Multi-dimensional analysis product pushing system, method and server

Abstract

The invention discloses a multi-dimensional analysis product pushing system, and the system comprises a data processing module which is used for carrying out the vectorization of the fitting data of a target product, and storing the fitting data in a vector database, and the fitting data comprises a right and interest type, a right and interest type, and a flow range type; the product integrating degree calculation module is used for performing integrating degree calculation on the user data and the integrating data to obtain the total integrating degree of the target product and the user; and the judging and pushing module is used for judging whether a notification of a target product is pushed to the user or not according to the total integrating degree. According to a specific scheme, the system can accurately push the notification to a user interested in the notification, and a user not interested in the notification cannot receive the notification. The risk of pushing the notification is reduced to the maximum extent, the effectiveness of the notification is improved to the maximum extent, the user who is interested in the notification and even has potential ordering can receive the notification, and the risk of disgusting of the user is reduced to the maximum extent while the ordering amount is improved.

Description

Technical Field

[0001] The present application belongs to the field of computers, and in particular, relates to a product push system, method and server for multi-dimensional analysis. Background Art

[0002] When a new product is launched on an existing website or APP, a message notification will be pushed to the user. In the traffic equity product scenario, if a new product is released, the information of the new product will be pushed to the user. After receiving the message notification, if the user is interested in the new product, he will click on the notification to view and order it. If the user is not interested in the new product, he will ignore the notification.

[0003] However, there are also users who are not interested in this push notification. After seeing the push notification, they ignore it directly or even disable the push permission of the APP. This makes it impossible to reach this user through push notifications later. Therefore, whether it is possible to accurately judge whether the user is interested in this product is related to the user experience and the actual push effect.

[0004] In the prior art, there are usually two solutions for product push: The first solution: When a new product is launched, a recommendation notification of the product is sent to all users. Users who are interested in the product may click on the notification to view it, while users who are not interested in the product may ignore the notification or even block the push permission of the APP. This solution is not targeted and pushes all users uniformly, which is very likely to cause a large number of users to be disgusted, causing users to block APP notifications or even uninstall the APP.

[0005] The second solution: When a new product is launched, query the order records of similar products. If there are users with similar product purchase records, push this product. For example, if the new product is 10G Tencent Video traffic, users with Tencent Video-related order records will be pushed notifications of the new product, and other users will not be pushed. This solution pushes to users with order records, involves a narrow group of people, and has a lower risk. However, it is not highly targeted. Users who have downloaded the same type of order may not be interested in this notification, and the push effect is not high.

[0006] Therefore, there is an urgent need for a traffic equity product push solution that can solve the above problems. Summary of the invention

[0007] In order to address the deficiencies of the prior art, the present application provides a product push system, method and server for multi-dimensional analysis, which calculates the degree of compatibility between traffic equity products and user preferences, and sends the products to users in a targeted manner, thereby minimizing the risk of user disgust.

[0008] The technical effects to be achieved by this application are achieved through the following solutions: According to a first aspect of the present application, a product push system for multi-dimensional analysis is provided, comprising: A data processing module, used to vectorize the matching data of the target product and store it in a vector database, wherein the matching data includes equity category, equity type and flow range type; A product compatibility calculation module, used to calculate the compatibility between the user data and the compatibility data to obtain the total compatibility between the target product and the user; The determination and push module is used to determine whether to push a notification of the target product to the user according to the total compatibility.

[0009] Preferably, the user data includes at least historical orders, search terms and high-frequency browsing. The user data is calculated with the three types of matching data to obtain the matching degree, and then the average of all matching degrees is calculated to obtain the total matching degree.

[0010] Preferably, the specific method for calculating the fit of historical orders is: Count the equity category, equity type and flow range type of all historical orders of the user; Calculate the proportion of the target product's equity category in the equity category of historical orders: T1; the proportion of the target product's equity type in the equity type of historical orders: T2; the proportion of the target product's flow range type in the flow range type of historical orders: T3; The fit of historical orders is calculated according to the following formula: QT=(T1+T2+T3) / 3.

