E-commerce product information recommendation method and system based on big data model

Abstract

The invention discloses an e-commerce product information recommendation method and system based on a big data model, and relates to the technical field of product information recommendation management. The e-commerce product information recommendation method based on the big data model comprises the following steps: obtaining an interaction information evaluation index; obtaining a user access similarity evaluation value; and obtaining a product information similarity evaluation value. According to the method, the first interaction information matrix is constructed, the interaction information evaluation index is obtained, whether the second interaction information matrix is constructed or not is judged on the basis of the obtained interaction information evaluation index, and then whether the user access recommendation data is obtained or not is judged on the basis of the obtained user access similarity evaluation value. And finally, whether the product information recommendation data is obtained is judged based on the obtained product information similarity evaluation value, the effect of improving the relevance between the e-commerce product information and the user in the recommendation process is achieved, and the problem that in the prior art, the relevance between the e-commerce product information and the user is low in the recommendation process is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of product information recommendation management, and in particular to an e-commerce product information recommendation method and system based on a big data model. Background Art

[0002] With the rapid development of information technology and the widespread popularity of the Internet, the e-commerce industry has ushered in unprecedented development opportunities. The types and quantities of products on e-commerce platforms have increased dramatically, providing consumers with a wealth of shopping options. However, this rich choice also brings a problem: how to quickly find products that meet your needs in a huge amount of product information? This requires an efficient and accurate product information recommendation method. The rise of big data technology provides a new solution for e-commerce product information recommendation. By collecting, processing and analyzing massive data, big data technology can mine potential information on consumer behavior, preferences and needs, thereby providing strong support for personalized recommendations. In the field of e-commerce, big data technology can be applied to user behavior analysis, product association analysis, market trend forecasting and other aspects, providing a scientific basis for e-commerce product information recommendation.

[0003] In the existing technology, user portraits and product portraits are constructed by collecting and analyzing user behavior data and product attribute data. Then, machine learning algorithms are used to match users and products to generate personalized recommendation lists, thereby improving the accuracy and satisfaction of recommendations.

[0004] For example, the invention patent with announcement number: CN113592588B announces an e-commerce platform product recommendation system and method based on big data collaborative filtering technology, including: collecting product browsing or transaction records of users on the e-commerce platform, and listing user ratings of products; establishing multi-dimensional attributes according to product categories, and dividing attributes and attribute values ​​for all products in the evaluation table; for each user, extracting evaluated products, counting the counts of each attribute value of all evaluated products under each dimension of attributes, and calculating the mean of each attribute value for each user in combination with the ratings; sorting the calculated similarities in descending order, extracting the users to be recommended corresponding to several similarity values ​​with high ranking as neighboring users, and then further calculating the recommendation index of the products to be recommended based on the neighboring users; sorting the recommendation index in descending order, and selecting products corresponding to several recommendation indexes with high ranking to recommend to the target users.

[0005] For example, the patent application with publication number: CN118735638A discloses a product selection method and system for a cross-border e-commerce platform based on big data, including: monitoring market changes in the cross-border e-commerce target market and generating a market change trend set; obtaining and recording the sales of merchants' historical product selections and generating a historical product selection statistical log; analyzing market change trends based on the market change trend set, and evaluating merchants' historical product selections based on the historical product selection statistical log; accessing the database of the cross-border e-commerce platform and generating a platform information feature set based on platform-related data in the database; building a product selection analysis model, analyzing candidate products provided by merchants based on the product selection analysis model, and generating a product selection recommendation strategy; managing product changes on the cross-border e-commerce platform, and monitoring relevant data on merchants' product selection on the cross-border e-commerce platform, and providing feedback to merchants based on the monitoring data.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the prior art, personalized recommendation systems rely on a large amount of user behavior data to build user portraits and product portraits. However, in actual applications, user and product data are often very sparse (i.e., there are few interaction records between most users and products). This sparsity makes it difficult for the recommendation system to accurately capture user preferences, thereby affecting the recommendation effect. At the same time, for new users or new products, due to the lack of historical behavior data, it is difficult for the recommendation system to make effective recommendations, and there is a problem of low relevance between e-commerce product information and users during the recommendation process. Summary of the invention

[0008] The embodiments of the present application provide an e-commerce product information recommendation method and system based on a big data model, thereby solving the problem of low relevance between e-commerce product information and users during the recommendation process in the prior art, and improving the relevance between e-commerce product information and users during the recommendation process.

[0009] The embodiment of the present application provides an e-commerce product information recommendation method based on a big data model, comprising the following steps: step 1, constructing a first interactive information matrix and obtaining an interactive information evaluation index based on user behavior data and product attribute data of the big data model within a preset time period, and judging whether to construct a second interactive information matrix based on the obtained interactive information evaluation index, wherein the interactive information evaluation index is used to evaluate the matching degree of interactive elements in the first interactive information matrix within the preset time period; step 2, if the second interactive information matrix is ​​constructed, obtaining a user access similarity evaluation value based on the similarity between user behaviors in the second interactive information matrix, and judging whether to obtain user access recommendation data based on the obtained user access similarity evaluation value, wherein the user access similarity evaluation value is used to evaluate the similarity between user access behaviors; step 3, constructing a third interactive information matrix and obtaining a product information similarity evaluation value based on the feature similarity between product information elements in the third interactive information matrix, and judging whether to obtain product information recommendation data based on the obtained product information similarity evaluation value, wherein the product information similarity evaluation value is used to evaluate the similarity between product information on the e-commerce platform.

