TWM589312U - Product purchase evaluation system based on user web browsing history - Google Patents

Abstract

A commodity purchase evaluation system based on a user's Internet browsing record. The commodity purchase evaluation system is used for signal connection with an operation terminal. The commodity purchase evaluation system includes a service providing device, wherein the operation terminal generates a canvas fingerprint based on the browsing screen, and the service providing device Used to provide browsing content, store browsing records, provide historical transaction data and user data, and perform concatenation and expansion operations to generate at least one browsing record feature vector and at least one historical transaction feature vector, and perform purchase prediction analysis based on product information, And generate a purchase prediction analysis result and a label, and link the label with the user's identity code, based on the user's Internet browsing history for product purchase evaluation to achieve the combination of online browsing and offline purchase behavior for product purchase prediction evaluation .

Description

Commodity purchase evaluation system based on user's Internet browsing history

This case is about a product purchase evaluation system based on the user's Internet browsing records, especially a system that uses canvas fingerprints and user offline behaviors to make purchase evaluations.

The trading behavior of traditional commodities mainly comes from physical channels. With the evolution of technology, network channels are also one of the channels used by users. It is also common for users to browse online and offline to contact physical channels; however, Using online information, you can only observe whether there is a purchase at the time of browsing, and you cannot effectively analyze the offline purchase behavior after browsing; in particular, the characteristics, purposes, uses and effects of financial commodities are different from general commodities. Financial commodities have their complexity and The rights and interests of users are very deep. Therefore, in the past, many sales and purchases required an offline review and evaluation process, which made the sales and consumption behavior of financial products different from those of ordinary products. In addition, in order to evaluate the user's offline behavior, the current commonly used method is to identify the user through the user's active login to the account or use the browser's cookie to record the user's browsing behavior using the browser to achieve tracking usage Person behavior. However, in order to maintain the performance of the device and reduce the storage space, users may often clear the cookies, resulting in the loss of information based on cookies to record user behaviors, or during the general Internet browsing process to browse the interested Products, but due to various factors that are not logged in or purchased in real time, the products that the user is interested in are not recorded or the records are lost, and the user's behavior cannot be evaluated and provided to the user for reference.

Therefore, how to effectively combine online and offline heterogeneous data, through the training results of historical data, to evaluate and predict whether the user may purchase commodities in the future, as a reference for the sale of general commodities and financial commodities Problems that must be solved urgently.

In view of this, this author proposes a product purchase evaluation system based on the user's Internet browsing history.

According to an embodiment of the present invention, a product purchase evaluation system based on the user's Internet browsing record, wherein the product purchase evaluation system based on the user's Internet browsing record is used to connect with an operation terminal signal and based on the user's network browsing record The commodity purchase evaluation system includes an operation terminal and a service providing device. The operation terminal includes an input module, a display module and a canvas fingerprint generation module, wherein the input module is used to provide a browsing request and the display The module is used to display a browsing screen, and the canvas fingerprint generating module is used to generate a canvas fingerprint according to the user environment of the operation terminal when the browsing screen is generated; the service providing device includes a content providing module and a browsing record Module, a user data module, a historical transaction module, a feature expansion module and a commodity module, wherein the content providing module stores a plurality of content information and provides corresponding browsing content according to the browsing request Information to the operation terminal and generate a browsing record; the browsing record module to store the browsing record and the canvas fingerprint; the user data module to store at least one user data, the user data at least Contains a user identification code; the historical transaction module is used to store a plurality of historical transaction data corresponding to the user data; the commodity module is connected to store a plurality of commodity information; and the feature expansion module, Query the corresponding user identity code according to the canvas fingerprint, and query the user's historical transaction data in the analysis interval with the user identity code and an analysis interval, and the feature expansion module based on the browsing record and the historical transaction The data is expanded to generate at least one browsing record feature vector and at least one historical transaction feature vector. The feature expansion module performs purchase prediction analysis based on the browsing record feature vector, the historical transaction feature vector, and a commodity information, which is generated after the analysis A purchase prediction analysis result and a label, and link the label with the user identity code.

