A method, system and storage medium for building user portraits based on a retail SaaS platform

Abstract

The present invention provides a construction method, system, storage medium and electronic device based on user portrait of retail SaaS platform, obtains commodity information and user information, portraits user attribute behavior, establishes user feature matrix, establishes commodity feature matrix, calculates commodity association and user interest, generates a first recommendation list, calculates the similarity between the characteristic value in the commodity feature matrix of each commodity whose user interest ranking is within a threshold range and the characteristic value in the user feature matrix of the current user, sorts the commodities according to the similarity, and selects commodities whose similarity meets the threshold for display. The present invention performs multi-layer convolution operation on commodity association and then calculates user interest, closely associates commodity association with user interest, and improves the accuracy of user interest commodity search.

Description

Technical Field

[0001] The present invention relates to the technical field of software SaaS platform, and in particular to a method, system and storage medium for constructing user portraits based on a retail SaaS platform. Background Art

[0002] The economic environment in the era of big data has undergone tremendous changes, and the changes in the environment have prompted customer needs to be more dynamic and multi-dimensional. In order to fully realize customer value and meet the requirements of the development of the times, the value creation mechanism of my country's retail enterprises needs to be optimized. The arrival of the big data era means a fundamental change in the production and operation methods of enterprises. As an important pillar industry of the service industry, the retail industry needs to respond to market changes and user needs in a timely manner and improve operational efficiency. How to effectively use big data resources and then realize enterprise value-added has become an urgent problem that every retail enterprise needs to solve.

[0003] In the prior art, precision marketing is achieved through user profiling. However, the existing method of implementing user profiling cannot accurately determine the products that users are interested in. Summary of the invention

[0004] In view of this, the present invention provides a method, system and storage medium for constructing user portraits based on the retail SaaS platform. According to the retail SaaS big data information, user portrait and product portrait analysis are performed, and a user feature matrix and a product feature matrix are established. According to the user's historical behavior information, the correlation and interest are calculated, and the eigenvalue calculation algorithm is used to realize the eigenvalue similarity calculation, thereby realizing accurate product push service and improving the accuracy of customer demand budget.

[0005] In a first aspect, an embodiment of the present invention provides a method for constructing a user portrait based on a retail SaaS platform, comprising:

[0006] Step 101, obtaining product information and user information;

[0007] Step 102, profile the user's attribute behavior and establish a user feature matrix;

[0008] Step 103, establishing a commodity feature matrix;

[0009] Step 104, calculating the commodity relevance and user interest, and generating a first recommendation list;

[0010] Step 105, calculating the similarity between the feature value in the product feature matrix of each product in the first recommendation list whose user interest ranking is within the threshold range and the feature value in the user feature matrix of the current user;

[0011] Step 106, sort the products according to their similarity, and select products whose similarity meets a threshold value for display.

[0012] Furthermore, the establishment of the user feature matrix is ​​specifically as follows:

[0013] Establish user feature matrix Y S (a,b,c,d,e,f), where S represents the user's identity number, a represents the user's gender information, b represents the user's age, c represents the deviation of the longitude and latitude of the user's frequently used address from the longitude and latitude of the center of Beijing, d represents the attribute type number value of the most purchased product by the user, e represents the attribute type number value of the most collected product by the user, and f represents the user's characteristic value;

[0014] The user characteristic value f is calculated as follows:

[0015] f=w1*a+w2*b+w3*c+w4*d+w5*e

[0016] Among them, the value range of w1, w2, w3, w4, and w5 is 0 to 1.

[0017] Furthermore, the establishment of the commodity feature matrix is ​​specifically as follows:

[0018] Establish product feature matrix W B (a1,b1,c1,d1,e1,f1),

[0019] Where B is the product ID, a1 is the probability that the user who purchases the product is male, b1 is the average age of users who purchase the product, c1 is the deviation of the longitude and latitude of the product location from the longitude and latitude of the center of Beijing, d1 is the product attribute type ID, e1 is the product attribute type ID with the highest similarity to itself, and f1 is the product feature value;

[0020] The user characteristic value f1 is calculated as follows:

[0021] f1=w6*a1+w7*b1+w8*c1+w9*d1+w10*e1

[0022] Among them, w6, w7, w8, w9, and w10 have values ​​ranging from 0 to 1.