[0010] Furthermore, the specific steps for obtaining the interaction information evaluation index are as follows: E1, within a preset time period, obtaining the average matrix dimension of the first interaction information matrix, and obtaining the average matrix dimension score when the number of non-zero elements in the rows and columns of the first interaction information matrix is ​​the same; E2, within a preset time period, counting the total number of elements and the total number of interaction elements in the first interaction information matrix to obtain the interaction element ratio; E3, judging whether the interaction element ratio is greater than the preset interaction element ratio, if so, counting the user's access timestamp within the preset time period and executing E4, otherwise no calculation is performed, the access timestamp indicates the duration of the user's visit to the e-commerce platform; E4, judging whether the access timestamp is not less than the preset access timestamp, if so, indicating that there is effective interaction between the corresponding user and the e-commerce platform and counting the number of interaction elements of effective interaction, otherwise no calculation is performed; E5, obtaining the interaction effectiveness evaluation value and the access timestamp score, and combining the obtained average matrix dimension score and interaction element ratio to obtain the interaction information evaluation index.

[0011] Furthermore, the specific restricted expression of the mutual information evaluation index is:

[0012]

[0013] Where t is the number of the preset time period, t = 1, 2, ..., T, T is the total number of preset time periods, i is the number of the number of interactive elements with effective interaction, i = 1, 2, ..., U, U is the total number of interactive elements with effective interaction, e is a natural constant, XI trepresents the interaction information evaluation index of the first interaction information matrix in the tth preset time period, H t represents the number of non-zero elements in the first interactive information matrix in the tth preset time period, L t represents the number of non-zero elements in the first interactive information matrix in the tth preset time period, W t represents the average matrix dimension score of the first interactive information matrix in the tth preset time period, Z t represents the ratio of interactive elements in the first interactive information matrix in the tth preset time period, Z 0 Indicates the preset interactive element ratio, Y t represents the interaction effectiveness evaluation value of the first interaction information matrix in the tth preset time period, Q i.t represents the access timestamp score of the interaction element of the i-th effective interaction in the first interaction information matrix within the t-th preset time period, Q1 i.t Indicates the access timestamp of the interaction element of the i-th valid interaction in the first interaction information matrix within the t-th preset time period, Q1 0 Indicates the preset access timestamp.

[0014] Furthermore, the user access similarity evaluation value is obtained by the following method: counting the number of designated products that are commonly visited by user elements in the second interaction information matrix in the e-commerce platform within a preset time period, and at the same time counting the average visit time of the user elements within the preset time period; processing the obtained number of designated products, average visit time and reference visit interaction time to obtain a user access correlation coefficient, and combining the obtained interaction information evaluation index to obtain a user access similarity evaluation value.

[0015] Furthermore, the product information similarity evaluation value is obtained by the following method: counting the number of user visits to the specified product pair elements in the third interactive information matrix within a preset time period; processing the obtained number of user visits and the reference average number of user visits to obtain a product correlation coefficient, and combining the obtained interactive information evaluation index and the user access similarity evaluation value to obtain the product information similarity evaluation value.

[0016] The embodiment of the present application provides an e-commerce product information recommendation system based on a big data model, including: an interaction information evaluation index acquisition module, a user access similarity evaluation value acquisition module and a product information similarity evaluation value acquisition module; wherein the interaction information evaluation index acquisition module is used to construct a first interaction information matrix and obtain interaction information evaluation indicators based on user behavior data and product attribute data of the big data model within a preset time period, and at the same time, based on the obtained interaction information evaluation indicators, determine whether to construct a second interaction information matrix, and the interaction information evaluation indicators are used to evaluate the matching degree of interaction elements in the first interaction information matrix within a preset time period; the user access similarity evaluation value acquisition module is used to determine whether to construct a second interaction information matrix The second interactive information matrix is ​​used to obtain the user access similarity evaluation value according to the similarity between user behaviors in the second interactive information matrix, and at the same time, determine whether to obtain user access recommendation data based on the obtained user access similarity evaluation value, and the user access similarity evaluation value is used to evaluate the similarity between user access behaviors; the product information similarity evaluation value acquisition module is used to construct the third interactive information matrix and obtain the product information similarity evaluation value according to the feature similarity between product information elements in the third interactive information matrix, and at the same time, determine whether to obtain product information recommendation data based on the obtained product information similarity evaluation value, and the product information similarity evaluation value is used to evaluate the similarity between product information on the e-commerce platform.

[0017] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0018] 1. The first interactive information matrix is ​​constructed through the user behavior data and product attribute data in the big data model and the interactive information evaluation index is obtained. At the same time, based on the obtained interactive information evaluation index, it is determined whether to construct the second interactive information matrix. Then, based on the obtained user access similarity evaluation value, it is determined whether to obtain user access recommendation data. Finally, based on the obtained product information similarity evaluation value, it is determined whether to obtain product information recommendation data. This achieves a more accurate evaluation of the degree of interaction between user access behavior and product information, and further improves the relevance of e-commerce product information to users in the recommendation process, effectively solving the problem of low relevance of e-commerce product information to users in the recommendation process in the prior art.

[0019] 2. By counting the average access time of user elements in the second interactive information matrix within a preset time period and the number of designated products jointly visited in the e-commerce platform, the obtained number of designated products, average access time and reference access interaction time are processed to obtain the user access correlation coefficient, and the user access similarity evaluation value is obtained in combination with the obtained interactive information evaluation index, thereby achieving an improvement in the accuracy of obtaining the user access similarity evaluation value, and further achieving a more accurate evaluation of the similarity between user access behaviors.

