Electronic commerce data pushing method and system based on big data

Abstract

The invention provides an electronic commerce data pushing method and system based on big data. The method comprises the following steps: receiving a real-time browsing behavior data stream from a user; dynamically constructing a user interest model according to the real-time browsing behavior data flow, wherein the user interest model is used for reflecting the interest preference of a user; performing real-time matching with commodity information in a commodity database based on the user interest model to find a commodity set matched with the interest preference; generating a personalized data push packet according to the commodity information in the commodity set in combination with a preset data push strategy; and the personalized data push packet is pushed to terminal equipment of a user in real time through a network, so that the user can receive commodity information push conforming to the current interest preference. According to the technical scheme provided by the invention, the method not only improves the operation efficiency of the e-commerce platform, but also provides a more intelligent and personalized shopping experience environment for the user.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of e-commerce data push based on big data, and in particular, to a method and system for e-commerce data push based on big data. Background Art

[0002] With the development of Internet technology, e-commerce platforms have become an indispensable part of people's daily lives. In this context, how to efficiently connect user needs with product information has become one of the important challenges facing e-commerce platforms. Specifically, in the application scenarios, technical requirements are mainly reflected in the following aspects:

[0003] E-commerce platforms need to be able to receive users' browsing behavior data streams in real time, and dynamically build user interest models based on this, so as to accurately reflect users' interest preferences. In addition, it is necessary to match this model with the information in the product database in real time to find a set of products that match the user's interests, and then generate personalized data push packages based on this information, and instantly push them to the user's terminal device through the network.

[0004] Most e-commerce platforms currently use a recommendation system based on historical data, which analyzes users' past behavior data (such as purchase records, search history, etc.) to infer the products they may be interested in. Although this method has achieved personalized recommendations to a certain extent, it is insufficient in terms of real-time and accuracy. In addition, traditional push strategies often ignore the user's current specific context (such as geographic location, time, etc.), resulting in the push content sometimes not meeting the user's current needs.

[0005] Traditional recommendation algorithms, due to their reliance on historical data, perform poorly in dealing with rapid changes in user interests, which can easily lead to lagging or irrelevant recommendations. In addition, the lack of effective processing methods for real-time data streams makes the recommendation system respond slowly, making it difficult to meet the needs of users for instant feedback in the modern e-commerce environment. Furthermore, the push strategy is not flexible enough to respond quickly to user behavior in different scenarios, affecting the pertinence and effectiveness of the push content. Summary of the invention

[0006] The embodiments of the present application provide an e-commerce data push method and system based on big data, so as to solve the problem that the prior art performs poorly in coping with the rapid changes in user interests.

[0007] In a first aspect, an embodiment of the present application provides an e-commerce data push method based on big data, comprising:

[0008] Receive a real-time browsing behavior data stream from a user; dynamically build a user interest model based on the real-time browsing behavior data stream, wherein the user interest model is used to reflect the user's interest preferences; perform real-time matching based on the user interest model with the product information in the product database to discover a product set that matches the interest preferences; generate a personalized data push package based on the product information in the product set and in combination with a preset data push strategy; and instantly push the personalized data push package to the user's terminal device via the network so that the user can receive product information push that matches the current interest preferences.

[0009] Optionally, the dynamically constructing the interest preference of the user interest model according to the real-time browsing behavior data stream includes:

[0010] The user's behavior data is collected in real time through the monitoring component on the user's terminal device to form a user behavior sequence, wherein the behavior data includes web browsing records, click behaviors, search queries, and page dwell time; the behavior data is analyzed and processed to extract user behavior characteristics, wherein the user's behavior characteristics include the user's attention to a specific type of content, browsing frequency, and interaction depth; based on the extracted behavior characteristics, a user interest model is constructed.

[0011] Optionally, the real-time matching based on the user interest model with product information in a product database to find a product set matching the interest preference includes:

[0012] Quickly retrieve product information associated with the keywords, category preferences, and brand tendencies in the user interest model in a product database; evaluate the relevance score of each product information in the product database based on the user's behavior data and the user interest model; implement a multi-dimensional matching strategy to match the product information with the interest preferences to obtain a multi-dimensional matching result; generate a candidate product set based on the relevance score of the product information and the multi-dimensional matching result, and sort the candidate product set to give priority to displaying the product set that best matches the interest preference.

[0013] Optionally, generating a personalized data push package based on the product information in the product set in combination with a preset data push strategy includes:

[0014] Extract key product features based on product information in the product set, the key product features include product category, price range, brand and promotion information; set a push rule set based on a preset data push strategy, the push rule set includes a push time window, user active period, and push frequency limit; use a machine learning algorithm to calculate the push weight of each product based on the key product features and the push rule set; customize the style of the push message in combination with the user's device type and operating system version information; design the content of the push message, and generate a corresponding personalized data push package based on the push weight of each product, the style of the push message, the content of the push message and the user's preference settings.

[0015] Optionally, the step of instantly pushing the personalized data push package to the user's terminal device via a network includes:

[0016] The optimal data transmission path is selected according to the user's geographic location information and network conditions; the personalized data push package is encrypted using encryption technology, and the encrypted personalized data push package is sent from the server to the user's terminal device based on the optimal data transmission path.