In order to make the technical content, purpose, and advantages of this creation easier to understand, the following describes the implementation of this creation in further detail with reference to the drawings. However, this description is a description of an exemplary embodiment, and is not intended to limit this creation Category.

As shown in FIG. 1, it is an embodiment of a product purchase evaluation system based on the user's Internet browsing record, in which the product purchase evaluation system based on the user's Internet browsing record is used to communicate with an operation terminal 100 via the network 110 signal connection, a service providing device 120 based on the commodity purchase evaluation system of the user's Internet browsing record. The operation terminal 100 and the service providing device 120 are connected to each other through the network 110. In the present creation, connection refers to a method of message transmission and exchange between modules, such as a wired connection or a wireless connection. In other embodiments, the operation terminal 100 may be directly connected to the service providing device 120.

In this embodiment, the operation terminal 100 is a computer device including an input module 101, a display module 102, and a canvas fingerprint generation module 103, such as a personal computer, a mobile phone, and a tablet computer. The input module 101 is used to provide a browsing request, the display module 102 is used to display a browsing screen, and the canvas fingerprint generation module 103 is used to generate a canvas fingerprint according to the user environment of the operation terminal 100 when the browsing screen is generated. The user environment refers to the browser settings, image processing settings, and hardware composition of the operating terminal 100, including proxy strings, screen color depth, language, plug-ins, compression settings, anti-aliasing, pixel rendering, drawing engine, driver And memory, but not limited to this.

The created service providing device 120 is used to provide browsing content, generate a purchase prediction analysis result and label, and link to the user's identity code to produce personalized product purchase evaluation results. The service providing device 120 includes a content providing module 121, a browsing record module 122, a user data module 123, a historical transaction data module 124, a feature expansion module 125, and a commodity module 125.

The content providing module 121 is a server, or a computer device with the same or similar functions, such as a web server (Web Server), which can store a plurality of content information and provide corresponding browsing according to the user's browsing request The content information is sent to the operation terminal 100, and a browsing record is generated. The browsing record includes a browsing attribute of browsing content information. The browsing attribute may be a web page URL, a web page attribute, a browsing time, a browsing day, a browsing product, a commodity amount, a click URL, a click time, or an order number, etc., but is not limited thereto.

The browsing record module 122 is a storage device for storing browsing records and canvas fingerprints to save the user's browsing records and canvas fingerprints at any time. Accordingly, even if the user's operation terminal 100 and the network 110 do not save the browsing records and For canvas fingerprints, the system can still perform commodity purchase evaluation based on the information in the browsing history module 122.

The user data module 123 is a storage device for storing at least one user data. The user data includes at least one user identity code. In other embodiments, the user data may also include a user account, password, and usage. Personal information such as the name, email address, telephone number, birthday, ID number, gender, age, education level, income, working experience, etc., but not limited to this.

The historical transaction data module 124 is used to store a plurality of historical transaction data corresponding to user data. The historical transaction data includes a transaction attribute. The transaction attribute may be the number of transactions, the transaction time, the number of transaction days, or the transaction amount, but is not limited to this .

The commodity module 125 is connected to store a plurality of commodity information. The commodity information may be a commodity type, a commodity amount, a commodity purchase restriction, a commodity content, etc. In this embodiment, when the commodity is a financial insurance commodity, the commodity information may be "insurance category", "insurance commodity name", "policy terms", "rate", "insured age", etc. In terms of insurance category It is also divided into "life insurance", "medical insurance", "accident insurance", "investment insurance", "international insurance", "travel insurance", "comprehensive insurance", etc., "policy terms" can be divided into "outpatient" , "Hospitalization", "Surgery", "Cancer", "Major Injuries", etc.