[0020] 3. By counting the number of user visits to the specified product elements in the third interactive information matrix within a preset time period, the obtained user visit times and the reference average user visit times are processed to obtain the product correlation coefficient, and the obtained interactive information evaluation index and the user visit similarity evaluation value are combined to obtain the product information similarity evaluation value, thereby achieving an improved accuracy in obtaining the product information similarity evaluation value, and further achieving a more accurate evaluation of the similarity between product information on the e-commerce platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of an e-commerce product information recommendation method based on a big data model provided in an embodiment of the present application;

[0022] Figure 2 A two-dimensional coordinate diagram of a user access correlation coefficient-user access similarity evaluation value provided in an embodiment of the present application;

[0023] Figure 3 A two-dimensional coordinate diagram of the interaction information evaluation index-user access similarity evaluation value provided in the embodiment of the present application;

[0024] Figure 4 A structural diagram of an e-commerce product information recommendation system based on a big data model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The embodiment of the present application solves the problem of low relevance between e-commerce product information and users in the recommendation process in the prior art by providing an e-commerce product information recommendation method and system based on a big data model. A first interaction information matrix is ​​constructed and an interaction information evaluation index is obtained based on user behavior data and product attribute data within a preset time period through a big data model. At the same time, it is determined whether to construct a second interaction information matrix based on the obtained interaction information evaluation index. If the second interaction information matrix is ​​constructed, a user access similarity evaluation value is obtained based on the similarity between user behaviors in the second interaction information matrix, and then user access recommendation data is obtained based on the obtained user access similarity evaluation value. Finally, a third interaction information matrix is ​​constructed and a product information similarity evaluation value is obtained based on the feature similarity between product information elements in the third interaction information matrix. At the same time, product information recommendation data is obtained based on the obtained product information similarity evaluation value, thereby improving the relevance between e-commerce product information and users in the recommendation process.

[0026] The technical solution in the embodiment of the present application is to solve the problem that the e-commerce product information has low relevance to the user during the recommendation process. The overall idea is as follows:

[0027] The first interactive information matrix is ​​constructed by using the user behavior data and product attribute data in the big data model and the interactive information evaluation index is obtained. At the same time, based on the obtained interactive information evaluation index, it is determined whether to construct the second interactive information matrix. Then, based on the obtained user access similarity evaluation value, it is determined whether to obtain user access recommendation data. Finally, based on the obtained product information similarity evaluation value, it is determined whether to obtain product information recommendation data, thereby achieving the effect of improving the relevance of e-commerce product information to users in the recommendation process.

[0028] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0029] like Figure 1 As shown, it is a flow chart of an e-commerce product information recommendation method based on a big data model provided by an embodiment of the present application. An e-commerce product information recommendation method based on a big data model provided by an embodiment of the present application includes the following steps: Step 1, constructing a first interactive information matrix according to user behavior data and product attribute data of the big data model within a preset time period to obtain an interactive information evaluation index, and judging whether to construct a second interactive information matrix based on the obtained interactive information evaluation index, the interactive information evaluation index is used to evaluate the matching degree of interactive elements in the first interactive information matrix within the preset time period, and the interactive elements include user elements and product information elements; Step 2, if the second interactive information matrix is ​​constructed, a user access similarity evaluation value is obtained according to the similarity between user behaviors in the second interactive information matrix, and judging whether to obtain user access recommendation data based on the obtained user access similarity evaluation value, the user access similarity evaluation value is used to evaluate the similarity between user access behaviors; Step 3, constructing a third interactive information matrix and obtaining a product information similarity evaluation value according to the feature similarity between product information elements in the third interactive information matrix, and judging whether to obtain product information recommendation data based on the obtained product information similarity evaluation value, the product information similarity evaluation value is used to evaluate the similarity between product information on the e-commerce platform.

[0030] Among them, user behavior data includes but is not limited to the number of clicks, views, and orders of users; product attribute data includes but is not limited to the name, category, brand, and description of the product; the first interaction information matrix (usually using binary values ​​to indicate whether there is interaction between the user and the product, where 1 indicates interaction and 0 indicates no interaction) is used to visualize the frequency of users visiting the e-commerce platform; the second interaction information matrix is ​​used to visualize the similarity relationship between user behavior data; the third interaction information matrix is ​​used to visualize the similarity relationship between product attribute data; user access recommendation data includes the recommendation time, recommendation method, and recommendation type of the target user; product information recommendation data includes the recommendation time, recommendation method, and recommendation type of the specified product on the e-commerce platform.

[0031] In this embodiment, the recommendation time is recorded in real time by a time counter, and the recommendation method and recommendation type are visually displayed through the recommendation interface of the recommendation system, and each recommendation method and recommendation type are represented by a unique code. Common recommendation methods are usually likes and frequency push in the user's social network, and recommendation types usually include new product recommendations, hot-selling recommendations and discount recommendations. For example, in the second interactive information matrix, the likes in the social network in the recommendation method are "coded as 11", and the frequency push is "coded as 22".

[0032] In the application field of information recommendation on e-commerce platforms, big data models can mine potential associations and patterns between users and product information by analyzing massive amounts of user behavior data and product attribute data. These associations and patterns are used to construct an interactive information matrix, which in turn evaluates the degree of match between user and product information, the degree of similarity between user access behaviors, and the degree of similarity between product information. Based on these evaluation results, the recommendation system can provide users with personalized e-commerce product information recommendations, thereby improving the accuracy of recommendations and user satisfaction.