[0017] Optionally, evaluating the relevance score of each commodity information in the commodity database according to the user's behavior data and the user interest model includes:

[0018] The relevance score of each product information in the product database is evaluated by the following calculation formula:

[0019]

[0020] Among them, rel i represents the relevance score of product information i, H u represents the behavior data of user u, M u (t) represents the interest model of user u at time t, and represents the user's interest preference through a vector or matrix, P i Represents the attribute vector of product information i, sim(H u ,P i ) represents the behavior data H of user u u and the attribute vector P of product information i i The similarity function, sim(M u (t),P i ) represents the interest model M of user u at time t u (t) and the attribute vector P of product information i iSimilarity function, α represents the weight factor, which is used to balance the importance of behavior data, β represents the weight factor, which is used to balance the importance of the current interest model, and γ represents an adjustable parameter, which is used to control the influence of time decay. represents the time when user u last interacted with product information i, T represents the current time point, Represents an exponential decay function, which is used to measure the distance between the time when user u last interacted with product information i and the current time point T. λ is a positive number that determines the decay rate.

[0021] Optionally, the real-time matching based on the user interest model with product information in a product database to find a product set matching the interest preference further includes:

[0022] When real-time matching is performed based on the user interest model and the product information in the product database, a comprehensive scoring function CSF (p k ,t), used to evaluate product p k Matching degree with user’s interest preferences:

[0023] CSF(p k ,t)=α(t)·RF(p k ,t)+β(t)·CF(p k ,t)+γ(t)·MD(p k ,t)+λ·DIF(p k ,t)+ξ·DLF(p k ,t)+ρRLW(p k ,t)+ψ·RNN(p k ,t)+v·LSTM(p k ,t)

[0024] Among them, p k represents the attribute vector of product k; t represents the current time point; α(t) represents the time-related weight factor, which is used to adjust the importance of the score based on the recommendation framework; RF(p k ,t): score based on the recommendation framework; β(t) represents the time-related weight factor, which is used to adjust the importance of the collaborative filtering score; CF(p k ,t) represents the collaborative filtering score, which is calculated based on the similarity of user behaviors; γ(t): represents the time-related weight factor, which is used to adjust the importance of the multi-dimensional matching score; MD(p k ,t): The score of multi-dimensional matching, which is the comprehensive matching score based on the multi-dimensional matching results; λ represents a constant factor used to adjust the difference factor; DIF(p k,t) represents the score of the difference factor, which refers to the degree of difference between the product and the behavior data; ξ represents the constant factor, which is used to adjust the importance of the attenuation factor; DLF(p k ,t) represents the decay factor score, which is used to measure the time when the user last interacted with the product; ρ represents the constant factor, which is used to adjust the importance of rule-based learning weights; RLW(p k ,t): score based on rule-based learning weights, which may be a score calculated by some predefined rules. ψ represents a constant factor used to adjust the importance of neural network-based scoring; RNN(p k ,t) represents the score based on the neural network, which may be the product score predicted by training the neural network; ν represents the constant factor, which is used to adjust the importance of the score based on the long short-term memory network; LSTM (p k ,t) represents the rating based on the long short-term memory network, which is the product score predicted by the trained long short-term memory network.

[0025] In a second aspect, an embodiment of the present application provides an e-commerce data push system based on big data, including:

[0026] A receiving module is used to receive a real-time browsing behavior data stream from a user; a building module is used to dynamically build a user interest model based on the real-time browsing behavior data stream, and the user interest model is used to reflect the user's interest preferences; a matching module is used to perform real-time matching based on the user interest model with the product information in the product database to find a product set that matches the interest preferences; a generating module is used to generate a personalized data push package based on the product information in the product set and a preset data push strategy; a pushing module is used to instantly push the personalized data push package to the user's terminal device via the network, so that the user can receive product information push that meets the current interest preferences.

[0027] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an e-commerce data push method based on big data as described in the first aspect.

[0028] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements an e-commerce data push method based on big data as described in the first aspect.

[0029] In an embodiment of the present application, a real-time browsing behavior data stream from a user is received; a user interest model is dynamically constructed based on the real-time browsing behavior data stream, and the user interest model is used to reflect the user's interest preferences; real-time matching is performed based on the user interest model and the product information in the product database to find a product set that matches the interest preferences; a personalized data push package is generated based on the product information in the product set and combined with a preset data push strategy; the personalized data push package is instantly pushed to the user's terminal device via the network, so that the user can receive product information push that meets the current interest preferences. The technical solution and method provided by the present application not only improve the operational efficiency of the e-commerce platform, but also provide users with a more intelligent and personalized shopping experience environment.

[0030] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 A flowchart of an e-commerce data push method based on big data provided in an embodiment of the present application;

[0033] Figure 2 A schematic diagram of the structure of an e-commerce data push system based on big data provided in an embodiment of the present application;

[0034] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0036] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0037] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0038] Figure 1 A flowchart of an e-commerce data push method based on big data is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0039] 101. Receive real-time browsing behavior data stream from users;

[0040] This data usually includes but is not limited to user click events, scrolling page actions, search queries, visits to product detail pages, adding to shopping carts, purchase behaviors, etc. These real-time data provide direct evidence of users' current activities and help the platform understand users' immediate interests and needs.

