E-commerce platform data analysis method

Abstract

The invention relates to the field of data analysis, in particular to an e-commerce platform data analysis method, and the method comprises the steps: creating a user behavior database and a commodity information database; calculating a feature highlighting representation value for the user behavior feature based on the user behavior feature of the user to determine whether the user behavior feature is a reference behavior feature, determining an association consistency coefficient based on an association degree between a commodity keyword corresponding to the reference behavior feature and a historical commodity keyword corresponding to a historical user behavior feature of the user, so as to judge a pointing category of the user behavior feature; according to the commodity pushing method and device, the accuracy of commodity pushing of the platform end to the user can be improved on the premise that the data reliability is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to an e-commerce platform data analysis method. Background Art

[0002] Consumers' shopping habits are gradually shifting from offline to online, and their demand for personalized products and services is growing. E-commerce platforms can use the massive user behavior data they have accumulated and analyze big data to help them better understand and meet consumer needs, further optimize products and services, provide consumers with a better online shopping experience, and improve operational efficiency and market competitiveness.

[0003] Chinese patent publication number: CN116342230A, discloses an e-commerce data storage platform based on big data analysis, including a data collection and cleaning module, a data storage and management module, and a data analysis and modeling module; the data collection and cleaning module includes a data collection unit and a data cleaning unit; the data collection unit is used to collect commodity-related data; the data cleaning unit includes a data value processing program, a data formatting program, a data integration program, a data deduplication program, and a data sorting program; the data storage and management module includes a database selection and database design unit, a database table building unit, a data import unit, and a data partitioning unit; the data analysis and modeling module analyzes user purchase behavior and commodity attributes based on the data processed by the data collection and cleaning module, and establishes a commodity recommendation model to improve user purchase conversion rate and sales.

[0004] However, there are still the following problems in the prior art:

[0005] Although the method of pushing relevant product information based on the user's behavioral characteristics is considered, it does not take into account that when the user's behavior is unclear, the pushed product information will be inaccurate and unreliable. Summary of the invention

[0006] To this end, the present invention provides an e-commerce platform data analysis method to overcome the problem in the prior art that, although relevant product information is pushed based on user behavior characteristics, it does not consider the unclear direction of user behavior, resulting in inaccurate and unreliable pushed product information.

[0007] To achieve the above object, the present invention provides an e-commerce platform data analysis method, which comprises:

[0008] Create a user behavior database and a product information database, wherein the user behavior database is used to obtain and store user behavior characteristics in real time, and the product information database is used to store the association relationship between each product keyword,

[0009] The user behavior characteristics include a product page refresh frequency and a product page pull completeness for a single product page, and the association relationship is determined based on a number of user behavior characteristics stored in the user behavior database;

[0010] Calculating a feature highlighting representation value for the user behavior feature based on the user behavior feature of the user to determine whether the user behavior feature is a reference behavior feature, and determining a correlation consistency coefficient based on the correlation between the commodity keyword corresponding to the reference behavior feature and the historical commodity keyword corresponding to the historical user behavior feature of the user to determine the pointing category of the user behavior feature;

[0011] Pushing products to users based on the categories of the user behavior characteristics includes:

[0012] Based on the product page corresponding to the user behavior feature, a reference keyword is locked, several derivative keywords associated with the reference keyword are determined, several product pages where the derivative keywords exist are determined, and a push frequency for each of the product pages is determined based on the feature highlighting representation value, so as to push the product page to the user;

[0013] Or, randomly screen the historical user behavior characteristics of the user, determine the corresponding product page, extract the product keywords, and push the product page containing the product keywords to the user.

[0014] Furthermore, the process of determining the association relationship of product keywords includes:

[0015] Calling the user behavior characteristics of several users to determine the product keywords in several product pages browsed by the users at one time;

[0016] Calculate the co-occurrence probability of a single product keyword and the remaining product keywords in a single browsing of a user;

[0017] If the co-occurrence probability between a single product keyword and any product keyword is greater than a predetermined co-occurrence probability threshold, it is determined that there is an association relationship between the single product keyword and any product keyword;

[0018] Among them, a single browsing is when a user starts browsing the product page through the platform end and closes the platform end.