[0023] Furthermore, the calculation of the commodity association degree is specifically as follows:

[0024] Calculate the product correlation g between the product i purchased by the current user and all other products ij :

[0025]

[0026] Among them, g ijrepresents the correlation between product i and product j, where T(i)∩T(j) represents the probability of purchasing both product i and product j in all purchase records on the SaaS platform; T(i) represents the probability of purchasing only product i in all purchase records on the SaaS platform; T(j) represents the probability of purchasing only product j in all purchase records on the SaaS platform;

[0027] Furthermore, the user interest degree calculation is specifically as follows:

[0028] Establish product association matrix K B (x, y, g ij ), sort the products according to the correlation g ij Load the product association matrix in descending order; where B represents the product number, x represents the row of the product in the matrix, and y represents the column of the product in the matrix;

[0029] The convolution mapping algorithm is used to calculate the optimal correlation product matrix, and the convolution kernel size is set to 3*3 and the convolution layer is two layers;

[0030] The convolution kernel is the first layer convolution matrix K B (x, y, g ij ) performs polling calculations, and the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each polling:

[0031] Step 1041, obtaining the commodity association information at the center of the submatrix;

[0032] Step 1042, calculating the first average correlation degree of the commodities around the commodity at the central position;

[0033] Specifically, the first average correlation g′ of the commodities around the commodity at the central position is calculated using the following formula:

[0034]

[0035] Where r is the radius, the center point is the central position product, r = 1; g ij Indicates the correlation degree of each commodity in the small matrix;

[0036] Step 1043, assigning the average relevance to the central location commodity relevance;

[0037] Step 1044, the calculated center position commodity association information is sequentially saved to the second layer convolution matrix;

[0038] The calculated center position product information is saved in turn to the second layer convolution matrix K2 B (x, y, g ij ) in the corresponding position;

[0039] Convolution kernel for the second layer convolution matrix K2 B (x, y, g ij ) performs polling calculations, and the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each polling:

[0040] Step 1045, obtaining the commodity association information at the center of the submatrix;

[0041] Step 1046, calculating the second average correlation degree of the commodities around the commodity at the central position;

[0042] Specifically, the second average correlation g′ of the commodities around the commodity at the central position is calculated using the following formula:

[0043]

[0044] Among them, r is the radius, the center point is the central position product, r = 1; g′ ij Indicates the correlation degree of each commodity in the small matrix;

[0045] Step 1047, assigning the average relevance to the center position commodity relevance;

[0046] Step 1048, the calculated center position commodity association information is sequentially saved to the third layer convolution matrix;

[0047] The calculated center position product information is saved in turn to the third layer convolution matrix K3 B (x, y, g ij ) in the corresponding position.

[0048] Furthermore, after obtaining the third layer convolution matrix, it also includes:

[0049] The third-layer convolution matrix K3 of the current user is calculated by the interest formula B (x, y, g ij ) is calculated as follows:

[0050] X hj =g ij *p hj

[0051] Among them, X hj Indicates user h's response to K3 B (x, y, g ij ) is the interest level of a product j; p hj Indicates the number of user h's behavior records on product j. The number of behaviors refers to the total number of purchases, favorites, browsing, likes, and searches of the user on product j on the SaaS platform.

[0052] According to Xhj The value is large or small, and the top ranked products whose user interest ranking is within the threshold range are selected.

[0053] Furthermore, when the current user performs a purchase operation, the calculation of the correlation between the product i purchased by the current user and all other products is triggered and started.

[0054] On the other hand, the present invention also provides a system for building a user portrait based on a retail SaaS platform, the system comprising:

[0055] Acquisition module 201, used to acquire product information and user information;

[0056] A portrait module 202 is used to profile user attributes and behaviors and establish a user feature matrix;

[0057] A matrix building module 203, used to build a commodity feature matrix;

[0058] A calculation module 204 is used to calculate the commodity association and user interest, and generate a first recommendation list;

[0059] Comparison module 205, used to calculate the similarity between the feature value in the product feature matrix of each product in the first recommendation list whose user interest ranking is within the threshold range and the feature value in the user feature matrix of the current user;

[0060] The display module 206 is used to sort the products according to their similarity and select products whose similarity meets a threshold for display.