The feature expansion module 125 queries the corresponding user identity code based on the canvas fingerprint, and the user identity code and an analysis interval query the user's historical transaction data in the analysis interval, and the feature expansion module 125 queries the historical records and historical transaction data Perform an expansion operation to generate at least one browsing record feature vector and at least one historical transaction feature vector. The feature expansion module 125 can perform purchase forecast analysis based on the browsing record feature vector, historical transaction feature vector, and a commodity information, generate a purchase forecast analysis result and a label after the analysis, and link the label with the user identity code.

Further, to achieve the above purpose, the feature expansion module 125 of the present invention further includes a plurality of functional units. As shown in FIG. 2, the feature expansion module 125 may include an identity serial unit 201, a historical data expansion unit 202, A network data expansion unit 203, an interactive expansion unit 204, and an analysis interval determination unit 205. Wherein, the identity concatenation unit 201 obtains the user identity code, browsing record and historical transaction data from the service providing device 120, and then concatenates the user identity code, browsing record and historical transaction data to make the browsing record and historical transaction data Form an association, and the user's identity code is associated with browsing records or historical transaction data.

The historical data expansion unit 202 is used to perform an expansion operation on the historical transaction data. The expansion operation is based on at least one transaction attribute. The transaction attribute is the number of transactions, transaction time, transaction days, or transaction amount, such as the total number of purchases. , The number of purchases per day, the total number of days purchased and the average purchase amount, etc. For example, if a user has purchased travel insurance, annuity insurance, investment insurance and policy loans, the historical data expansion unit 202 based on the historical transaction data purchased by the user and the transaction attributes of the historical transaction data such as the number of transactions, transaction time, and transactions The number of days or transaction amount generates a historical transaction feature vector of (0.3, 0.7, 0, 0).

The network data expansion unit 203 is used to perform an expansion operation on the browsing record, wherein the expansion operation is calculated based on at least one browsing attribute of the content corresponding to the browsing record, the browsing attribute is a web page URL, a web page attribute, a browsing time, a browsing Days, browsed products, product amount, or number of purchases, such as splitting and identifying URL data, browsing behavior characteristics (such as browsing time and days), web page attributes (such as marketing activities, marketing time), browsed product name, amount, etc. To generate a browsing record feature vector.

The interactive expansion unit 204 is used to expand the interactive relationship between the browsing record after the expansion operation of the network data expansion unit 203 and the historical operation after the expansion operation of the historical data expansion unit 202 to generate an interactive expansion feature vector. The interaction relationship can be It is the action before each browsing (Internet 110 browsing/commodity purchase), the total number of webpages browsed before each purchase, etc., but it is not limited to this. Purchase forecast analysis

The feature determination unit 206 is used to analyze the correlation between the browsing record feature vector, the historical transaction feature vector, and the interactive expansion feature vector and the actual purchase, so as to eliminate the interference of unnecessary information and enhance the accuracy of machine learning. For example, the feature determination unit 206 takes the absolute value of the correlation of the fields of the browsing record feature vector or the historical transaction feature vector to generate a predicted label, and excludes the fields with low correlation. The method of determining correlation is a product correlation Analytical methods, such as Pearson Correlation. In this embodiment, when the feature determination unit 206 uses the Pearson product difference correlation analysis to browse the record feature vector or the historical transaction feature vector, if the absolute value of the correlation generated is greater than or equal to the preset value of a predicted label, it means the aforementioned browsing Record feature vectors, historical transaction feature vectors, and interactive expansion feature vectors are related to actual purchases. For example, the correlation threshold of a product's "willing to buy" prediction tag is 0.3. When the feature determination unit 206 calculates a user's browsing record When the absolute value of the correlation between the feature vector, the historical transaction feature vector, and the interactive extended feature vector is 0.5, because 0.5 is greater than the threshold value of 0.3, it is determined to be related, and the user is associated with the predicted tag with the purchase intention. However, those with ordinary knowledge in the technical field should understand that the product difference correlation analysis method and the absolute value are not limited to this.