[0041] Assume that an e-commerce platform has a front-end application that users access through mobile devices. When users scroll through the product list, click to enter a product detail page, and search for a specific product in the application, the front-end application will record these behaviors and send these data to the back-end server in real time through the API (application programming interface):

[0042] JavaScript code or other front-end technologies are embedded in the e-commerce application on the user's terminal device to monitor every operation event of the user. For example, when a user clicks on a product image, the front-end triggers an event and records the click behavior; whenever a user operation event is detected, the front-end application packages the data of the event into a message in JSON format and sends it to the back-end server through an HTTP / HTTPS request. The data packet usually contains information such as the event type (such as click), event timestamp, user ID, device information (such as device type, operating system), and the ID of the relevant product; after receiving the data sent by the front-end, the back-end server stores it in the database and triggers the corresponding business logic, such as updating the user's browsing history, adjusting the user's interest model, etc.; in order to achieve true real-time performance, the e-commerce platform may use stream processing frameworks (such as Apache Kafka, Apache Flink, etc.) to process these data streams to ensure that the data can be processed quickly and used in subsequent recommendation systems.

[0043] In this way, e-commerce platforms can continuously obtain the latest user behavior data and make real-time responses based on this data, such as immediately updating recommendation lists to provide product information that is more in line with the user's current interests.

[0044] 102. Dynamically construct a user interest model based on the real-time browsing behavior data stream, wherein the user interest model is used to reflect the user's interest preferences;

[0045] The core of building a user interest model is to extract useful features from a large amount of user behavior data and use these features to describe the user's interest preferences. Common features include user attention to a specific category of goods, browsing frequency, click-through rate, etc. Model construction usually involves the application of data mining technology and machine learning algorithms to update user interest preferences in real time, thereby providing more accurate product recommendations.

[0046] Suppose there is such a scenario: user Alice is using an application of an e-commerce platform. She has frequently browsed sports shoes in the past hour and is particularly interested in a certain brand of sports shoes. Now let's see how to dynamically build her interest model based on her behavior data:

[0047] First, the front-end application has collected Alice's browsing behavior data in real time, including information such as the types of goods she has viewed, brands, and price ranges; the back-end system performs preliminary cleaning on the received data to remove invalid or duplicate data entries; extracts useful features from the behavior data, such as the category of goods Alice has recently browsed (sports shoes), brand (a certain brand), browsing frequency (multiple browsing within an hour), page dwell time (average long dwell time per page), etc.; uses machine learning algorithms (such as collaborative filtering, content-based recommendation, etc.) to build Alice's interest model. For example, based on her recent browsing behavior, the model may conclude that Alice is particularly interested in sports shoes, especially sports shoes of a certain brand; as Alice continues to browse other goods or returns to view sports shoes again, the back-end system will continue to update her behavior data and adjust her interest model accordingly. If she starts browsing other types of shoes, such as casual shoes, her interest model will gradually reflect this; once Alice's interest model is updated, the system can immediately apply it to the product recommendation algorithm. For example, when Alice opens the application again, the homepage will give priority to displaying sports shoes and other types of shoes that match her interest model to improve her shopping experience. Through the above steps, the e-commerce platform can dynamically adjust and optimize the user interest model based on the user's real-time behavior data, thereby providing more personalized and user-friendly product recommendation services. This not only improves user satisfaction, but also increases the platform's conversion rate.

[0048] This application takes into account that in existing e-commerce platforms, although the user interest model constructed through historical data can provide a certain degree of personalized recommendations, since the user's interest preferences will change over time, especially when the user's behavior pattern changes significantly in the short term, the model based on historical data may lag behind the user's actual interests. In addition, the existing data collection and processing methods may have certain limitations, such as incomplete data collection, inaccurate feature extraction, etc., resulting in inaccurate recommendation results. Therefore, an embodiment of the present invention proposes an optional solution, which further refines the user behavior characteristics through in-depth analysis and processing of real-time browsing behavior data, so as to solve the above technical problems and improve the real-time and accuracy of the recommendation system.

[0049] The options are as follows:

[0050] Optionally, the “dynamically constructing the interest preference of the user interest model according to the real-time browsing behavior data stream” in step 102 includes:

[0051] The monitoring component on the user terminal device collects the user's behavior data in real time to form a user behavior sequence, wherein the behavior data includes web browsing records, click behaviors, search queries, and page dwell time; the behavior data is analyzed and processed to extract user behavior features, wherein the user's attention to a specific type of content, browsing frequency, and interaction depth are included; based on the extracted behavior features, a user interest model is constructed; assuming that user Alice browses a variety of sports shoes on an e-commerce platform continuously and has a strong interest in a certain brand of sports shoes; Alice browses 5 different models of sports shoes in half an hour, 3 of which belong to the same brand, and the time spent on these 3 product pages is The time she spends on a certain style is significantly longer than other styles; the monitoring component on the terminal device records her every click, page dwell time and other behaviors, and forms a user behavior sequence; analysis and processing show that Alice is very interested in sports shoes, especially sports shoes of a certain brand; the extracted behavioral features include: sports shoe category (high attention), a certain brand (high attention), and long page dwell time (high interaction depth); using the extracted behavioral features, the system constructs Alice's interest model, which reflects her high interest in sports shoes, especially sports shoes of a certain brand; after the model is built, the system recommends more products that match her interests to Alice based on the model, and gives priority to displaying these products when she visits next time. Through the above steps, the system can more accurately capture the user's immediate interest changes and adjust the recommended content in a timely manner, thereby providing a more personalized shopping experience.