[0061] On the other hand, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 7 when running.

[0062] On the other hand, the present invention also provides an electronic device, comprising a memory and a processor, wherein the memory is used to store executable instructions; when the processor is used to run the executable instructions stored in the memory, the method described in any one of claims 1 to 7 is implemented.

[0063] The present invention provides a construction method, system, storage medium and electronic device based on user portrait of retail SaaS platform, obtains commodity information and user information, profiles user attribute behavior, establishes user feature matrix, establishes commodity feature matrix, calculates commodity association and user interest, generates a first recommendation list, calculates the similarity between the characteristic value in the commodity feature matrix of each commodity whose user interest ranking is within the threshold range and the characteristic value in the user feature matrix of the current user, sorts the commodities according to the similarity, and selects commodities whose similarity meets the threshold for display. The present invention performs multi-layer convolution operation on commodity association and then calculates user interest, closely associates commodity association with user interest, improves the accuracy of user interest commodity search, and effectively reduces the cost of user portrait generation without increasing hardware investment. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0065] Figure 1 It is a flowchart of the process of constructing the user portrait of the present invention;

[0066] Figure 2 It is a system structure diagram for constructing the user portrait of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0068] The present invention proposes a method, system and storage medium for constructing user portraits based on a retail SaaS platform.

[0069] Specifically, the method for constructing a user portrait based on a retail SaaS platform of the present invention is as shown in the flowchart attached. Figure 1 shown.

[0070] Specifically, the method comprises:

[0071] Step 101, obtain product information and user information; product information includes the product number, the gender of the user who purchased the product, the age of the user who purchased the product, the longitude and latitude of the product location, the product attribute type, etc. User information includes user identification number, user gender information, user age, longitude and latitude of the user's common address location, user purchase product information, user browsing product information, etc.

[0072] This step is information collection, which collects data on product information and user information in all operating processes in the SaaS retail platform.

[0073] Step 102, profile the user's attribute behavior and establish a user feature matrix;

[0074] Specifically, the user attributes are profiled to establish the user feature matrix Y S (a,b,c,d,e,f), where S represents the user's identity number, a represents the user's gender information, and in the process of program implementation, a=1 for males and a=0 for females are assigned. b represents the user's age, c represents the deviation of the longitude and latitude of the user's common address from the longitude and latitude of the center of Beijing (the center's longitude and latitude are 39°54′20″N, 116°25′29″E), d represents the attribute type number value of the most purchased product by the user, e represents the attribute type number value of the most collected product by the user, and f represents the user's characteristic value.

[0075] The user feature value f is calculated as follows:

[0076] f=w1*a+w2*b+w3*c+w4*d+w5*e

[0077] Among them, the value range of w1, w2, w3, w4, and w5 is 0 to 1, and w1 is initialized to 0.3, w2 = 0.5, w3 = 0.2, w4 = 0.2, and w5 = 0.2.

[0078] Step 103, establishing a commodity feature matrix;

[0079] Specifically, this step establishes the product feature matrix W B (a1, b1, c1, d1, e1, f1), where B is the product ID. a1 is the probability that the user who purchases the product is male, b1 is the average age of users who purchase the product, c1 is the deviation of the longitude and latitude of the product from the longitude and latitude of the center of Beijing, d1 is the product attribute type ID, e1 is the product attribute type ID with the highest similarity to itself, and f1 is the product feature value.

[0080] The user characteristic value f1 is calculated as follows:

[0081] f1=w6*a1+w7*b1+w8*c1+w9*d1+w10*e1

[0082] Among them, the value range of w6, w7, w8, w9, and w10 is 0 to 1, and w6 is initialized to 0.3, w7 is initialized to 0.5, w8 is initialized to 0.2, w9 is initialized to 0.2, and w10 is initialized to 0.2.