[0019] Furthermore, the process of calculating the feature highlighting representation value for the user behavior feature based on the user behavior feature includes:

[0020] Calculate the ratio of the product page refresh frequency to the preset refresh frequency threshold as the first feature highlighting representation coefficient;

[0021] The completeness of the product page is used as the second feature to highlight the representation coefficient;

[0022] A feature salient characterization value is obtained by weightedly summing the first feature salient characterization coefficient and the second feature salient characterization coefficient.

[0023] Further, determining whether the user behavior feature is a reference behavior feature includes:

[0024] If the feature prominence characterization value is greater than or equal to the feature prominence characterization value threshold, the user behavior feature is determined to be a reference behavior feature.

[0025] Further, the process of determining the correlation consistency coefficient includes,

[0026] Determine the product page corresponding to the user's reference behavior characteristics, and extract product keywords from the product page;

[0027] Determine the historical user behavior characteristics of the user within a predetermined period in the user behavior database, and extract historical product keywords from the product page;

[0028] Determine the average value of the correlation between each of the product keywords and the historical product keywords;

[0029] The association degree mean is determined as the association consistency coefficient.

[0030] Further, determining the direction category of the user behavior feature includes:

[0031] If the correlation consistency coefficient is greater than or equal to the correlation consistency coefficient threshold, the user behavior feature is determined to be a clear pointing category;

[0032] If the correlation consistency coefficient is less than the correlation consistency coefficient threshold, the user behavior feature is determined to be a fuzzy pointing category.

[0033] Furthermore, pushing products to users according to the category of the user behavior characteristics includes:

[0034] If the user behavior feature is a clear pointing category, then based on the product page corresponding to the user behavior feature, a reference keyword is locked, several derivative keywords associated with the reference keyword are determined, several product pages where the derivative keywords exist are determined, and a push frequency for each of the product pages is determined based on the feature highlighting representation value, so as to push the product page to the user;

[0035] If the user behavior feature is a fuzzy pointing category, the historical user behavior features of the user are randomly screened, the corresponding product page is determined to extract product keywords, and the product page containing the product keywords is pushed to the user.

[0036] Furthermore, the process of locking reference keywords on the product page corresponding to the user behavior characteristics includes:

[0037] Determine the product page corresponding to the user behavior feature;

[0038] A number of product keywords are extracted from the product page, and the occurrence frequency of each of the product keywords is calculated, and the product keyword corresponding to the highest frequency is locked as a reference keyword.

[0039] Furthermore, the process of determining a number of derived keywords associated with the reference keyword includes:

[0040] Retrieving from the product information database a number of other product keywords that are associated with the reference keyword;

[0041] The called commodity keywords are determined as derived keywords.

[0042] Further, the push frequency for each of the product pages is determined based on the feature highlighting representation value, including:

[0043] The push frequency for each of the product pages is increased, and the increase in the push frequency is positively correlated with the feature highlighting representation value.

[0044] Compared with the prior art, the present invention creates a user behavior database and a product information database; calculates a feature highlighting representation value for the user behavior feature based on the user behavior feature of the user to determine whether the user behavior feature is a reference behavior feature, determines a correlation consistency coefficient based on the correlation between the product keyword corresponding to the reference behavior feature and the historical product keyword corresponding to the user's historical user behavior feature to determine the direction category of the user behavior feature; and adaptively pushes products to users based on the direction category of the user behavior feature. The present invention can improve the accuracy of product push to users on the platform side while ensuring data reliability.

[0045] In particular, the present invention takes into account the association relationship between several keywords in each product page covered by the platform, and determines other products that the user may have potential purchasing intention through the association relationship. In actual situations, the user behavior can reflect whether the user has potential associated purchasing intention for the product. By identifying the keywords in each product page to build an association relationship, after subsequently locking in the products that the user currently has potential associated purchasing intention, support is provided for identifying the products corresponding to other product keywords that have an association relationship with the product keywords of the above-mentioned products, so as to facilitate accurate product push to users.