[0052] This approach not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment. In particular, when dealing with short-term changes in user behavior, it can respond faster and improve the accuracy and real-time nature of recommendations.

[0053] 103. Performing real-time matching based on the user interest model and product information in a product database to find a product set matching the interest preference;

[0054] This step means that after building the user interest model, the system needs to compare the model with the product information in the product database in real time to find a set of products that match the user's current interest preferences.

[0055] Assume that user Bob has frequently browsed outdoor sports equipment in the past period of time, and has paid special attention to a certain brand of tents. Now we need to match suitable products according to Bob's interest model: the system extracts the features that Bob is interested in according to his interest model, such as "outdoor sports equipment", "tent", and "a certain brand"; in the product database, the system quickly retrieves product information that matches Bob's interest features. For example, all products marked as "outdoor sports equipment", especially those tents with the brand "a certain brand"; the system calculates the relevance score of each product based on Bob's historical behavior data (such as his recent browsing history, number of clicks, etc.) and interest model. For example, a certain brand of tent that Bob has frequently browsed recently will get a higher relevance score; the system scores the relevance of each product; the system comprehensively considers multiple dimensions such as user behavior and time decay, matches product information with user interest preferences, and generates a candidate product set; the system sorts the candidate product set and prioritizes those products that best match Bob's interest preferences; finally, the system generates a set of products that highly match Bob's interest preferences and sorts them by relevance scores for subsequent personalized data push. Through the above steps, the system can discover and recommend product sets that match the user's current interests and preferences in real time, thereby improving user satisfaction and shopping experience. This method not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment.

[0056] The present application takes into account that in the existing e-commerce recommendation system, although the product recommendation based on the user interest model has achieved certain results, there are still some problems. For example, the traditional recommendation algorithm may only rely on the user's historical behavior data, ignoring the user's real-time behavior and interest changes, resulting in the lack of timeliness and accuracy of the recommended content. In addition, a single-dimensional matching strategy may also cause the recommendation results to be not comprehensive and accurate enough. Therefore, an embodiment of the present invention proposes an optional solution, which solves the above-mentioned technical problems by performing multi-dimensional matching of product information and conducting a comprehensive evaluation in combination with the user's real-time behavior data, thereby improving the real-time response capability of the recommendation system and the accuracy of the recommendation results.

[0057] The options are as follows:

[0058] Optionally, the “matching in real time based on the user interest model with product information in a product database to find a product set matching the interest preference” in step 103 includes:

[0059] Quickly retrieve product information associated with the keywords, category preferences, and brand tendencies in the user interest model in a product database; evaluate the relevance score of each product information in the product database based on the user's behavior data and the user interest model; implement a multi-dimensional matching strategy to match the product information with the interest preferences to obtain a multi-dimensional matching result; generate a candidate product set based on the relevance score of the product information and the multi-dimensional matching result, and sort the candidate product set to give priority to displaying the product set that best matches the interest preference.

[0060] Assume that user Bob has frequently browsed outdoor sports equipment in the past period of time, and is particularly interested in a certain brand of tents. Now we need to match suitable products based on Bob's interest model. Quickly retrieve product information associated with the keywords, category preferences, and brand preferences in the user's interest model in the product database: Bob's interest model contains keywords such as "outdoor sports equipment", "tents", and "certain brand"; the product database quickly retrieves product information related to these keywords through indexing technology. Based on the user's behavior data and the user's interest model, evaluate the relevance score of each product information in the product database:

[0061] Optionally, evaluating the relevance score of each commodity information in the commodity database according to the user's behavior data and the user interest model includes:

[0062] The relevance score of each product information in the product database is evaluated by the following calculation formula:

[0063]

[0064] Among them, rel i represents the relevance score of product information i, H u represents the behavior data of user u, M u (t) represents the interest model of user u at time t, and represents the user's interest preference through a vector or matrix, P i Represents the attribute vector of product information i, sim(H u ,P i ) represents the behavior data H of user u u and the attribute vector P of product information i i The similarity function, sim(M u (t),P i ) represents the interest model M of user u at time t u (t) and the attribute vector P of product information i iSimilarity function, α represents the weight factor, which is used to balance the importance of behavior data, β represents the weight factor, which is used to balance the importance of the current interest model, and γ represents an adjustable parameter, which is used to control the influence of time decay. represents the time when user u last interacted with product information i, T represents the current time point, Represents an exponential decay function, which is used to measure the distance between the time when user u last interacted with product information i and the current time point T. λ is a positive number that determines the decay rate.

[0065] Assume that Bob has frequently browsed the brand's account in the past week and stayed on these pages for a long time. Among them, H u represents Bob’s behavior data, M u (t) represents Bob’s interest model at time t, P i represents the attribute vector of product i, sim represents the similarity function, α, β and γ are weight factors, T is the current time point, is the time when Bob last interacted with item i, and λ is the decay rate.

[0066] Assuming α = 0.4, β = 0.4, γ = 0.2, λ = 0.1, the relevance score of a certain brand of tent is 0.4 0.9 + 0.4 0.8 + 0.2 e -0.1·(7×24) =0.36+0.32+0.2·0.779≈0.74; implement a multi-dimensional matching strategy to match the product information with the interest preferences, and obtain a multi-dimensional matching result: considering that Bob's recent active period is between 8 and 10 pm, the system will give a higher weight to the behavior during this period; combined with the time decay factor, products that are frequently browsed in the past week receive higher scores. Based on the relevance score and multi-dimensional matching results, the system generates a set of candidate products and sorts them by score. In the end, the system prioritizes the product set that best matches Bob's interest preferences, such as a certain brand of tents ranked first.

