar users includes: Select the first T interests in the interest distribution data of the similar users, where T is a natural number greater than 1; The repeated interests among the first T interests of all similar users are eliminated to obtain the target interest.

6. The advertisement recommendation method based on big data according to claim 4, characterized in that: The determining of the first correlation according to the first characteristic distribution data of the target interest of the similar user and the second characteristic distribution data of the advertisement comprises: Determine a first mean and a first standard deviation of the first characteristic distribution data of the target interest, and determine a second mean and a second standard deviation of the second characteristic distribution data; Calculate a first difference between the first characteristic distribution data of the target interest and the first mean, and a second difference between the second characteristic distribution data and the second mean; determining a first product of the first difference and the second difference, and a second product of the first standard deviation and the second standard deviation; A ratio of a mathematical expectation of the first product to the second product is determined as the first correlation.

7. The advertisement recommendation method based on big data according to claim 1, characterized in that: Determining the interest preference of the current user according to the search keywords in the user behavior data of the current user includes: Counting a third number of all search keywords and a fourth number of each search keyword in the user behavior data of the current user; Calculating a ratio of the fourth number of the search keywords to the third number to obtain a frequency of occurrence of the search keywords; The frequencies of occurrence of the search keywords are sorted in descending order, and the top N search keywords are selected as the interest preferences of the current user, where N is a natural number greater than 1.

8. The method for advertising recommendation based on big data according to any one of claims 1 to 7, characterized in that: Determining the second advertisement recommendation list that the current user is interested in according to the interest preference of the current user comprises: Obtain second feature distribution data of all advertisements to be recommended; Determine a second correlation according to the first characteristic distribution data of the interest preference of the current user and the second characteristic distribution data of the advertisement, wherein the second correlation represents the degree of correlation between the interest preference of the current user and the advertisement to be recommended; Advertisements whose second relevance is greater than a third threshold are selected to obtain the second advertisement recommendation list.

9. The advertisement recommendation method based on big data according to any one of claims 1 to 7, characterized in that: The recommending an advertisement to the current user according to the third advertisement recommendation list comprises: Determining from the third advertisement recommendation list the viewed advertisement clicked by the current user; Deleting the viewed advertisement from the third advertisement recommendation list to obtain a fourth advertisement recommendation list; Recommending advertisements in the fourth advertisement recommendation list to the current user.

10. An advertising recommendation system based on big data, characterized in that: The big data-based advertising recommendation system includes: a processor and a memory; wherein the memory is used to store computer programs that can be run on the processor; the processor is used to execute the program stored in the memory to implement the steps of the big data-based advertising recommendation method as described in any one of claims 1-9.

Images