[0046] In particular, the present invention calculates a feature-highlighting representation value for a user behavior feature based on the user behavior feature. In actual situations, due to the unclear directionality of a single behavior, it is impossible to clearly determine the user's purchasing intention, which may lead to inaccurate content pushed by the platform. The present application calculates a feature-highlighting representation value for the user behavior feature to characterize the degree of potential purchasing intention reflected by the user behavior, and determines the user behavior feature with clarity as the reference behavior feature. Furthermore, in order to avoid the low accuracy of data represented by the reference behavior feature alone, the correlation consistency coefficient is determined based on the correlation between the product keywords corresponding to the reference behavior feature and the historical product keywords corresponding to the historical user behavior feature, to determine whether the user behavior feature is a clearly directed category, and then adaptively select the product push method to improve the accuracy of product push to users.

[0047] In particular, for clearly pointed categories, consider locking in reference keywords, and determining derivative keywords based on the association between the keywords of each product, considering other potential products to be purchased, and pushing product pages based on the derivative keywords, thereby improving the accuracy of product push to users while ensuring data reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the steps of the e-commerce platform data analysis method according to an embodiment of the invention;

[0049] Figure 2 A logical schematic diagram for determining whether a user behavior feature is a reference behavior feature according to an embodiment of the invention;

[0050] Figure 3 A logical diagram of the categories of user behavior characteristics determined by the embodiment of the invention;

[0051] Figure 4 This is a logical diagram of pushing merchandise to users according to an embodiment of the invention. DETAILED DESCRIPTION

[0052] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0053] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0054] See also Figures 1 to 4 As shown, Figure 1 This is a schematic diagram of the steps of the e-commerce platform data analysis method according to an embodiment of the present invention. Figure 2A logical diagram for determining whether a user behavior feature is a reference behavior feature according to an embodiment of the present invention. Figure 3 A logical diagram of determining the pointing category of user behavior characteristics according to an embodiment of the present invention. Figure 4 This is a logical diagram of pushing products to users according to an embodiment of the present invention. The e-commerce platform data analysis method according to an embodiment of the present invention includes:

[0055] Create a user behavior database and a product information database, wherein the user behavior database is used to obtain and store user behavior characteristics in real time, and the product information database is used to store the association relationship between each product keyword,

[0056] The user behavior characteristics include a product page refresh frequency and a product page pull completeness for a single product page, and the association relationship is determined based on a number of user behavior characteristics stored in the user behavior database;

[0057] Calculating a feature highlighting representation value for the user behavior feature based on the user behavior feature of the user to determine whether the user behavior feature is a reference behavior feature, and determining a correlation consistency coefficient based on the correlation between the commodity keyword corresponding to the reference behavior feature and the historical commodity keyword corresponding to the historical user behavior feature of the user to determine the pointing category of the user behavior feature;

[0058] Pushing products to users based on the categories of the user behavior characteristics includes:

[0059] Based on the product page corresponding to the user behavior feature, a reference keyword is locked, several derivative keywords associated with the reference keyword are determined, several product pages where the derivative keywords exist are determined, and a push frequency for each of the product pages is determined based on the feature highlighting representation value, so as to push the product page to the user;

[0060] Or, randomly screen the historical user behavior characteristics of the user, determine the corresponding product page, extract the product keywords, and push the product page containing the product keywords to the user.

[0061] Specifically, there is no limitation on the specific method of constructing the user behavior database. After obtaining the user's authorization, relevant information about the user browsing the product interface can be collected to obtain the user behavior characteristics, which will not be elaborated here.

[0062] Specifically, the process of determining the association relationship of product keywords includes:

[0063] Calling the user behavior characteristics of several users to determine the product keywords in several product pages browsed by the users at one time;

[0064] Calculate the co-occurrence probability of a single product keyword and the remaining product keywords in a single browsing of a user;

[0065] If the co-occurrence probability between a single product keyword and any product keyword is greater than a predetermined co-occurrence probability threshold, it is determined that there is an association relationship between the single product keyword and any product keyword;

[0066] Among them, a single browsing is when a user starts browsing the product page through the platform end and closes the platform end.