[0067] Through the above steps, the system can discover and recommend product sets that match the user's current interests and preferences in real time, thereby improving user satisfaction and shopping experience. This method not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment.

[0068] The present application takes into account that in existing e-commerce recommendation systems, although the relevance scores of commodities can be evaluated through user behavior data and interest models, a single evaluation method may not be able to fully cover the complex needs of users. Existing recommendation algorithms may fail to fully utilize the advantages of multiple recommendation technologies, resulting in incomplete and inaccurate recommendation results. Therefore, an embodiment of the present invention proposes an optional solution, which introduces a comprehensive scoring function (CSF) to comprehensively evaluate the matching degree between commodities and user interest preferences, so as to solve the problem of incomplete and inaccurate recommendation results in the prior art, and improve the accuracy of the recommendation system and user satisfaction.

[0069] The options are as follows:

[0070] Optionally, the real-time matching based on the user interest model with product information in a product database to find a product set matching the interest preference further includes:

[0071] When real-time matching is performed based on the user interest model and the product information in the product database, a comprehensive scoring function CSF (p k ,t), used to evaluate product p k Matching degree with user’s interest preferences:

[0072] CSF(p k ,t)=α(t6)·RF(p k ,t)+β(t)·CF(p k ,t)+γ(t)·MD(p k ,t)+λ·DIF(p k ,t)+ξ·DLF(p k ,t)+ρRLW(p k ,t)+ψ·RNN(p k ,t)+v·LSTM(p k ,t)

[0073] Among them, p k represents the attribute vector of product k; t represents the current time point; α(t) represents the time-related weight factor, which is used to adjust the importance of the score based on the recommendation framework; RF(p k ,t): score based on the recommendation framework; β(t) represents the time-related weight factor, which is used to adjust the importance of the collaborative filtering score; CF(p k ,t) represents the collaborative filtering score, which is calculated based on the similarity of user behaviors; γ(t): represents the time-related weight factor, which is used to adjust the importance of the multi-dimensional matching score; MD(p k,t): The score of multi-dimensional matching, which is the comprehensive matching score based on the multi-dimensional matching results; λ represents a constant factor used to adjust the difference factor; DIF(p k ,t) represents the score of the difference factor, which refers to the degree of difference between the product and the behavior data; ξ represents the constant factor, which is used to adjust the importance of the attenuation factor; DLF(p k ,t) represents the decay factor score, which is used to measure the time when the user last interacted with the product; ρ represents the constant factor, which is used to adjust the importance of rule-based learning weights; RLW(p k ,t): score based on rule-based learning weights, which may be a score calculated by some predefined rules. ψ represents a constant factor used to adjust the importance of neural network-based scoring; RNN(p k ,t) represents the score based on the neural network, which may be the product score predicted by training the neural network; v represents the constant factor, which is used to adjust the importance of the score based on the long short-term memory network; LSTM (p k ,t) represents the rating based on the long short-term memory network, which is the product score predicted by the trained long short-term memory network.

[0074] Assume that user Bob has frequently browsed outdoor sports equipment in the past period of time, and is particularly interested in a certain brand of tents. Now we need to match the appropriate products based on Bob's interest model:

[0075] Define the comprehensive scoring function CSF(p k ,t), to evaluate the product p k The matching degree with the user’s interest preference at time t. Assume that CSF(p k ,t)=α(t)·RF(p k ,t)+β(t)·CF(p k ,t)+γ(t)·MD(p k ,t)+λ·DIF(p k ,t)+ξ·DLF(p k ,t)+ρ·RLW(p k ,t)+ψ·RNN(p k ,t)+v·LSTM(p k ,t). Calculate product p through recommendation algorithm k Score at time t. Assume RF(p k ,t)=0.8. Calculate product p based on the similarity of user behavior k Assume that CF(p k ,t)=0.7; Comprehensively evaluate product p through multi-dimensional matching results k Assuming MD(p k,t)=0.9; measure product p k The degree of difference between the DIF(p k ,t)=0.3; measures the user’s most recent interaction with product p k The distance between the interaction time and the current time point; assuming that DLF (p k ,t)=0.6; calculate product p through predefined rules k Assume that RLW(p k ,t)=0.5; predict product p through the trained neural network k Score. Assuming RNN(p k ,t)=0.75; predict product p through the trained long short-term memory network k Assume that LSTM(p k ,t)=0.85; assuming that the weight factors are: α(t)=0.2, β(t)=0.15, γ(t)=0.2, λ=0.1, ξ=0.1, ρ=0.05, ψ=0.1, v=0.1.

[0076] Then the comprehensive scoring function CSF(p k ,t) is calculated as:

[0077] CSF(p k ,t)=0.2·0.8+0.15·0.7+0.2·0.9+0.1·0.3+0.1·0.6+0.05·0.5+0.1·0. 75+0.1·0.85=0.16+0.105+0.18+0.03+0.06+0.025+0.075+0.085=0.72

[0078] Through the above steps, the system can comprehensively evaluate the match between products and user interests and preferences, thereby improving the comprehensiveness and accuracy of recommendation results.