[0083] Step 104, calculating the commodity relevance and user interest, and generating a first recommendation list;

[0084] In this embodiment, this step calculates the commodity association and the user interest to form a preliminary first recommendation list. The recommendation list in this embodiment specifically selects the optimized top 9 as the recommendation list items, which can be recorded as top9 recommendations. It should be emphasized that the top9 here is the threshold selection of the preferred embodiment of the present invention, and the protection scope of the present invention is not limited to this.

[0085] When user h completes the purchase payment on the retail SaaS platform, the calculation of the correlation between the product i purchased by user h and all other products on the retail SaaS platform is triggered. The product correlation calculation formula is as follows:

[0086]

[0087] Among them, g ij represents the correlation between product i and product j, where T(i)∩T(j) represents the probability of purchasing both product i and product j in all purchase records on the SaaS platform. T(i) represents the probability of purchasing only product i in all purchase records on the SaaS platform. T(j) represents the probability of purchasing only product j in all purchase records on the SaaS platform.

[0088] In this embodiment, the top n items with the greatest relevance to the item i purchased by the user h are obtained by calculation. 2 Products are counted as Stopn 2 If not satisfied with n 2 The unsatisfactory parts are filled with null values. Establish the commodity association matrix K B (x, y, g ij ), the matrix is ​​n rows * n columns, where n is a power of 3 and n is greater than or equal to 27. 2 According to the relevance g ij The product association matrix is ​​loaded from large to small. Among them, B represents the product number, x is the row of the product in the matrix, y is the column of the product in the matrix, and g ij Represents the correlation between product i and product j.

[0089] The convolution mapping algorithm is used to calculate the optimal correlation product matrix. In this embodiment, the convolution kernel size is set to 3*3. The convolution kernel is an intelligent computing body that can execute the algorithm, and there are two convolution layers.

[0090] The first layer of the convolution mapping algorithm: the convolution kernel is the first layer convolution matrix K B (x, y, g ij ) for polling calculation, the matrix size is n rows * n columns, and the total number of products is n 2 The convolution kernel calculates the 3*3 column submatrix within the coverage range of each poll. The calculation formula is as follows:

[0091] Step 1041, obtaining the commodity association information at the center of the submatrix;

[0092] The sub-matrix can usually be selected as a 3*3 column sub-matrix.

[0093] Step 1042, calculating the first average correlation degree of the commodities around the commodity at the central position;

[0094] Specifically, the first average correlation g′ of the commodities around the commodity at the central position is calculated using the following formula:

[0095]

[0096] Among them, r is the radius, the center point is the central position product, and r = 1. ij Represents the correlation degree of each product in the small matrix.

[0097] Step 1043, assigning the average relevance to the central location commodity relevance;

[0098] Specifically, the first average correlation g′ is replaced by g of the central position product. ij .

[0099] Step 1044, the calculated center position commodity association information is sequentially saved to the second layer convolution matrix;

[0100] The calculated center position product information is saved in turn to the second layer convolution matrix K2 B (x, y, g ij ) in the corresponding position.

[0101] The second layer of the convolution mapping algorithm: convolution kernel for matrix K2 B (x, y, g ij ) for polling calculation, the matrix size is n / 3 rows * n / 3 columns, and the total number of products is n 2 / 9, the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each poll. The calculation formula is as follows:

[0102] Step 1045, obtaining the commodity association information at the center of the submatrix;

[0103] The sub-matrix can usually be selected as a 3*3 column sub-matrix.

[0104] Step 1046, calculating the second average correlation degree of the commodities around the commodity at the central position;

[0105] Specifically, the second average correlation g′ of the commodities around the commodity at the central position is calculated using the following formula:

[0106]

[0107] Among them, r is the radius, the center point is the central position product, and r = 1. g′ ij Represents the correlation degree of each product in the small matrix.

[0108] Step 1047, assigning the average relevance to the center position commodity relevance;

[0109] Specifically, the first average correlation g″ is replaced by g′ of the product at the center. ij .

[0110] Step 1048, the calculated center position commodity association information is sequentially saved to the third layer convolution matrix;

[0111] The calculated center position product information is saved in turn to the third layer convolution matrix K3 B (x, y, g ij ) in the corresponding position.