[0067] The predetermined co-occurrence probability threshold is pre-set, wherein the user behavior characteristics of several users are statistically determined to determine the co-occurrence probability of different product keywords in a single browsing of the user, the co-occurrence probability is used as a random variable to construct a probability density function to determine the normal distribution curve, the co-occurrence probability corresponding to the midpoint of the 95% confidence interval is determined, and 0.75 to 0.85 times the co-occurrence probability is used as the predetermined co-occurrence probability threshold.

[0068] It is understandable that the platform side can be any e-commerce platform that provides e-commerce services. When the user side purchases goods, he logs in to the e-commerce platform, browses the product page, and closes the e-commerce platform after completing the browsing. This will not be elaborated here.

[0069] The present invention takes into account the association relationship between several keywords in each product page covered by the platform, and determines other products that the user may have potential purchasing intention through the association relationship. In actual situations, the user behavior can reflect whether the user has potential associated purchasing intention for the product. By identifying the keywords in each product page to build an association relationship, after subsequently locking in the products that the user currently has potential associated purchasing intention, support is provided for identifying the products corresponding to other product keywords that have an association relationship with the product keywords of the above-mentioned products, so as to facilitate accurate product push to users.

[0070] Specifically, the process of calculating the feature highlighting representation value for the user behavior feature based on the user behavior feature includes:

[0071] Calculate the ratio of the product page refresh frequency to the preset refresh frequency threshold as the first feature highlighting representation coefficient;

[0072] The completeness of the product page is used as the second feature to highlight the representation coefficient;

[0073] A feature salient characterization value is obtained by weightedly summing the first feature salient characterization coefficient and the second feature salient characterization coefficient.

[0074] In this embodiment, when performing weighted summation, the weight of the first feature highlighting characterization coefficient is set to 0.6, and the weight of the second feature highlighting characterization coefficient is set to 0.4;

[0075] The refresh frequency threshold P0 is pre-determined by obtaining relevant data on historical user usage of the platform, extracting the refresh frequency of the product page, and solving the average refresh frequency of the product page. The refresh frequency threshold is set to between 1.15 and 1.2 times the average refresh frequency.

[0076] The present invention calculates a feature highlighting representation value for the user behavior feature based on the user's user behavior feature. In actual situations, due to the unclear directionality of a single behavior, it is impossible to clearly determine the user's purchasing intention, which may cause the content pushed by the platform to be inaccurate. The present application calculates a feature highlighting representation value for the user behavior feature to characterize the degree of potential purchasing intention reflected by the user behavior, and determines the user behavior feature with clarity as the reference behavior feature. Furthermore, in order to avoid the low accuracy of the data represented by the reference behavior feature alone, the correlation consistency coefficient is determined based on the correlation between the product keywords corresponding to the reference behavior feature and the historical product keywords corresponding to the historical user behavior feature, and it is determined whether the user behavior feature is a clearly directed category. Subsequently, the product push method is adaptively selected to improve the accuracy of the products pushed to the user.

[0077] Specifically, the area covered by sliding from the top of the product page to the end of the page is determined as the total area of ​​the product page, and the ratio of the area covered by sliding when the user pulls the product page to the total area of ​​the product page is determined as the completeness of the product page pull.

[0078] Specifically, determining whether the user behavior feature is a reference behavior feature includes:

[0079] If the feature prominence characterization value is greater than or equal to the feature prominence characterization value threshold, the user behavior feature is determined to be a reference behavior feature;

[0080] If the feature prominence characterization value is less than the feature prominence characterization value threshold, it is determined that the user behavior feature is not a reference behavior feature.

[0081] The feature highlighting value threshold F0 is selected in the interval [0.75, 0.95].

[0082] Specifically, the process of determining the correlation consistency coefficient includes,

[0083] Determine the product page corresponding to the user's reference behavior characteristics, and extract product keywords from the product page;

[0084] Determine the historical user behavior characteristics of the user within a predetermined period in the user behavior database, and extract historical product keywords from the product page;

[0085] Determine the average value of the correlation between each of the product keywords and the historical product keywords;

[0086] The association degree mean is determined as the association consistency coefficient.