[0033] It should be supplemented that, by integrating the user behavior data and product output data within a preset time period into a matrix, a first interaction information matrix is ​​obtained, in which the row elements (i.e., user elements) represent visiting users, and the column elements (i.e., product information elements) represent designated products in the e-commerce platform, and one user element represents a visiting user, and one product information element represents a designated product in the e-commerce platform; the behavior data between adjacent users in the big data model are integrated into a new matrix, i.e., the second interaction information matrix, in which each interaction element represents the similarity between two users; the data between product attributes in the big data model are integrated into a new matrix, i.e., the third interaction information matrix, in which each interaction element represents the similarity between two product information.

[0034] This example can more accurately evaluate the similarity between user access behavior and product information by calculating the user access similarity evaluation value and the product information similarity evaluation value, thereby providing users with more accurate recommendations. Secondly, this example constructs three interactive information matrices (including the first interactive information matrix, the second interactive information matrix, and the third interactive information matrix) and determines whether to obtain user access recommendation data and product information recommendation data based on the evaluation data obtained from the three constructed matrices (including interactive information evaluation indicators, user access similarity evaluation values, and product information similarity evaluation values). Compared with the existing technology, this example has higher recommendation accuracy and personalization, ensuring that subsequent e-commerce product information recommendation work is carried out in an orderly manner, thereby improving the relevance of e-commerce product information to users during the recommendation process.

[0035] Furthermore, the specific steps for obtaining the mutual information evaluation index are as follows: E1, within a preset time period, obtaining the average matrix dimension of the first mutual information matrix, and obtaining the average matrix dimension score (i.e., W in the restriction expression of the mutual information evaluation index) when the number of non-zero elements in the rows and columns of the first mutual information matrix is ​​the same t ), the average matrix dimension score is the ratio of the average matrix dimension to the maximum allowed matrix dimension; E2, within a preset time period, the total number of elements in the first interactive information matrix and the total number of interactive elements are counted to obtain the interactive element ratio (i.e., the Z in the restricted expression of the interactive information evaluation index) t ), the total number of elements is the product of the number of user elements and the number of product information elements, and the interactive element ratio is the ratio of the total number of interactive elements to the total number of elements; E3, determine whether the interactive element ratio is greater than the preset interactive element ratio. If so, count the user's access timestamps within the preset time period and execute E4, otherwise no calculation is performed. The access timestamp indicates the duration of the user's access to the e-commerce platform; E4, determine whether the access timestamp is not less than the preset access timestamp. If so, it indicates that the corresponding user has effective interaction with the e-commerce platform and count the number of interactive elements with effective interaction, otherwise no calculation is performed; E5, obtain the interaction effectiveness evaluation value (that is, Y in the restriction expression of the interactive information evaluation index t ) and the access timestamp score (i.e., Q in the restricted expression of the mutual information evaluation index i.t ), and the interaction information evaluation index is obtained by combining the obtained average matrix dimension score and the proportion of interactive elements. The interaction effectiveness evaluation value is the ratio of the number of interactive elements to the total number of interactive elements, and the access timestamp score is the ratio of the difference between the access timestamp and the preset access timestamp to the preset access timestamp.

[0036] In this embodiment, the maximum allowable matrix dimension is the maximum value of the historical matrix dimension of the interaction information matrix in the preset database (stored in the big data model of the recommendation system) within the historical time period, the preset interaction element ratio is represented by the sum and average of the historical interaction element ratios of the interaction elements in the preset database within the historical time period, and the preset access timestamp is represented by the sum and average of the historical access timestamps of the user elements in the preset database to the e-commerce platform within the historical time period.

[0037] It should be understood that the first interaction information matrix is ​​usually used as the input of the recommendation system. Therefore, considering that "the number of non-zero elements in rows and columns in the first interaction information matrix is ​​the same" helps to more comprehensively understand the correlation between the data in the first interaction information matrix, ensuring the balance of the number of non-zero elements in rows and columns (i.e. avoiding the bias of the recommendation system towards certain users or products), thereby achieving improved accuracy and reliability in obtaining interaction information evaluation indicators.

[0038] Furthermore, the specific restricted expression of the mutual information evaluation index is:

[0039]

[0040] Where t is the number of the preset time period, t = 1, 2, ..., T, T is the total number of preset time periods, i is the number of the number of interactive elements with effective interaction, i = 1, 2, ..., U, U is the total number of interactive elements with effective interaction (U ≥ 2), e is a natural constant, XI t represents the interaction information evaluation index of the first interaction information matrix in the tth preset time period, H t represents the number of non-zero elements in the first interactive information matrix in the tth preset time period, L t represents the number of non-zero elements in the first interactive information matrix in the tth preset time period, W t represents the average matrix dimension score of the first interactive information matrix in the t-th preset time period, represents the average matrix dimension of the first interactive information matrix in the tth preset time period, W1 max represents the maximum allowed matrix dimension, Z t represents the ratio of interactive elements in the first interactive information matrix in the tth preset time period, Z 0 Indicates the preset interactive element ratio, Z1 t represents the total number of interactive elements in the first interactive information matrix in the tth preset time period, Z2 t represents the number of user elements in the first interaction information matrix in the tth preset time period, Z3 t represents the number of product information elements in the first interactive information matrix in the tth preset time period, Y t represents the interaction effectiveness evaluation value of the first interaction information matrix in the tth preset time period, Y1 t represents the number of effective interactive elements of the first interactive information matrix in the tth preset time period, Q i.t represents the access timestamp score of the interaction element of the i-th effective interaction in the first interaction information matrix within the t-th preset time period, Q1 i.t Indicates the access timestamp of the interaction element of the i-th valid interaction in the first interaction information matrix within the t-th preset time period, Q1 0 Indicates the preset access timestamp.