[0011] Preferably, the specific method for calculating the fit of the search terms is: Count the equity categories, equity types, and traffic range types of all search terms of the user; Calculate the proportion of the target product's equity category in the search term's equity category: S1; the proportion of the target product's equity type in the search term's equity type: S2; the proportion of the target product's traffic range type in the search term's traffic range type: S3; The fit of the search term is calculated according to the following formula: QS=(S1+S2+S3) / 3.

[0012] Preferably, the specific method for calculating the compatibility of high-frequency browsing is: Count all the high-frequency browsing rights and interests categories, rights and interests types, and traffic range types of the user; Calculate the proportion of the target product's equity category in the frequently browsed equity category: F1; the proportion of the target product's equity type in the frequently browsed equity type: F2; the proportion of the target product's traffic range type in the frequently browsed traffic range type: F3; The fit of high-frequency browsing is calculated according to the following formula: QF=(F1+F2+F3) / 3.

[0013] Preferably, when calculating the degree of fit: If the current user has no historical orders or the number of historical orders is less than 5, QT will be counted as empty; If the current user has no search terms or the number of search terms is less than 20, the QS is counted as empty; If the number of products frequently browsed by the current user is less than 100, QF is counted as empty.

[0014] Preferably, determining whether to push a notification of a target product to the user according to the total degree of fit is specifically as follows: A compatibility threshold for the pushed product is preset, and the total compatibility and favorability are added together and then compared with the compatibility threshold. If the compatibility is greater than the compatibility threshold, a notification of the target product is pushed to the user, otherwise, no notification is pushed; The favorability is a score of the user's acceptance of historical push notifications.

[0015] Preferably, it also includes a favorability adjustment module, which is used to adjust the favorability score, specifically: Set the initial value of favorability; The favorability is adjusted according to the user's processing of the target product notification: If the user ignores the notification, the favorability score will be reduced by 0.02; If the user clicks on the notification to view but does not order, the favorability score will be increased by 0.03; If the user clicks on the notification to view it and successfully places an order, the favorability score will be increased by 0.05.

[0016] According to a second aspect of the present application, a product push method using the product push system of the multi-dimensional analysis is provided, characterized in that it comprises the following steps: Step 1: vectorize the matching data of the target product and store it in a vector database, wherein the matching data includes equity category, equity type and flow range type; Step 2: Calculate the degree of fit between the user data and the fit data to obtain the total degree of fit between the target product and the user; Step 3: Determine whether to push a notification of the target product to the user based on the total compatibility.

[0017] According to a third aspect of the present application, there is provided a server, comprising: a memory and at least one processor; The memory stores a computer program, and the at least one processor executes the computer program stored in the memory to implement the above-mentioned product push method.

[0018] According to an embodiment of the present application, the beneficial effect of the product push system using the multi-dimensional analysis of the present application is that: according to the specific scheme, the notification is accurately pushed to the users who are interested in the notification, and the users who are not interested will not receive the notification. The risk of push notification is minimized, the effectiveness of the notification is maximized, and the users who are interested in the notification or even potential ordering receive the notification, which increases the order volume while minimizing the risk of user disgust. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the existing technical solutions, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0020] Figure 1 This is a structural block diagram of a product push system for multi-dimensional analysis in one embodiment of the present application; Figure 2 This is a flow chart of a product push method for multi-dimensional analysis in one embodiment of the present application; Figure 3 This is a structural block diagram of a server in an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. 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 ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0022] like Figure 1 As shown, the product push system for multi-dimensional analysis in one embodiment of the present application includes: A data processing module, used to vectorize the matching data of the target product and store it in a vector database, wherein the matching data includes equity category, equity type and flow range type; Among them, the target product is the traffic or equity product involved in the planned push notification; the equity category is, for example, travel, video, music, social communication, tools, and shopping; the equity type is, for example, multiple equity or single equity; the traffic range type is, for example, general traffic or directional traffic.