[0079] This method not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment. By introducing a comprehensive scoring function, the recommendation system can more comprehensively consider the results of multiple scoring methods, improve the accuracy and real-time nature of the recommendation, and thus further improve user satisfaction.

[0080] 104. Generate a personalized data push package based on the commodity information in the commodity set and in combination with a preset data push strategy;

[0081] This step means that after finding a set of products that match the user's interest preferences, the system needs to select appropriate content according to the preset data push strategy and generate a personalized push package so that the personalized content can be pushed to the user through the user's terminal device.

[0082] Assume that user Carol has recently browsed and purchased baby products frequently, and has shown a special interest in a certain brand of diapers. Now we need to generate a personalized data push package based on Carol's interest model:

[0083] The system extracts key features from a set of products that match Carol's interests, such as "baby products," "diapers," and "a certain brand." Push time window: Based on Carol's active time periods, choose to push during the time periods when she is most likely to check her phone; suppose data analysis shows that Carol is most active between 8 and 10 p.m.; To avoid overly disturbing users, set a push every two days; The system uses machine learning algorithms (such as logistic regression, decision trees, etc.) to calculate the push weight for each product. Assume that diapers of "a certain brand" are given a higher weight because they are highly correlated with user historical behavior; Choose the most suitable push style based on the type of device Carol uses (such as iOS or Android devices) and the operating system version. For example, iOS devices may support a richer range of push notification styles; Design the content of the push message to ensure that it is both attractive and not too intrusive. For example, a push message can include a short greeting, a product highlight introduction, and an obvious "View Details" button; Integrate all of the above information to generate a personalized data push package. This push package will contain selected product information, a suitable time window, an optimized push style, and carefully designed push content.

[0084] Through the above steps, the system finally generates a personalized data push package for user Carol. This push package will be sent to her mobile phone between 8pm and 10pm to increase the possibility of her viewing the push information and increase the probability of her interacting with the push content. This method not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment.

[0085] This application takes into account that although existing e-commerce platforms can recommend products based on user interests and preferences when generating personalized data push packages, there is still room for improvement in the refined management of push strategies. For example, existing push strategies may not fully consider the impact of factors such as user device type and operating system version on push effects, nor do they fully utilize machine learning algorithms to optimize push weights. Therefore, an embodiment of the present invention proposes an optional solution to solve the problems of insufficient refinement of push strategies and unsatisfactory push effects in the prior art by extracting key product features, setting a push rule set, calculating product push weights, and customizing push message styles and content, thereby improving the relevance of push content and user acceptance.

[0086] The options are as follows:

[0087] Optionally, the step 104 of “generating a personalized data push package based on the product information in the product set and in combination with a preset data push strategy” includes:

[0088] Extract key product features based on product information in the product set, the key product features include product category, price range, brand and promotion information; set a push rule set based on a preset data push strategy, the push rule set includes a push time window, user active period, and push frequency limit; calculate the push weight of each product based on the key product features and the push rule set using a machine learning algorithm; customize the style of the push message in combination with the user's device type and operating system version information; design the content of the push message, and generate a corresponding personalized data push package based on the push weight of each product, the style of the push message, the content of the push message and the user's preference settings.

[0089] Suppose user Carol has recently browsed and purchased baby products frequently, and has shown a particular interest in a certain brand of diapers. Now we need to generate a personalized data push package based on Carol's interest model; extract key features from the set of products that match Carol's interest preferences, such as "baby products", "diapers", "a certain brand", and "promotional information"; based on Carol's active time period, choose to push during the time period when she is most likely to check her phone. Suppose data analysis shows that Carol is most active between 8 and 10 pm; to avoid excessively disturbing users, set a push every two days; use machine learning algorithms (such as logistic regression, decision trees, etc.) to calculate the push weight of each product. Suppose "a certain brand" of diapers has a higher weight because it is highly correlated with the user's historical behavior; select the most suitable push message style based on the type of device Carol uses (such as iOS device) and the operating system version. For example, iOS devices may support a richer push notification style; design the content of the push message to ensure that it is both attractive and not too intrusive. For example, a push message can include a brief greeting, a product highlight introduction, and an obvious "View Details" button; integrating all of the above information can generate a personalized data push package, which will include selected product information, a suitable time window, an optimized push style, and carefully designed push content.

[0090] Through the above steps, the system can generate more personalized and accurate data push packages based on the user's interest preferences and device characteristics. This approach not only improves user acceptance and interaction probability, but also improves the operational efficiency of the e-commerce platform, providing users with a more intelligent and personalized shopping experience environment.

[0091] 105. Push the personalized data push package to the user's terminal device via the network in real time, so that the user can receive product information push that meets the current interest preference.

[0092] This step refers to sending personalized data push packages to the user's terminal device via the network immediately after the personalized data push packages are generated, ensuring that the user can receive product information that matches his or her current interests and preferences in a timely manner.

[0093] Assume that user David often browses electronic products at night and is very interested in a new smart watch recently. Now we need to send a personalized push package related to this smart watch to David immediately; the system selects an optimal data transmission path based on David's geographic location information (for example, knowing that he is at home through GPS positioning) and the current network conditions (such as good Wi-Fi connection quality); before sending, the system uses an encryption algorithm (such as AES or RSA) to encrypt the push package to ensure that even if the data is intercepted during transmission, it cannot be interpreted by a third party; the encrypted personalized data push package is sent from the server to David's smartphone through the optimal data transmission path. Assuming that David is using an iOS device, the push package will be transmitted through Apple's push service (APNs); when the push package arrives at David's smartphone, the push service client on the device decrypts the push package and displays the push notification according to the preset push style. For example, the push notification may be displayed at the top of the screen, containing a picture of the smart watch, a brief description, and a "View Now" button.