[0041] In this embodiment, assuming that the preset interactive element ratio is 0.75 and the preset access timestamp is 5 minutes, the change statistics of the interactive information evaluation index are shown in Table 1:

[0042] Table 1 Statistics of changes in mutual information evaluation indicators

[0043]

[0044]

[0045] It should be understood that, as can be seen from Table 1, the interaction information evaluation index increases with the increase of the interactive element ratio, the interaction effectiveness evaluation value and the access timestamp score, and decreases with the increase of the average matrix dimension score. Among them, the average matrix dimension score increases with the increase of the average matrix dimension, and the access timestamp score increases with the increase of the access timestamp.

[0046] It should be noted that (1) the interaction effectiveness evaluation value also indirectly affects the value of the access timestamp score. When the interaction effectiveness evaluation value increases (i.e., the user's satisfaction with the product increases), it indicates that the user's visit frequency to the e-commerce platform increases, thereby increasing the user's visit timestamp to the e-commerce platform (i.e., the access timestamp score also increases).

[0047] (2) The average matrix dimension score also indirectly affects the value of the interactive element ratio. When the average matrix dimension score increases, the interactive elements in the corresponding first interactive information matrix become more dispersed, that is, the correlation between the interactive elements decreases, which in turn leads to a decrease in the interactive element ratio within the preset time period.

[0048] Through the above-mentioned indirect influence mechanism, the recommendation system can better understand the user's behavior patterns, improve the reliability of subsequent recommended content, and thereby improve the relevance of e-commerce product information to users during the recommendation process, effectively solving the problem of low relevance of e-commerce product information to users during the recommendation process in the prior art.

[0049] Furthermore, the specific process of judging whether to construct the second interactive information matrix based on the obtained interactive information evaluation index is as follows: judging whether the obtained interactive information evaluation index is not less than the preset interactive information evaluation index: if the obtained interactive information evaluation index is not less than the preset interactive information evaluation index, it indicates that the interactive elements in the first interactive information matrix meet the construction requirements of the second interactive information matrix and the second interactive information matrix is ​​constructed; if the obtained interactive information evaluation index is less than the preset interactive information evaluation index, it indicates that the interactive elements in the first interactive information matrix do not meet the construction requirements of the second interactive information matrix. At this time, the second interactive information matrix is ​​not constructed and the changes in the interactive elements in the first interactive information matrix continue to be monitored.

[0050] In this embodiment, the preset interaction information evaluation index is represented by the sum and average of the historical interaction information evaluation indexes of the interaction elements in the first interaction information matrix in the preset database within the historical time period; by setting the preset interaction information evaluation index, this example can ensure that the second interaction information matrix is ​​constructed only when the obtained interaction information evaluation index is not less than the preset interaction information evaluation index, which helps to avoid inaccurate recommendations due to low data quality, thereby achieving accuracy and reliability in the construction of the second interaction information matrix.

[0051] Furthermore, the user access similarity evaluation value is obtained by the following method: when the obtained interaction information evaluation index is not less than the preset interaction information evaluation index, the number of designated products commonly visited by the user elements in the second interaction information matrix in the e-commerce platform within the preset time period is counted, and the average access time of the user elements in the preset time period is counted; the obtained number of designated products, the average access time and the reference access interaction time (represented by the result of summing and averaging the historical average access time of users in the preset database to the e-commerce platform within the historical time period) are processed to obtain the user access correlation coefficient (i.e., SIM(a, b) in the restricted expression of the user access similarity evaluation value), and combined with the obtained interaction information evaluation index (i.e., XI in the restricted expression of the user access similarity evaluation value) t )Get the user access similarity evaluation value.

[0052] In this embodiment, assuming that the preset interaction information evaluation index is 2, the specific restriction expression of the user access similarity evaluation value is:

[0053]

[0054] Where t is the number of the preset time period, t = 1, 2, ..., T, T is the total number of preset time periods, a, b are the numbers of users in the second interaction information matrix, m is the number of the specified product in the e-commerce platform, m = 1, 2, ..., M, M is the number of the specified products visited together, e is a natural constant, HU t represents the user access similarity evaluation value of the user element in the second interaction information matrix in the tth preset time period, XI t represents the interaction information evaluation index of the first interaction information matrix in the tth preset time period, XI 0 represents the preset interaction information evaluation index, SIM(a,b) represents the user access correlation coefficient of the user element in the second interaction information matrix in the tth preset time period, represents the average access time of user element a in the second interaction information matrix in the tth preset time period, A represents the average access time of user element b in the second interaction information matrix in the tth preset time period, 0 Indicates the reference average access duration.

[0055] Specifically, (assuming the preset mutual information evaluation index is 2) Figure 2 As shown, it is a two-dimensional coordinate diagram of the user access correlation coefficient-user access similarity evaluation value provided by the embodiment of the present application (the interaction information evaluation index is a fixed value of 1.8 at this time), by Figure 2 It can be seen that the user access similarity evaluation value increases with the increase of the user access correlation coefficient; Figure 3 As shown in FIG. 1 , a two-dimensional coordinate diagram of the interaction information evaluation index-user access similarity evaluation value provided in the embodiment of the present application (the user access correlation coefficient is a fixed value of 0.8 at this time), is obtained by Figure 3 It can be seen that the user access similarity evaluation value increases with the increase of the interaction information evaluation index.