[0112] The third convolution matrix K3 obtained by two layers of convolution mapping calculation B (x, y, g ij ) is n / 9*n / 9 columns, and the total number is n 2 / 81 product matrix.

[0113] Calculate the third-layer convolution matrix K3 of user h through the interest formula B (x, y, g ij ) The calculation formula is as follows:

[0114] X hj =g ij *p hj

[0115] Among them, X hj Indicates user h's response to K3 B (x, y, g ij ) is the interest level of a product j; g ij Indicates the correlation between product i and product j stored in the product matrix.hj It indicates the number of times user h has recorded his / her behaviors on product j. The number of behaviors refers to the total number of purchases, favorites, browsing, likes, and searches of the user on product j in the SaaS platform.

[0116] According to X hj The value is large or small, and the top 9 commodities are selected and counted as the top 9 commodities whose user interest ranking is within the threshold range. The threshold is selected as 9, which is a preferred embodiment of the present invention. Here, the top 9 is stored in the first recommendation list.

[0117] Step 105, calculating the similarity between the feature value in the product feature matrix of each product in the first recommendation list whose user interest ranking is within the threshold range and the feature value in the user feature matrix of the current user;

[0118] For the top 9 products, calculate the matrix W for each product B The feature matrix Y of the product feature value f1 and the user h in (a1,b1,c1,d1,e1,f1) S The similarity sim between the user feature values ​​f in (a,b,c,d,e,f).

[0119] Step 106, sort the products according to their similarity, and select products whose similarity meets a threshold value for display.

[0120] Specifically, the top three commodities with similarity sim values ​​are selected or commodities with similarity values ​​greater than a preset threshold are selected, and the commodities are recommended to the current user h in the front-end display module.

[0121] In addition, the present invention also provides a construction system based on retail SaaS platform user portraits, as shown in the attached Figure 2 As shown, the system comprises:

[0122] Acquisition module 201, used to acquire product information and user information;

[0123] A portrait module 202 is used to profile user attributes and behaviors and establish a user feature matrix;

[0124] A matrix building module 203, used to build a commodity feature matrix;

[0125] A calculation module 204 is used to calculate the commodity association and user interest, and generate a first recommendation list;

[0126] Comparison module 205, used to calculate the similarity between the feature value in the product feature matrix of each product in the first recommendation list whose user interest ranking is within the threshold range and the feature value in the user feature matrix of the current user;

[0127] The display module 206 is used to sort the products according to their similarity and select products whose similarity meets a threshold for display.

[0128] This embodiment further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the method described in any one of the above technical solutions when running.

[0129] The computer-readable storage medium may include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the technical field.

[0130] This embodiment also provides an electronic device, which includes: a memory: used to store executable instructions; a processor: used to implement the method for constructing user portraits based on a retail SaaS platform of the present invention when running the executable instructions stored in the memory.

[0131] The present invention provides a construction method, system, storage medium and electronic device based on user portrait of retail SaaS platform, obtains commodity information and user information, profiles user attribute behavior, establishes user feature matrix, establishes commodity feature matrix, calculates commodity association and user interest, generates a first recommendation list, calculates the similarity between the characteristic value in the commodity feature matrix of each commodity whose user interest ranking is within the threshold range and the characteristic value in the user feature matrix of the current user, sorts the commodities according to the similarity, and selects commodities whose similarity meets the threshold for display. The present invention performs multi-layer convolution operation on commodity association and then calculates user interest, closely associates commodity association with user interest, improves the accuracy of user interest commodity search, and effectively reduces the cost of user portrait generation without increasing hardware investment.