[0087] Specifically, the correlation between product keywords is determined based on historical user behavior characteristics, and the co-occurrence probability of two product keywords in a single browsing in the historical user behavior characteristics is determined as the correlation.

[0088] Specifically, determining the direction category of the user behavior feature includes:

[0089] If the correlation consistency coefficient is greater than or equal to the correlation consistency coefficient threshold, the user behavior feature is determined to be a clear pointing category;

[0090] If the correlation consistency coefficient is less than the correlation consistency coefficient threshold, the user behavior feature is determined to be a fuzzy pointing category.

[0091] The association consistency coefficient threshold is determined based on a predetermined co-occurrence probability threshold, and the association consistency coefficient threshold is set to 0.75 times the predetermined co-occurrence probability threshold.

[0092] Specifically, pushing products to users based on the categories of the user behavior characteristics includes:

[0093] If the user behavior feature is a clear pointing category, then based on the product page corresponding to the user behavior feature, a reference keyword is locked, several derivative keywords associated with the reference keyword are determined, several product pages where the derivative keywords exist are determined, and a push frequency for each of the product pages is determined based on the feature highlighting representation value, so as to push the product page to the user;

[0094] If the user behavior feature is a fuzzy pointing category, the historical user behavior features of the user are randomly screened, the corresponding product page is determined to extract product keywords, and the product page containing the product keywords is pushed to the user.

[0095] Specifically, the process of locking reference keywords on the product page corresponding to the user behavior characteristics includes:

[0096] Determine the product page corresponding to the user behavior feature;

[0097] A number of product keywords are extracted from the product page, and the occurrence frequency of each of the product keywords is calculated, and the product keyword corresponding to the highest frequency is locked as a reference keyword.

[0098] It is understandable that a product page contains several product keywords, and the product keyword that appears most frequently in the product page is obtained and locked as a reference keyword, which will not be repeated here.

[0099] Specifically, the process of determining a number of derived keywords that are associated with the reference keyword includes:

[0100] Retrieving from the product information database a number of other product keywords that are associated with the reference keyword;

[0101] The called commodity keywords are determined as derived keywords.

[0102] For clearly targeted categories, consider locking in reference keywords, and determine derivative keywords based on the associations between the keywords of each product. Consider other potential products for purchase, and push product pages based on the derivative keywords. This improves the accuracy of product push to users while ensuring data reliability.

[0103] Specifically, the push frequency for each product page is determined based on the feature highlighting representation value, including:

[0104] The push frequency for each of the product pages is increased, and the increase in the push frequency is positively correlated with the feature highlighting representation value.

[0105] In this embodiment, optionally,

[0106] The feature salient characterization value F is compared with the first feature salient characterization value comparison threshold F1 and the second feature salient characterization value comparison threshold F2,

[0107] If F>F2, the push frequency increase amount is determined to be the first push frequency increase amount h1, and h1=0.42h0;

[0108] If F1≤F≤F2, the push frequency increase amount is determined to be the second push frequency increase amount h2, and h2=0.37h0 is set;

[0109] If F<F1, the push frequency increase amount is determined to be the third push frequency increase amount h3, and h3=0.26h0;

[0110] Among them, h0 represents the initial push frequency, F1=1.2F0, F2=1.4F0.

[0111] It is understandable that the initial push frequency can be determined based on the total amount of goods on the platform and the sales of goods, which will not be elaborated here.

[0112] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A method for analyzing data on an e-commerce platform, characterized in that: include: Create a user behavior database and a product information database, wherein the user behavior database is used to obtain and store user behavior characteristics in real time, and the product information database is used to store the association relationship between each product keyword, The user behavior characteristics include a product page refresh frequency and a product page pull completeness for a single product page, and the association relationship is determined based on a number of user behavior characteristics stored in the user behavior database; Calculating a feature highlighting representation value for the user behavior feature based on the user behavior feature of the user to determine whether the user behavior feature is a reference behavior feature, and determining a correlation consistency coefficient based on the correlation between the commodity keyword corresponding to the reference behavior feature and the historical commodity keyword corresponding to the historical user behavior feature of the user to determine the pointing category of the user behavior feature; Pushing products to users based on the categories of the user behavior characteristics includes: Based on the product page corresponding to the user behavior feature, a reference keyword is locked, several derivative keywords associated with the reference keyword are determined, several product pages where the derivative keywords exist are determined, and a push frequency for each of the product pages is determined based on the feature highlighting representation value, so as to push the product page to the user; Or, randomly screen the historical user behavior characteristics of the user, determine the corresponding product page, extract the product keywords, and push the product page containing the product keywords to the user.