[0056] It should be noted that the interactive information evaluation index also indirectly affects the value of the user access correlation coefficient. Assuming there are two users a and b, when the interactive information evaluation index increases, it indicates that the frequency of interaction between users a and b and the e-commerce platform increases. At this time, the corresponding behavioral similarity between users a and b increases, that is, the user access correlation coefficient will also increase. By considering the above indirect influence mechanism, it helps the e-commerce platform to more accurately understand the similarities and differences between user behaviors, thereby achieving an improvement in the relevance of e-commerce product information to users in the recommendation process, and effectively solving the problem of low relevance of e-commerce product information to users in the recommendation process in the prior art.

[0057] Furthermore, the specific process of determining whether to obtain user access recommendation data based on the obtained user access similarity evaluation value is as follows: determining whether the obtained user access similarity evaluation value is greater than a preset user access similarity evaluation value: if the obtained user access similarity evaluation value is greater than the preset user access similarity evaluation value, then it indicates that the similarity between user access behaviors in the second interaction information matrix meets the construction requirements of the third interaction information matrix and user access recommendation data is obtained; if the obtained user access similarity evaluation value is not greater than the preset user access similarity evaluation value, then it indicates that the similarity between user access behaviors in the second interaction information matrix does not meet the construction requirements of the third interaction information matrix, and user access recommendation data is not obtained at this time.

[0058] In this embodiment, the preset user access similarity evaluation value is represented by the sum and average of the historical user access similarity evaluation values ​​of the user elements in the second interaction information matrix in the preset database within the historical time period; this example sets the preset user access similarity evaluation value so that the recommendation system can more effectively screen and process user behavior data and reduce unnecessary resource consumption. By determining whether the user access similarity evaluation value is greater than the preset user access similarity evaluation value and obtaining the user access recommendation data when the conditions are met, the accuracy of the recommendation system can be significantly improved as well as the flexibility and adaptability of the recommendation system can be improved.

[0059] Further, the product information similarity evaluation value is obtained by the following method: when the obtained user access similarity evaluation value is greater than the preset user access similarity evaluation value, the number of user accesses to the specified product pair element (for example, the combination of product a and product b) in the third interactive information matrix within a preset time period is counted; the obtained number of user accesses and the reference average number of user accesses are processed to obtain the product correlation coefficient (i.e., the SI in the restriction expression of the product information similarity evaluation value); n.t ), and combined with the obtained mutual information evaluation index (i.e., XI in the restricted expression of product information similarity evaluation value) t ) and the user access similarity evaluation value (i.e., the HU in the restriction expression of the product information similarity evaluation value t ) to obtain the product information similarity evaluation value.

[0060] In this embodiment, assuming that the preset user access similarity evaluation value is 2.25, the specific restriction expression of the product information similarity evaluation value is:

[0061]

[0062] Where t is the number of the preset time period, t = 1, 2, ..., T, T is the total number of preset time periods, m is the number of the specified product, m = 1, 2, ..., M, M is the number of the specified products that are accessed together, n is the number of the specified product pair element, n = 1, 2, ..., N, N is the total number of the specified product pair elements e is a natural constant, GU t represents the product information similarity evaluation value of the specified product pair element in the third interactive information matrix within the t-th preset time period, HU t represents the user access similarity evaluation value of the user element in the second interaction information matrix in the tth preset time period, HU 0 Represents the preset user access similarity evaluation value, XI t represents the interaction information evaluation index of the first interaction information matrix in the tth preset time period, SI n.trepresents the product correlation coefficient of the nth specified product pair element in the third interactive information matrix in the tth preset time period, SI1 n.t represents the number of user visits to the nth specified product pair element in the third interactive information matrix within the tth preset time period, Represents the reference average number of user visits (represented by summing and averaging the historical average number of user visits of the designated product elements in the third interactive information matrix in the preset database within the historical time period).

[0063] It should be understood that the product information similarity evaluation value increases with the increase of the interaction information evaluation index, the product relevance coefficient and the user visit similarity evaluation value. Among them, the product relevance coefficient increases with the increase of the number of user visits. When the number of user visits increases, the recommendation system in the e-commerce platform will count more user behavior data, thereby more accurately identifying the relationship between user behaviors and improving the correlation between user behaviors and product information recommendations (that is, as the product relevance coefficient increases, the corresponding product information similarity evaluation value also increases).

[0064] It should be noted that the user access similarity evaluation value also indirectly affects the value of the product relevance coefficient. For example, by calculating the user access similarity evaluation value, the recommendation system finds that user A and user B have similar browsing behaviors in the field of electronic products (for example, they both browsed the pages of smartphones and tablets). Although the specific preferences of the users are different, this similar browsing behavior indicates that they are all looking for electronic products and may have a common interest in certain types of electronic products. As a result, when the user access similarity evaluation value increases, the corresponding product relevance coefficient will also increase.

[0065] By considering the above-mentioned indirect influence mechanism, the recommendation system can more accurately identify the similarities between user behaviors, thereby improving the matching degree between recommended products and user needs, and further improving the relevance of e-commerce product information to users in the recommendation process, effectively solving the problem of low relevance of e-commerce product information to users in the recommendation process in the prior art.