[0094] Through the above steps, the system successfully sends a personalized push package containing smart watch information to David in real time. This not only ensures that David receives product information that matches his interests and preferences at the best time, but also enhances his trust and satisfaction with the platform because he knows that his privacy is properly protected; this method not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment.

[0095] The present application takes into account that, in the existing e-commerce platform, although the function of personalized data push has been realized, there are still some problems in the process of data transmission, such as transmission delay, insufficient security, etc. In particular, when users are in different geographical locations and network conditions, the immediacy and security of the push package may be affected. Therefore, the embodiment of the present invention proposes an optional solution, which selects the optimal data transmission path according to the user's geographical location information and network conditions, and uses encryption technology to encrypt the push package, so as to solve the problems of data transmission delay and insufficient security in the prior art, and improve the immediacy of push and the security of user data.

[0096] The options are as follows:

[0097] Optionally, the “instantly pushing the personalized data push package to the user's terminal device via the network” in step 105 includes:

[0098] According to the user's geographic location information and network conditions, select the optimal data transmission path; use encryption technology to encrypt the personalized data push package, and send the encrypted personalized data push package from the server to the user's terminal device based on the optimal data transmission path; assume that user David often browses electronic products at night and is very interested in a new smart watch recently. Now we need to send the personalized push package related to this smart watch to David immediately; know David's location at home through GPS positioning; detect that David is currently using a Wi-Fi connection with good signal quality; select a path that can ensure fast and stable data transmission, such as connecting through a home Wi-Fi network; use the AES algorithm to encrypt the personalized data push package; include a picture of the smart watch, a brief description and a "view now" button; send the encrypted data packet from the server to David's smartphone through the selected home Wi-Fi network path.

[0099] Through the above steps, the system successfully sends a personalized push package containing smartwatch information to David in real time. This not only ensures that David receives product information that matches his interests and preferences at the best time, but also enhances his trust and satisfaction with the platform because he knows that his privacy is properly protected. This approach not only improves the operational efficiency of the e-commerce platform, but also provides users with a more intelligent and personalized shopping experience environment.

[0100] Figure 2 A structural diagram of an e-commerce data push system based on big data is provided for the embodiment of the present application, such as Figure 2 As shown, the device comprises:

[0101] Receiving module 21, used for receiving real-time browsing behavior data stream from users;

[0102] A construction module 22, used to dynamically construct a user interest model based on the real-time browsing behavior data stream, wherein the user interest model is used to reflect the user's interest preferences;

[0103] A matching module 23, configured to perform real-time matching based on the user interest model and the product information in the product database to find a product set matching the interest preference;

[0104] A generating module 24, configured to generate a personalized data push package based on the commodity information in the commodity set and in combination with a preset data push strategy;

[0105] The push module 25 is used to push the personalized data push package to the user's terminal device via the network in real time, so that the user can receive the product information push that meets the current interest preference.

[0106] Figure 2 The e-commerce data push system based on big data can execute Figure 1 The implementation principle and technical effect of the e-commerce data push method based on big data described in the embodiment shown are not repeated here. The specific way in which each module and unit performs operations in the e-commerce data push system based on big data in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0107] In one possible design, Figure 2 The e-commerce data push system based on big data of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0108] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0109] The processing component 32 is used to: receive a real-time browsing behavior data stream from a user; dynamically construct a user interest model based on the real-time browsing behavior data stream, wherein the user interest model is used to reflect the user's interest preferences; perform real-time matching based on the user interest model with the product information in the product database to discover a product set that matches the interest preferences; generate a personalized data push package based on the product information in the product set and in combination with a preset data push strategy; and instantly push the personalized data push package to the user's terminal device via the network so that the user can receive product information push that matches the current interest preferences.

[0110] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0111] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0112] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0113] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0114] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0115] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0116] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1The illustrated embodiment is an e-commerce data push method based on big data.

[0117] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0118] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0119] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for pushing e-commerce data based on big data, characterized in that: include: Receive real-time browsing behavior data streams from users; Dynamically constructing a user interest model based on the real-time browsing behavior data stream, wherein the user interest model is used to reflect the user's interest preferences; Performing real-time matching based on the user interest model with product information in a product database to discover a product set that matches the user's interest preferences; Generate a personalized data push package based on the product information in the product set and in combination with a preset data push strategy; The personalized data push package is instantly pushed to the user's terminal device via the network, so that the user can receive product information push that meets the current interest preference.

2. The method according to claim 1, characterized in that The dynamically constructing the user interest model based on the real-time browsing behavior data stream includes: The monitoring component on the user terminal device collects user behavior data in real time to form a user behavior sequence. The behavior data includes web browsing history, click behavior, search queries, and page dwell time; Analyzing and processing the behavior data to extract user behavior characteristics, wherein the user behavior characteristics include the user's attention level, browsing frequency, and interaction depth for a specific type of content; Based on the extracted behavioral features, a user interest model is constructed.