[0023] Vectorize the fit data, converting words or phrases in the text into numerical vectors. These vectors capture the semantic and syntactic properties of the vocabulary, and mathematical operations such as addition and subtraction can be performed on these vectors to reveal the relationships between words, and they can be used in machine learning and deep learning models.

[0024] For example: for the sentence "I want to buy international data traffic", first build a vocabulary that includes all the unique words that appear in the corpus. The vocabulary is represented as: {'I': 0, 'want': 1, 'buy': 2, 'international': 3, 'data traffic': 4}; Then create One-Hot vectors. For each word, a vector with a length equal to the size of the vocabulary will be created. Except for the position corresponding to the vocabulary being 1, all other positions are 0. For example, for the word "international", its One-Hot vector representation is "international": [0, 0, 0, 1, 0].

[0025] The product fit calculation module is used to calculate the fit between the user data and the fit data to obtain the total fit between the target product and the user; The user data includes at least: historical orders, search terms, and high-frequency browsing. Calculate the fit between the user data and three types of fit data respectively, and then calculate the average value of all fits to obtain the total fit.

[0026] The specific method for calculating the fit of historical orders is: count the rights and interests categories, rights and interests types, and traffic range categories of all the user's historical orders, and segment the product name part in the order. After segmentation, it is also added to the statistical category list. Calculate the proportion of the target product in this dimension in all historical orders, where the cosine similarity of the word vector > 0.8 is considered a match: The proportion of the rights and interests category of the target product in the rights and interests categories of historical orders: T1; the proportion of the rights and interests type of the target product in the rights and interests types of historical orders: T2; the proportion of the traffic range type of the target product in the traffic range types of historical orders: T3; Calculate the fit of historical orders according to the following formula: QT = (T1 + T2 + T3) / 3.

[0027] If the user has no historical orders or the number of historical orders is less than 5, then QT is not calculated, that is, QT is empty.

[0028] The specific method for calculating the fit of search terms is: count all valid search terms of users, count them according to the three dimensions of equity category, equity type, and traffic range category, and segment the search content, and add it to the statistical category list after segmentation. If the search term is a specific dimension, such as: video, music, etc., it is directly counted; if the search term is the name of a partner, such as: Tencent, query the equity type and equity category corresponding to the partner, and then count it in the corresponding dimension, and finally count the proportion of the fit dimension of the target product in the search term, where the cosine similarity of the word vector> 0.8 is a match: Calculate the proportion of the target product's equity category in the search term's equity category: S1; the proportion of the target product's equity type in the search term's equity type: S2; the proportion of the target product's traffic range type in the search term's traffic range type: S3; calculate the fit of the search term according to the following formula: QS=(S1+S2+S3) / 3.

[0029] If the user has no search record, or the number of valid search terms is less than 20, QS is not calculated, that is, QS is empty.

[0030] The specific method for calculating the fit of high-frequency browsing is: count the user's browsing history products, count the rights and interests categories, rights and interests types, and traffic range categories corresponding to the products, and segment the product names, and add them to the statistical category list after segmentation. Calculate the product fit dimension of the target product and the proportion of the browsed products. The cosine similarity of the word vector>0.8 is a match: Calculate the proportion of the target product's equity category in the frequently browsed equity category: F1; the proportion of the target product's equity type in the frequently browsed equity type: F2; the proportion of the target product's traffic range type in the frequently browsed traffic range type: F3; calculate the fit of high-frequency browsing according to the following formula: QF=(F1+F2+F3) / 3.

[0031] If the number of products in the user's browsing history is less than 100, QF is not calculated, that is, QF is empty.

[0032] In the process of word segmentation for historical orders, search terms, and high-frequency browsing, deep learning models (such as recurrent neural networks (RNN) or long short-term memory (LSTM)) are used for word segmentation.

[0033] Cosine similarity is used to measure the angle between two non-zero vectors, thereby determining the similarity of the two vectors. In this embodiment, it is used to determine the matching degree of the target product with different user data, thereby determining the proportion.