[0132] It should be noted that, unless otherwise defined, the technical terms or scientific terms used herein shall have the usual meaning understood by persons with ordinary skills in the field to which the invention belongs. The words "first", "second" and similar words used in the invention specification and claims of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "one" do not indicate a quantity limitation, but indicate the existence of at least one. Words such as "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0133] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for constructing user portraits based on a retail SaaS platform, characterized in that: Step 101, obtaining product information and user information; Step 102, profile the user's attribute behavior and establish a user feature matrix; Step 103, establishing a commodity feature matrix; Step 104, calculating the commodity relevance and user interest, and generating a first recommendation list; The user interest calculation is specifically as follows: Establish the product association matrix K B (x,y,g ij ), sort the products according to the correlation g ij Load the product association matrix in descending order; where B represents the product number, x represents the row of the product in the matrix, and y represents the column of the product in the matrix; The convolution mapping algorithm is used to calculate the optimal correlation product matrix, and the convolution kernel size is set to 3*3 and the convolution layer is two layers; The convolution kernel is the first layer convolution matrix K B (x, y, g ij ) performs polling calculations, and the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each polling: Step 1041, obtaining the commodity association information at the center of the submatrix; Step 1042, calculating the first average correlation degree of the commodities around the commodity at the central position; Specifically, the first average correlation g′ of the commodities around the central commodity is calculated using the following formula: Where r is the radius, the center point is the central position product, r = 1; g ij Indicates the correlation degree of each commodity in the small matrix; Step 1043, assigning the first average relevance to the center position commodity relevance; Step 1044, the calculated center position commodity association information is sequentially saved to the second layer convolution matrix; The calculated center position product information is saved in the second layer convolution matrix K2 in sequence B (x,y,g ij ) in the corresponding position; Convolution kernel for the second layer convolution matrix K2 B (x,y,g ij ) performs polling calculation, and the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each polling; Step 1045, obtaining the commodity association information at the center of the submatrix; Step 1046, calculating the second average correlation degree of the commodities around the commodity at the central position; Specifically, the second average correlation g′ of the commodities around the commodity at the central position is calculated using the following formula: Among them, r is the radius, the center point is the central position product, r = 1; g′ ij Indicates the correlation degree of each commodity in the small matrix; Step 1047, assigning the second average relevance to the center position commodity relevance; Step 1048, the calculated center position commodity association information is sequentially saved to the third layer convolution matrix; The calculated center position product information is saved in the third layer convolution matrix K3 in sequence B (x,y,g ij ) in the corresponding position; After obtaining the third layer convolution matrix, it also includes: The third-layer convolution matrix K3 of the current user is calculated by the interest formula B (x, y, g ij ) is calculated as follows: X hj =g ij *p hj Among them, X hj Indicates user h's response to K3 B (x, y, g ij ) is the interest level of a product j; p hj Indicates the number of user h's behavior records on product j. The number of behaviors refers to the total number of purchases, favorites, browsing, likes, and searches of the user on product j on the SaaS platform. According to X hj Value size, select the top ranked products whose user interest ranking is within the threshold range; Step 105, calculating the similarity between the feature value in the product feature matrix of each product in the first recommendation list whose user interest ranking is within the threshold range and the feature value in the user feature matrix of the current user; Step 106, sort the products according to their similarity, and select products whose similarity meets a threshold value for display.

2. The method according to claim 1, characterized in that: The specific steps of establishing the user feature matrix are as follows: Establish user feature matrix Y S (a,b,c,d,e,f), where S represents the user's identity number, a represents the user's gender information, b represents the user's age, c represents the deviation of the longitude and latitude of the user's frequently used address from the longitude and latitude of the center of Beijing, d represents the attribute type number value of the most purchased product by the user, e represents the attribute type number value of the most collected product by the user, and f represents the user's characteristic value; The user characteristic value f is calculated as follows: f=w1*a+w2*b+w3*c+w4*d+w5*e Among them, the value range of w1, w2, w3, w4, and w5 is 0 to 1.

3. The method according to claim 1, characterized in that The specific steps of establishing the commodity feature matrix are as follows: Establish product feature matrix W B (a1,b1,c1,d1,e1,f1), Where B is the product ID, a1 is the probability that the user who purchases the product is male, b1 is the average age of users who purchase the product, c1 is the deviation of the longitude and latitude of the product location from the longitude and latitude of the center of Beijing, d1 is the product attribute type number, e1 is the product attribute type number with the highest similarity to itself, and f1 is the product feature value; The commodity characteristic value f1 is calculated as follows: f1=w6*a1+w7*b1+w8*c1+w9*d1+w10*e1 Among them, w6, w7, w8, w9, and w10 have values ​​ranging from 0 to 1.