The analysis interval determination unit 205 is used to determine the analysis interval, use different lengths of time as the analysis interval for backtesting calculation, and browse the record feature vector, historical transaction feature vector and commodity information for purchase prediction analysis tags in a continuously advancing manner. For example, the analysis interval can be 5, 10, 20, 30, 90, 180, 365, or 730, etc., and the purchase probability is calculated individually to determine the optimal analysis interval of the product. In this embodiment In the backtesting calculation, it is a multi-category classification algorithm, and generates an F1 score as the purchase probability, and the highest analysis interval is the one with the highest F1 score. However, those with ordinary knowledge in the technical field should understand that the method of determining the backtest calculation and the optimal analysis interval is not limited to this.

As shown in FIG. 3, an implementation process of the commodity purchase evaluation system based on the browsing records of the user network 110 is used for the operation of the aforementioned operation terminal 100 and the service providing device 120. The steps include: Step 301, from an operation terminal 100 provides a browsing request; step 302, the service providing device 120 provides a corresponding browsing content information to the operating terminal 100 according to the browsing request and generates a browsing record, the operating terminal 100 generates a browsing screen according to the browsing content information; step 303, operating the terminal 100 generates a canvas fingerprint based on the browsing screen; step 304, the service providing device 120 stores the browsing record and canvas fingerprint; step 305, queries a corresponding user identity code based on the canvas fingerprint; step 306, based on the user identity code and an analysis interval Query the user's historical transaction data in the analysis interval; Step 307, perform an extended operation on the browsing record and historical transaction data to generate at least one browsing record feature vector and at least one historical transaction feature vector; Step 308, obtain a commodity information, based on browsing The recorded feature vector, the historical transaction feature vector, and the commodity information are used to make a purchase prediction analysis, and a purchase prediction analysis result is generated after the analysis; and step 309, a label is generated according to the purchase prediction analysis result and linked to the user identity code.

Next, the steps of the implementation process of a product purchase evaluation system based on the browsing records of the user's network 110 will be further described. Please refer to FIGS. 1 to 3. In step 301, the user can use the input module of the operation terminal 100 101 For example, the keyboard of a personal computer enters a webpage or product to be queried for inquiries. When the server of the network 110 replies to the corresponding webpage, the user clicks the webpage to be opened to issue a browsing request to the network 110110 and the service providing device 120.

Step 302, after receiving the aforementioned browsing request, the service providing device 120 generates a corresponding browsing content information from the content providing module 121 and provides it to the operation terminal 100. The operation terminal 100 generates a browsing screen according to the browsing content information and passes the display module 102 provides the user with browsing, and the operation terminal 100 generates a browsing record accordingly.

Step 303: The canvas fingerprint generation module 103 of the operation terminal 100 generates a canvas fingerprint according to the browsing screen. The canvas fingerprint is calculated based on the browser information and device information of the operation terminal 100. The foregoing information may include: browser brand, Browser settings, browser version, browser language, browser font, software environment when the browser is running, screen quality, system version, hardware information, etc.; due to the different hardware and software settings of the operating terminal 100 for different users All the same, based on different graphics processing engines, different image generation methods, different compression ratios, anti-aliasing, rendering operations, etc., the canvas fingerprints generated by the same image are not the same, so they can be commended for specific users.

In step 304, the operation terminal 100 provides the browsing record and the canvas fingerprint to the service providing device 120, and the browsing record module 122 of the service providing device 120 is used to store the browsing record and the canvas fingerprint.

In step 305, the identity serial unit 201 of the feature expansion module 125 can query the user data module 123 for a corresponding user identity code according to the canvas fingerprint.