3. The method according to claim 1, characterized in that The real-time matching of the user interest model with the product information in the product database to find a product set matching the interest preference includes: Quickly retrieve product information associated with the keywords, category preferences, and brand tendencies in the user interest model from a product database; Evaluate the relevance score of each product information in the product database based on the user's behavior data and the user interest model; Implementing a multi-dimensional matching strategy to match the product information with the interest preferences to obtain a multi-dimensional matching result; Based on the relevance score of the product information and the multi-dimensional matching result, a candidate product set is generated, and the candidate product set is sorted, with the product set that best matches the interest preference being displayed first.

4. The method according to claim 1, wherein The generating of a personalized data push package based on the product information in the product set and in combination with a preset data push strategy includes: Extracting key product features based on product information in the product set, wherein the key product features include product category, price range, brand, and promotion information; According to the preset data push strategy, set the push rule set, which includes the push time window, user active period, and push frequency limit; Utilizing a machine learning algorithm, based on the key product features and the push rule set, calculate the push weight of each product; Customize the push message style based on the user's device type and operating system version information; The content of the push message is designed, and a corresponding personalized data push package is generated according to the push weight of each product, the style of the push message, the content of the push message and the user's preference settings.

5. The method according to claim 1, characterized in that The step of instantly pushing the personalized data push package to the user's terminal device via a network includes: Selecting an optimal data transmission path based on the user's geographic location information and network conditions; The personalized data push package is encrypted using encryption technology, and the encrypted personalized data push package is sent from the server to the user's terminal device based on the optimal data transmission path.

6. The method according to claim 3, characterized in that The step of evaluating the relevance score of each product information in the product database based on the user's behavior data and the user interest model includes: The relevance score of each product information in the product database is evaluated by the following calculation formula: Among them, rel i represents the relevance score of product information i, H u represents the behavior data of user u, M u (t) represents the interest model of user u at time t, and expresses the user's interest preference through a vector or matrix, P i Represents the attribute vector of product information i, sim(H u , P i ) represents the behavior data H of user u u and the attribute vector P of product information i i The similarity function, sim(M u (t), P i ) represents the interest model M of user u at time t u (t) and the attribute vector P of product information i i Similarity function, α represents the weight factor, which is used to balance the importance of behavior data, β represents the weight factor, which is used to balance the importance of the current interest model, and γ represents the adjustable parameter, which is used to control the influence of time decay. represents the time when user u last interacted with product information i, T represents the current time point, Represents an exponential decay function, which is used to measure the distance between the time of the most recent interaction between user u and product information i and the current time point T. λ is a positive number that determines the decay rate.

7. The method according to claim 3, characterized in that The real-time matching of the user interest model with product information in a product database to find a product set that matches the user interest preference further includes: When the user interest model is matched with the product information in the product database in real time, a comprehensive scoring function CSF (p k , t), used to evaluate product p k Matching degree with user’s interests and preferences: CSF(p k ,t)=α(t)·RF(p k ,t)+β(t)·CF(p k ,t)+γ(t)·MD(p k ,t)+λ·DIF(p k ,t)+ξ·DLF(p k ,t)+ρ RLW(p k ,t)+ψ·RNN(p k ,t)+v·LSTM(p k ,t) Among them, p k represents the attribute vector of product k; t represents the current time point; α(t) represents the time-related weight factor, which is used to adjust the importance of the score based on the recommendation framework; RF(p k , t): score based on the recommendation framework; β(t) represents the time-related weight factor used to adjust the importance of collaborative filtering scores; CF(p k , t) represents the collaborative filtering score, which is calculated based on the similarity of user behavior; γ(t): represents the time-related weight factor, which is used to adjust the importance of the multi-dimensional matching score; MD(p k , t): The score of multi-dimensional matching is the comprehensive matching score based on the multi-dimensional matching results; λ represents a constant factor used to adjust the difference factor; DIF(p k , t) represents the score of the difference factor, which refers to the degree of difference between the product and the behavioral data; ξ represents the constant factor, which is used to adjust the importance of the attenuation factor; DLF(p k , t) represents the decay factor score, which is used to measure the time of the user's most recent interaction with the product; ρ represents the constant factor, which is used to adjust the importance of rule-based learning weights; RLW(p k , t): score based on rule-based learning weights, which may be scores calculated by some predefined rules. ψ represents a constant factor used to adjust the importance of neural network-based scoring; RNN(p k , t) represents the rating based on the neural network, which may be the product score predicted by training the neural network; v represents the constant factor, which is used to adjust the importance of the rating based on the long short-term memory network; LSTM (p k , t) represents the rating based on the long short-term memory network, which is the product score predicted by the trained long short-term memory network.

8. An e-commerce data push system based on big data, characterized in that: include: A receiving module, configured to receive a real-time browsing behavior data stream from a user; A construction module, configured to dynamically construct a user interest model based on the real-time browsing behavior data stream, wherein the user interest model is configured to reflect the user's interest preferences; A matching module, configured to perform real-time matching based on the user interest model with product information in a product database to discover a set of products that match the user's interest preferences; A generating module, configured to generate a personalized data push package based on the product information in the product set and in combination with a preset data push strategy; The push module is used to push the personalized data push package to the user's terminal device via the network in real time, so that the user can receive product information push that meets the current interest preferences.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an e-commerce data push method based on big data as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, the e-commerce data push method based on big data as described in any one of claims 1 to 7 is implemented.

Images