[0066] Furthermore, the specific process of determining whether to obtain product information recommendation data based on the obtained product information similarity evaluation value is as follows: determining whether the obtained product information similarity evaluation value is greater than a preset product information similarity evaluation value: if the obtained product information similarity evaluation value is greater than the preset product information similarity evaluation value, then it indicates that the similarity between product information on the e-commerce platform meets the expected needs of product information recommendation and product information recommendation data is obtained; if the obtained user access similarity evaluation value is not greater than the preset user access similarity evaluation value, then it indicates that the similarity between product information on the e-commerce platform does not meet the expected needs of product information recommendation, and product information recommendation data is not obtained at this time.

[0067] In this embodiment, the preset product information similarity evaluation value is represented by the sum and average of the historical product information similarity evaluation values ​​of the product information elements in the third interactive information matrix in the preset database within the historical time period; if the product information similarity evaluation value is not greater than the preset user access similarity evaluation value, the recommendation system will not make a recommendation, which helps to reduce incorrect recommendations caused by inaccurate data or mismatched user interests, so that the recommendation system can quickly screen out product information that is highly similar to user needs.

[0068] like Figure 4 As shown, it is a structural schematic diagram of an e-commerce product information recommendation system based on a big data model provided in an embodiment of the present application. An e-commerce product information recommendation system based on a big data model provided in an embodiment of the present application includes: an interactive information evaluation index acquisition module, a user access similarity evaluation value acquisition module and a product information similarity evaluation value acquisition module; wherein the interactive information evaluation index acquisition module is used to construct a first interactive information matrix according to user behavior data and product attribute data of the big data model within a preset time period to obtain interactive information evaluation indicators, and at the same time, based on the obtained interactive information evaluation indicators, determine whether to construct a second interactive information matrix, and the interactive information evaluation indicators are used to evaluate the matching degree of interactive elements in the first interactive information matrix within a preset time period, and the interactive elements include user elements and product information elements; the user access similarity evaluation value acquisition module is used to obtain the user access similarity evaluation value according to the similarity between user behaviors in the second interactive information matrix if the second interactive information matrix is ​​constructed, and at the same time, determine whether to obtain user access recommendation data based on the obtained user access similarity evaluation value, and the user access similarity evaluation value is used to evaluate the similarity between user access behaviors; the product information similarity evaluation value acquisition module is used to construct the third interactive information matrix and obtain the product information similarity evaluation value according to the feature similarity between product information elements in the third interactive information matrix, and at the same time, determine whether to obtain product information recommendation data based on the obtained product information similarity evaluation value, and the product information similarity evaluation value is used to evaluate the similarity between product information on the e-commerce platform.

[0069] In this embodiment, the evaluation indicators such as the interaction information evaluation index, the user access similarity evaluation value and the product information similarity evaluation value together constitute the core of the e-commerce product information recommendation system. The interaction and mutual influence between them help to more comprehensively understand the relationship between user behavior and product attributes. At the same time, by considering the correlation between these evaluation indicators, the recommendation system can provide users with more accurate and personalized recommendations, thereby achieving an improvement in the relevance of e-commerce product information to users in the recommendation process, and effectively solving the problem of low relevance of e-commerce product information to users in the recommendation process in the prior art.

[0070] To summarize, the embodiment of the present application constructs a first interaction information matrix and obtains interaction information evaluation indicators through user behavior data and product attribute data in the big data model, and at the same time determines whether to construct a second interaction information matrix based on the obtained interaction information evaluation indicators, and then determines whether to obtain user access recommendation data based on the obtained user access similarity evaluation value, and finally determines whether to obtain product information recommendation data based on the obtained product information similarity evaluation value, thereby achieving a more accurate evaluation of the degree of interaction between user access behavior and product information, and then achieving an improvement in the relevance of e-commerce product information to users in the recommendation process, effectively solving the problem of low relevance of e-commerce product information to users in the recommendation process in the prior art.

[0071] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0073] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0075] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0076] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for recommending e-commerce product information based on a big data model, characterized in that: The following steps are involved: Step 1: construct a first interaction information matrix and obtain an interaction information evaluation index based on the user behavior data and product attribute data of the big data model within a preset time period, and determine whether to construct a second interaction information matrix based on the obtained interaction information evaluation index, wherein the interaction information evaluation index is used to evaluate the matching degree of the interaction elements in the first interaction information matrix within the preset time period; Step 2: if the second interaction information matrix is ​​constructed, a user access similarity evaluation value is obtained according to the similarity between user behaviors in the second interaction information matrix, and at the same time, based on the obtained user access similarity evaluation value, it is determined whether to obtain user access recommendation data, wherein the user access similarity evaluation value is used to evaluate the similarity between user access behaviors; Step three, construct a third interactive information matrix and obtain a product information similarity evaluation value based on the feature similarity between product information elements in the third interactive information matrix, and at the same time, determine whether to obtain product information recommendation data based on the obtained product information similarity evaluation value. The product information similarity evaluation value is used to evaluate the similarity between product information on the e-commerce platform.

2. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The first interaction information matrix is ​​used to visualize the frequency of users visiting the e-commerce platform; The second interaction information matrix is ​​used to visualize similarity relationships between user behavior data; The third interaction information matrix is ​​used to visualize the similarity relationship between product attribute data; The user access recommendation data includes the target user's recommendation time, recommendation method and recommendation type; The product information recommendation data includes the recommendation time, recommendation method and recommendation type of the specified product in the e-commerce platform.

3. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The specific steps for obtaining the interaction information evaluation index are as follows: E1, obtaining an average matrix dimension of the first interactive information matrix within a preset time period, and obtaining an average matrix dimension score when the number of non-zero elements in the rows and columns of the first interactive information matrix is ​​the same; E2, within a preset time period, counting the total number of elements and the total number of interactive elements in the first interactive information matrix to obtain a proportion of interactive elements; E3, determine whether the interactive element ratio is greater than a preset interactive element ratio. If so, count the user's access timestamps within a preset time period and execute E4. Otherwise, no calculation is performed. The access timestamp indicates the duration of the user's access to the e-commerce platform. E4, judging whether the access timestamp is not less than the preset access timestamp, if so, it indicates that the corresponding user has effective interaction with the e-commerce platform and the number of effective interaction elements is counted, otherwise no calculation is performed; E5, obtain the interaction effectiveness evaluation value and access timestamp score, and combine the obtained average matrix dimension score and interactive element ratio to obtain the interaction information evaluation index.

4. The e-commerce product information recommendation method based on a big data model as claimed in claim 3, characterized in that: The specific limiting expression of the interaction information evaluation index is: Where t is the number of the preset time period, t = 1, 2, ..., T, T is the total number of preset time periods, i is the number of the number of interactive elements with effective interaction, i = 1, 2, ..., U, U is the total number of interactive elements with effective interaction, e is a natural constant, XI t represents the interaction information evaluation index of the first interaction information matrix in the tth preset time period, H t represents the number of non-zero elements in the first interactive information matrix in the tth preset time period, L t represents the number of non-zero elements in the first interactive information matrix in the tth preset time period, W t represents the average matrix dimension score of the first interactive information matrix in the tth preset time period, Z t represents the ratio of interactive elements in the first interactive information matrix in the tth preset time period, Z0 represents the preset ratio of interactive elements, and Y t represents the interaction effectiveness evaluation value of the first interaction information matrix in the tth preset time period, Q i.t represents the access timestamp score of the interaction element of the i-th effective interaction in the first interaction information matrix within the t-th preset time period, Q1 i.t It represents the access timestamp of the interaction element of the i-th valid interaction in the first interaction information matrix within the t-th preset time period, and Q10 represents the preset access timestamp.

5. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The specific process of judging whether to construct the second mutual information matrix based on the obtained mutual information evaluation index is as follows: Determine whether the obtained interaction information evaluation index is not less than the preset interaction information evaluation index: If the obtained mutual information evaluation index is not less than the preset mutual information evaluation index, constructing a second mutual information matrix; If the obtained interaction information evaluation index is less than the preset interaction information evaluation index, the second interaction information matrix is ​​not constructed and the changes of the interaction elements in the first interaction information matrix are continuously monitored.

6. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The user access similarity evaluation value is obtained by the following method: Counting the number of designated products that the user elements in the second interaction information matrix have visited together on the e-commerce platform within a preset time period, and counting the average visit duration of the user elements within the preset time period; The obtained number of specified products, average access time and reference access interaction time are processed to obtain the user access correlation coefficient, and the user access similarity evaluation value is obtained in combination with the obtained interaction information evaluation index.

7. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The specific process of determining whether to obtain user access recommendation data based on the obtained user access similarity evaluation value is as follows: Determine whether the obtained user access similarity evaluation value is greater than the preset user access similarity evaluation value: If the obtained user access similarity evaluation value is greater than the preset user access similarity evaluation value, obtaining user access recommendation data; If the obtained user access similarity evaluation value is not greater than the preset user access similarity evaluation value, the user access recommendation data is not obtained.

8. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The product information similarity evaluation value is obtained by the following method: Counting the number of user visits to the specified product-to-element in the third interactive information matrix within a preset time period; The obtained user access times and the reference average user access times are processed to obtain the product correlation coefficient, and the obtained interaction information evaluation index and the user access similarity evaluation value are combined to obtain the product information similarity evaluation value.

9. The e-commerce product information recommendation method based on a big data model as claimed in claim 1, characterized in that: The specific process of determining whether to obtain product information recommendation data based on the obtained product information similarity evaluation value is as follows: Determine whether the obtained product information similarity evaluation value is greater than the preset product information similarity evaluation value: If the obtained product information similarity evaluation value is greater than the preset product information similarity evaluation value, then obtaining product information recommendation data; If the obtained user access similarity evaluation value is not greater than the preset user access similarity evaluation value, the product information recommendation data is not obtained.

10. An e-commerce product information recommendation system based on a big data model, characterized in that: include: An interactive information evaluation index acquisition module, a user access similarity evaluation value acquisition module, and a product information similarity evaluation value acquisition module; The interactive information evaluation index acquisition module is used to construct a first interactive information matrix and obtain interactive information evaluation indexes based on the user behavior data and product attribute data of the big data model within a preset time period, and to determine whether to construct a second interactive information matrix based on the obtained interactive information evaluation indexes. The interactive information evaluation indexes are used to evaluate the matching degree of interactive elements in the first interactive information matrix within a preset time period. The user access similarity evaluation value acquisition module is used to acquire the user access similarity evaluation value according to the similarity between user behaviors in the second interactive information matrix when constructing the second interactive information matrix, and to determine whether to acquire user access recommendation data based on the acquired user access similarity evaluation value, wherein the user access similarity evaluation value is used to evaluate the similarity between user access behaviors; The product information similarity evaluation value acquisition module is used to construct a third interactive information matrix and obtain a product information similarity evaluation value based on the feature similarity between product information elements in the third interactive information matrix, and at the same time, based on the obtained product information similarity evaluation value, determine whether to obtain product information recommendation data. The product information similarity evaluation value is used to evaluate the similarity between product information on the e-commerce platform.

Images