In step 306, the identity concatenation unit 201 of the feature expansion module 125 queries the historical transaction data module 124 for historical transaction data of the user in the analysis interval according to an analysis interval provided by the user identity code and the analysis interval determination unit 205. The analysis interval may be a preset period, such as a preset 5, 10, 20, 30, 90, 180, 365 or 730 days, or the analysis interval determining unit 205 may use different lengths The time is used as the analysis interval for backtesting calculations, and the record feature vector, historical transaction feature vector and product information are browsed in a forward-going manner to carry out the purchase probability analysis label, and the purchase probability is calculated individually to determine the best analysis interval for the product. For example, the best analysis interval is determined to be 180 days from 5, 10, 20, 30, 90, 180, 365 or 730 days. In other embodiments, after the historical transaction data is obtained, a serial connection step may be further performed to connect the user identity code, browsing records and historical transaction data.

In step 307, after receiving the browsing record feature vector and the at least one historical transaction feature vector, the feature expansion module 125 will first perform preprocessing. Next, the network data expansion unit 203 performs an expansion operation on the browsing record. The content corresponding to the browsing record contains at least one browsing attribute. The browsing attribute may be a web page URL, a web page attribute, a browsing time, a browsing day, a browsing commodity, and a commodity amount. Or the number of purchases, here, the calculation of browsing attributes, such as splitting and identifying URL data, browsing behavior characteristics (such as browsing time and browsing days), web page attributes (such as marketing activities, marketing time), and the name of the browsed product , Amount confirmation, etc. to generate at least one browsing record feature vector. The historical data expansion unit 202 performs an expansion operation on the transaction attributes in the historical transaction data. The transaction attributes may be the number of transactions, the transaction time, the number of transaction days, or the transaction amount, such as the total number of purchases, the number of purchases per day, the total number of days purchased, and the average purchase The amount, etc., generates at least one historical transaction feature vector.

Step 308, the feature expansion module 125 obtains a commodity information from the commodity module 125, and the interactive expansion unit 204 performs a purchase prediction analysis based on the browsing record feature vector, the historical transaction feature vector and the commodity information, and generates a purchase prediction analysis result after the analysis, To determine the probability of each user buying a single type of goods. In other embodiments, the purchase prediction analysis system includes a purchase intention vector generation step,

Step 309, generate a label according to the result of the purchase prediction analysis and link it with the user's identity code. In this embodiment, the label may be purchased or not purchased. In other embodiments, the label may be the value of the purchase probability (20%, 40 %, 60%, 80%, 90%, 99%, etc.) or the conclusion corresponding to this value (low, slightly lower, average, slightly higher, high, extremely high, etc.).

In other embodiments, the implementation process of the created product purchase evaluation system further includes a correlation analysis step for analyzing the correlation between the browsing record feature vector, the historical transaction feature vector, and the commodity information, when the browsing When the value of the record feature vector and the historical transaction feature vector is greater than one of the thresholds of the commodity information, it is determined to be highly correlated; when the values of the browse record feature vector and the historical transaction feature vector are less than the threshold of the commodity information, Judgment is not relevant.

Although this creation has disclosed the above embodiments with examples, it is not intended to limit this creation. Anyone with ordinary knowledge in the technical field of the subject can make some changes and modifications without departing from the spirit and scope of this creation. , Should be the scope of rights claimed by the patent, so the scope of protection of this patent shall be deemed as defined by the scope of the attached patent application.

100‧‧‧Operation terminal 125‧‧‧Feature expansion module 101‧‧‧ input module 126‧‧‧Commodity module 102‧‧‧Display module 201‧‧‧ Identity Series Unit 103‧‧‧Canvas fingerprint generation module 202‧‧‧Historical data expansion unit 110‧‧‧ Internet 203‧‧‧Network data expansion unit 120‧‧‧Service provision device 204‧‧‧Interactive Expansion Unit 121‧‧‧Content providing module 205‧‧‧Analysis interval determination unit 122‧‧‧Browsing record module 206‧‧‧Feature determination unit 123‧‧‧ User Data Module 301-309‧‧‧Step 124‧‧‧ Historical Transaction Data Module

FIG. 1 is a schematic diagram of an exemplary system embodiment of a merchandise purchase evaluation system based on a user’s Internet browsing history. FIG. 2 is a schematic diagram of an exemplary module embodiment of the created feature expansion module. FIG. 3 is a schematic diagram of an exemplary system implementation process of a product purchase evaluation system based on the user’s Internet browsing records.