4. The method according to claim 1, characterized in that: The calculation of commodity association is specifically as follows: Calculate the product correlation g between the product i purchased by the current user and all other products ij : Among them, g i j represents the correlation between product i and product j, where T(i)∩T(j) represents the probability of purchasing both product i and product j in all purchase records on the SaaS platform; T(i) represents the probability of purchasing only product i in all purchase records on the SaaS platform; T(j) represents the probability of purchasing only product j in all purchase records on the SaaS platform.

5. The method according to claim 1, characterized in that When the current user performs a purchase operation, the calculation of the correlation between the product i purchased by the current user and all other products is triggered and started.

6. A system for building user portraits based on a retail SaaS platform, characterized in that: The system comprises: Acquisition module 201, used to acquire product information and user information; A portrait module 202 is used to profile user attributes and behaviors and establish a user feature matrix; A matrix building module 203 is used to build a commodity feature matrix; A calculation module 204 is used to calculate the commodity association and user interest, and generate a first recommendation list; The user interest calculation is specifically as follows: Establish the product association matrix K B (x,y,g ij ), sort the products according to the correlation g ij Load the product association matrix in descending order; where B represents the product number, x represents the row of the product in the matrix, and y represents the column of the product in the matrix; The convolution mapping algorithm is used to calculate the optimal correlation product matrix, and the convolution kernel size is set to 3*3 and the convolution layer is two layers; The convolution kernel is the first layer convolution matrix K B (x,y,g ij ) performs polling calculations, and the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each polling: Step 1041, obtaining the commodity association information at the center of the submatrix; Step 1042, calculating the first average correlation degree of the commodities around the commodity at the central position; Specifically, the first average correlation g′ of the commodities around the central commodity is calculated using the following formula: Among them, r is the radius, the center point is the central position product, r = 1; g ij Indicates the correlation degree of each commodity in the small matrix; Step 1043, assigning the first average relevance to the center position commodity relevance; Step 1044, the calculated center position commodity association information is sequentially saved to the second layer convolution matrix; The calculated center position product information is saved in the second layer convolution matrix K2 in sequence B (x,y,g ij ) in the corresponding position; Convolution kernel for the second layer convolution matrix K2 B (x, y, g ij ) performs polling calculations, and the convolution kernel calculates the 3*3 column sub-matrix within the coverage range of each polling: Step 1045, obtaining the commodity association information at the center of the submatrix; Step 1046, calculating the second average correlation degree of the commodities around the commodity at the central position; Specifically, the second average correlation g′ of the commodities around the commodity at the central position is calculated using the following formula: Among them, r is the radius, the center point is the central position product, r = 1; g′ ij Indicates the correlation degree of each commodity in the small matrix; Step 1047, assigning the second average relevance to the center position commodity relevance; Step 1048, the calculated center position commodity association information is sequentially saved to the third layer convolution matrix; The calculated center position product information is saved in the third layer convolution matrix K3 in sequence B (x, y, g ij ) in the corresponding position; After obtaining the third layer convolution matrix, it also includes: The third-layer convolution matrix K3 of the current user is calculated by the interest formula B (x,y,g ij ) is calculated as follows: X hj =g ij *p hj Among them, X hj Indicates user h's response to K3 B (x, y, g ij ) is the interest level of a product j; p hj Indicates the number of user h's behavior records on product j. The number of behaviors refers to the total number of purchases, favorites, browsing, likes, and searches of the user on product j on the SaaS platform. According to X hj Value size, select the top ranked products whose user interest ranking is within the threshold range; Comparison module 205, used to calculate the similarity between the feature value in the product feature matrix of each product in the first recommendation list whose user interest ranking is within the threshold range and the feature value in the user feature matrix of the current user; The display module 206 is used to sort the products according to their similarity and select products whose similarity meets a threshold for display.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 5 when executed.

8. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein the memory is used to store executable instructions; when the processor is used to run the executable instructions stored in the memory, the method described in any one of claims 1 to 5 is implemented.

Images