[0034] The determination and push module is used to determine whether to push the notification of the target product to the user according to the total fit. Specifically: The preset match threshold QW=0.8 for the pushed product is used to determine whether to push notifications of the target product to the user based on the match and favorability of the product: If Q + L>= QW, the notification of the target product is pushed to the user, otherwise it is not pushed.

[0035] L is the favorability, that is, the score of the user's acceptance of the historical push notification. The initial value is 0. According to whether the user accepts the notification, the favorability adjustment module adjusts the user's favorability score and stores it. Specifically: After pushing notifications to users, adjust the user's favorability based on their response.

[0036] If the user ignores the notification, the favorability decreases: L=L-0.02; If the user clicks on the notification to view the product but does not order it, the favorability will increase by: L=L+0.03; If the user clicks on the notification to view the product and successfully places an order, the favorability will increase by: L=L+0.05.

[0037] like Figure 2 As shown, the product push method of the product push system using the multi-dimensional analysis is characterized by comprising the following steps: Step 1: vectorize the matching data of the target product and store it in a vector database, wherein the matching data includes equity category, equity type and flow range type; Step 2: Calculate the degree of fit between the user data and the fit data to obtain the total degree of fit between the target product and the user; The user data in this step includes historical orders, search terms, and high-frequency browsing. The fit between the target product and historical orders, QT, the fit between the target product and search terms, QS, and the fit between the target product and high-frequency browsing are calculated respectively. The product fit is calculated based on historical orders, search terms, and high-frequency browsing. The total fit between the target product and the user is: Q = (QT + QS + QF) / 3.

[0038] Step 3: Determine whether to push a notification of the target product to the user based on the total compatibility.

[0039] In this step, based on the overall product fit and favorability, determine whether to push notifications of the target product to the user: If Q + L>= QW, the notification of the target product is pushed to the user, otherwise it is not pushed. Q is the total fit, L is the favorability, and QW is the fit threshold.

[0040] Other steps may include: Adjust the user's favorability based on whether the user receives the notification and store it. After pushing the notification to the user, adjust the user's favorability based on the user's response: If the user ignores the notification, the favorability decreases: L = L - 0.02; If the user clicks on the notification to view the product but does not order it, the favorability will increase by: L = L + 0.03; If the user clicks on the notification to view the product and successfully places an order, the user’s favorability will increase by: L = L + 0.05.

[0041] like Figure 3 As shown, a server in an embodiment of the present application includes: a memory 301 and at least one processor 302; The memory 301 stores a computer program, and the at least one processor 302 executes the computer program stored in the memory 301 to implement the above-mentioned product push method.

[0042] According to an embodiment of the present application, the beneficial effect of the product push system using the multi-dimensional analysis of the present application is that: according to the specific scheme, the notification is accurately pushed to the users who are interested in the notification, and the users who are not interested will not receive the notification. The risk of push notification is minimized, the effectiveness of the notification is maximized, and the users who are interested in the notification or even potential ordering receive the notification, which increases the order volume while minimizing the risk of user disgust.

[0043] It should be noted that the above detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present application belongs.

[0044] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should also be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0045] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0046] In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products, or apparatuses.

[0047] For ease of description, spatially relative terms, such as "above", "above", "on the upper surface of", "above", etc., may be used herein to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" may include both "above" and "below". The device may also be positioned in other different ways, such as rotated 90 degrees or in other orientations, and the spatially relative descriptions used herein are interpreted accordingly.

[0048] In the above detailed description, reference is made to the accompanying drawings, which form a part of this document. In the accompanying drawings, similar symbols typically identify similar components unless the context indicates otherwise. The illustrated embodiments described in the detailed description, drawings, and claims are not meant to be limiting. Other embodiments may be used, and other changes may be made, without departing from the spirit or scope of the subject matter presented herein.