2. The e-commerce platform data analysis method according to claim 1, characterized in that: The process of determining the association relationship between product keywords includes: Calling the user behavior characteristics of several users to determine the product keywords in several product pages browsed by the users at one time; Calculate the co-occurrence probability of a single product keyword and the remaining product keywords in a single browsing of a user; If the co-occurrence probability between a single product keyword and any product keyword is greater than a predetermined co-occurrence probability threshold, it is determined that there is an association relationship between the single product keyword and any product keyword; Among them, a single browsing is when a user starts browsing the product page through the platform end and closes the platform end.

3. The e-commerce platform data analysis method according to claim 1, characterized in that: The process of calculating the feature highlighting representation value for the user behavior characteristics based on the user's user behavior characteristics includes: Calculate the ratio of the product page refresh frequency to the preset refresh frequency threshold as the first feature highlighting representation coefficient; The completeness of the product page is used as the second feature to highlight the representation coefficient; A feature salient characterization value is obtained by weightedly summing the first feature salient characterization coefficient and the second feature salient characterization coefficient.

4. The e-commerce platform data analysis method according to claim 1, characterized in that: Determining whether the user behavior feature is a reference behavior feature includes: If the feature prominence characterization value is greater than or equal to the feature prominence characterization value threshold, the user behavior feature is determined to be a reference behavior feature.

5. The e-commerce platform data analysis method according to claim 1, characterized in that: The process of determining the correlation coefficient includes, Determine the product page corresponding to the user's reference behavior characteristics, and extract product keywords from the product page; Determine the historical user behavior characteristics of the user within a predetermined period in the user behavior database, and extract historical product keywords from the product page; Determine the average value of the correlation between each of the product keywords and the historical product keywords; The association degree mean is determined as the association consistency coefficient.

6. The e-commerce platform data analysis method according to claim 1, characterized in that: Determine the category of the user behavior feature, including: If the correlation consistency coefficient is greater than or equal to the correlation consistency coefficient threshold, the user behavior feature is determined to be a clear pointing category; If the correlation consistency coefficient is less than the correlation consistency coefficient threshold, the user behavior feature is determined to be a fuzzy pointing category.

7. The e-commerce platform data analysis method according to claim 1, characterized in that: Pushing products to users based on the categories of the user behavior characteristics includes: If the user behavior feature is a clear pointing category, then based on the product page corresponding to the user behavior feature, a reference keyword is locked, several derivative keywords associated with the reference keyword are determined, several product pages where the derivative keywords exist are determined, and a push frequency for each of the product pages is determined based on the feature highlighting representation value, so as to push the product page to the user; If the user behavior feature is a fuzzy pointing category, the historical user behavior features of the user are randomly screened, the corresponding product page is determined to extract product keywords, and the product page containing the product keywords is pushed to the user.

8. The e-commerce platform data analysis method according to claim 1, characterized in that: The process of locking reference keywords on the product page corresponding to the user behavior characteristics includes: Determine the product page corresponding to the user behavior feature; A number of product keywords are extracted from the product page, and the occurrence frequency of each of the product keywords is calculated, and the product keyword corresponding to the highest frequency is locked as a reference keyword.

9. The e-commerce platform data analysis method according to claim 1, characterized in that ,The process of determining several derivative keywords that are associated with the reference keyword includes, Retrieving from the product information database a number of other product keywords that are associated with the reference keyword; The called commodity keywords are determined as derived keywords.

10. The e-commerce platform data analysis method according to claim 1, characterized in that: Determining the push frequency for each of the product pages based on the feature highlighting representation value includes: The push frequency for each of the product pages is increased, and the increase in the push frequency is positively correlated with the feature highlighting representation value.

Images