100‧‧‧Operation terminal

101‧‧‧ input module

102‧‧‧Display module

103‧‧‧Canvas fingerprint generation module

110‧‧‧ Internet

120‧‧‧Service provision device

121‧‧‧Content providing module

122‧‧‧Browsing record module

123‧‧‧ User Data Module

124‧‧‧ Historical Transaction Data Module

125‧‧‧Feature expansion module

126‧‧‧Commodity module

Claims (11)

A merchandise purchase evaluation system based on the user's Internet browsing records is used to connect with an operation terminal signal, wherein the operation terminal includes an input module, a display module, and a canvas fingerprint generation module, wherein the input The module is used to provide a browse request, the display module is used to display a browse screen, and the canvas fingerprint generation module is used to generate a canvas fingerprint based on the browse screen. The system includes: A service providing device including a content providing module, a browsing record module, a user data module, a historical transaction module, a feature expansion module, and a commodity module, wherein: The content providing module stores a plurality of content information, provides corresponding browsing content information to the operation terminal according to the browsing request, and generates a browsing record; The browsing record module is used to store the browsing record and the canvas fingerprint; The user data module is used to store at least one user data, and the user data includes at least one user identity code; The historical transaction module is used to store a plurality of historical transaction data corresponding to the user data; The commodity module is connected to store a plurality of commodity information; and The feature expansion module queries the corresponding user identity code according to the canvas fingerprint, and the user identity code and an analysis interval query the user's historical transaction data in the analysis interval, and the feature expansion module is based on the The browsing record and the historical transaction data are expanded to generate at least one browsing record feature vector and at least one historical transaction feature vector. The feature expansion module makes a purchase prediction based on the browsing record feature vector, the historical transaction feature vector, and a commodity information After the analysis, a purchase prediction analysis result and a label are generated, and the label is linked to the user identity code. The commodity purchase evaluation system according to claim 1, wherein the feature expansion module includes an identity serial unit, a historical data expansion unit, a network data expansion unit, an interactive expansion unit, and an analysis interval determination unit. The commodity purchase evaluation system according to claim 1 or 2, wherein the feature expansion module further includes a feature determination unit for analyzing the browsing record feature vector, the historical transaction feature vector and the product information Of relevance. The commodity purchase evaluation system according to claim 3, wherein when the value of the browsing record feature vector and the historical transaction feature vector is greater than one of the thresholds of the commodity information, it is determined to be highly correlated; when the browsing record feature vector, the When the value of the historical transaction feature vector is less than the threshold of the commodity information, it is judged as not relevant. The commodity purchase evaluation system according to claim 2, wherein the identity concatenation unit is based on concatenating the user identity code, the browsing record and the historical transaction data. The commodity purchase evaluation system according to claim 2, wherein the historical data expansion unit is used to perform an expansion operation on the historical transaction data, wherein the expansion operation is calculated based on at least one transaction attribute, and the transaction attribute is the number of transactions, Transaction time, transaction days or transaction amount. The commodity purchase evaluation system according to claim 2, wherein the network data expansion unit is used to perform an expansion operation on the browsing record, wherein the expansion operation is calculated based on at least one browsing attribute, the browsing attribute is a web page URL, Web page attributes, browsing time, browsing days, browsing products, product amount or number of purchases. The commodity purchase evaluation system as described in claim 2, wherein the interactive expansion unit is used to perform purchase prediction analysis. The commodity purchase evaluation system according to claim 2, wherein the purchase prediction analysis includes a purchase intention vector. The commodity purchase evaluation system according to claim 2, wherein the analysis interval determination unit is used to determine the analysis interval. The commodity purchase evaluation system according to claim 10, wherein the analysis interval is 5, 10, 20, 30, 90, 180, 365 or 730 days.

Images