[0049] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A product push system for multi-dimensional analysis, characterized in that: include: A data processing module, used to vectorize the matching data of the target product and store it in a vector database, wherein the matching data includes equity category, equity type and flow range type; A product compatibility calculation module, used to calculate the compatibility between the user data and the compatibility data to obtain the total compatibility between the target product and the user; The determination and push module is used to determine whether to push a notification of the target product to the user according to the total compatibility.

2. The product push system of multi-dimensional analysis according to claim 1 is characterized in that: The user data at least includes: historical orders, search terms and high-frequency browsing. The user data is calculated with the three types of matching data to obtain the matching degree, and then the average of all the matching degrees is calculated to obtain the total matching degree.

3. The product push system of multi-dimensional analysis according to claim 2 is characterized in that: The specific method for calculating the fit of historical orders is: Count the equity category, equity type and flow range type of all historical orders of the user; Calculate the proportion of the target product's equity category in the equity category of historical orders: T1; the proportion of the target product's equity type in the equity type of historical orders: T2; the proportion of the target product's flow range type in the flow range type of historical orders: T3; The fit of historical orders is calculated according to the following formula: QT=(T1+T2+T3) / 3.

4. The product push system of multi-dimensional analysis according to claim 3 is characterized in that: The specific method for calculating the fit of the search terms is: Count the equity categories, equity types, and traffic range types of all search terms of the user; Calculate the proportion of the target product's equity category in the search term's equity category: S1; the proportion of the target product's equity type in the search term's equity type: S2; the proportion of the target product's traffic range type in the search term's traffic range type: S3; The fit of the search term is calculated according to the following formula: QS=(S1+S2+S3) / 3.

5. The product push system of multi-dimensional analysis according to claim 4 is characterized in that: The specific method for calculating the relevance of high-frequency browsing is: Count all the high-frequency browsing rights and interests categories, rights and interests types, and traffic range types of the user; Calculate the proportion of the target product's equity category in the frequently browsed equity category: F1; the proportion of the target product's equity type in the frequently browsed equity type: F2; the proportion of the target product's traffic range type in the frequently browsed traffic range type: F3; The fit of high-frequency browsing is calculated according to the following formula: QF=(F1+F2+F3) / 3.

6. The product push system of multi-dimensional analysis according to claim 5, characterized in that: When calculating the fit: If the current user has no historical orders or the number of historical orders is less than 5, QT will be counted as empty; If the current user has no search terms or the number of search terms is less than 20, the QS is counted as empty; If the number of products frequently browsed by the current user is less than 100, QF is counted as empty.

7. The product push system of multi-dimensional analysis according to claim 5, characterized in that: Determining whether to push a notification of a target product to the user according to the total fit is specifically as follows: A compatibility threshold for the pushed product is preset, and the total compatibility and favorability are added together and then compared with the compatibility threshold. If the compatibility is greater than the compatibility threshold, a notification of the target product is pushed to the user, otherwise, no notification is pushed; The favorability is a score of the user's acceptance of historical push notifications.

8. The product push system of multi-dimensional analysis according to claim 7, characterized in that: It also includes a favorability adjustment module, which is used to adjust the favorability score, specifically: Set the initial value of favorability; The favorability is adjusted according to the user's processing of the target product notification: If the user ignores the notification, the favorability score will be reduced by 0.02; If the user clicks on the notification to view but does not order, the favorability score will be increased by 0.03; If the user clicks on the notification to view it and successfully places an order, the favorability score will be increased by 0.

05.

9. A product push method using the product push system of the multi-dimensional analysis according to any one of claims 1 to 8, characterized in that: The steps include: Step 1: vectorize the matching data of the target product and store it in a vector database, wherein the matching data includes equity category, equity type and flow range type; Step 2: Calculate the degree of compatibility between the user data and the matching data to obtain the total degree of compatibility between the target product and the user; Step 3: Determine whether to push a notification of the target product to the user based on the total compatibility.

10. A server, characterized in that: include: memory and at least one processor; The memory stores a computer program, and the at least one processor executes the computer program stored in the memory to implement the product push method described in claim 9.

Images