Mobile advertisement effect analysis method and system based on user behaviors

Abstract

The invention relates to the technical field of advertisement data analysis, in particular to a mobile advertisement effect analysis method and system based on user behaviors. According to the method, detailed data of user behaviors and advertisement putting are captured in real time through an SDA interface and an API interface of a mobile advertisement platform, users are subdivided into user groups with different activeness, activeness and consumption levels by calculating activeness parameters, advertisement activeness parameters and consumption level parameters, accurate description and classification of the user behaviors are achieved, and the user experience is improved. The method helps advertisers to more accurately position target users, can mine potential user demands, improves the precision and conversion rate of advertisements, and reduces the cost of the advertisements at the same time. The advertisement putting effect is comprehensively evaluated by calculating key indexes such as the advertisement revenue index, the advertisement adaptation degree index and the shopping willingness index, a scientific decision basis is provided for advertisers, the advertisement putting strategy is optimized, and the advertisement effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of advertising data analysis, and particularly to a method and system for analyzing the effect of mobile advertising based on user behavior. Background Art

[0002] Today, with the rapid development of the mobile Internet, mobile advertising has become an important bridge connecting users with goods or services. However, with the intensification of market competition and the diversification of user needs, how to accurately push advertisements, improve the conversion rate of advertisements, and reduce advertising costs have become important challenges faced by advertisers. For this reason, a method and system for analyzing the effect of mobile advertising based on user behavior have emerged as the times require.

[0003] In the prior art, there have been various methods attempting to solve the problem of mobile advertising effect analysis. For example, in the patent document CN105677844A, a method for targeted pushing of mobile advertising big data and user cross-screen recognition is proposed. This method obtains the historical browsing and behavior data of multiple users when browsing web pages or Apps through a promotion and delivery server, and generates browsing habit and behavior interest data; when it detects that a user is browsing again in the advertising network according to the visitor ID of the user, combined with the historical interest classification and user demand tags of the user, it selects advertisement data that conforms to the browsing habit of the user for pushing. This method improves the accuracy and conversion rate of advertisements to a certain extent, but there are still some problems. For example, it may ignore the similarity between users, resulting in less personalized advertisement recommendations; at the same time, the processing and analysis of user behavior data may not be deep enough to fully explore the potential needs of users.

[0004] In order to overcome the deficiencies of the prior art, this paper proposes a method and system for analyzing the effect of mobile advertising based on user behavior. This method and system not only collect the basic information and behavior data of users, but also deeply explore the interest preferences and consumption habits of users through steps such as preprocessing and behavior data analysis. On this basis, artificial intelligence technologies such as collaborative filtering algorithms are used to achieve accurate advertisement pushing and effect evaluation. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems existing in the background art, and to propose a method and system for analyzing the effect of mobile advertising based on user behavior.

[0006] The purpose of the present invention can be achieved through the following technical solutions: In a first aspect, the present invention provides a method for analyzing the effect of mobile advertising based on user behavior, including the following steps: Step 1: Crawling user behavior data; Real-time crawling of relevant data on user behavior and advertisement delivery through the SDA interface and API interface of the mobile advertising platform application: Obtain user-related information, including user ID encoding, user gender encoding, user age encoding, and user IP encoding; The user identity encoding is UID; Among them, the user gender encoding is UGE, and a value of 1 represents female; a value of -1 represents male; a value of 0 represents not noted; Among them, the user age encoding is UAE, and a value of 0 represents 18 - 24 years old; a value of 1 represents 25 - 34 years old; a value of 2 represents 35 - 44 years old; a value of 3 represents 45 - 54 years old; a value of 4 represents 55 years old and above; Among them, the user IP encoding is UIE and is in the integer format after conversion of the IPv4 format; Obtain user behavior data, including the number of newly added items in the user's favorites in the recent month N11, the number of times the user clicks on advertisements in the recent month N12, the user browsing commodity page type encoding and the corresponding browsing time in the recent month; Among them, the browsing commodity page type is page_type, and page_type_1 represents sports goods; page_type_2 represents cosmetics; page_type_3 represents daily necessities; page_type_4 represents cultural goods; page_type_5 represents clothing goods; page_type_6 represents luxury goods; page_type_7 represents industrial production accessories; page_type_8 represents electronic products; page_type_9 represents maternal and child products; page_type_10 represents other goods.

[0007] The browsing time T includes the browsing time of sports goods Ttype_1, the browsing time of cosmetics Ttype_2, the browsing time of daily necessities Ttype_3, the browsing time of cultural goods Ttype_4, the browsing time of clothing goods Ttype_5, the browsing time of luxury goods Ttype_6, the browsing time of industrial production accessories Ttype_7, the browsing time of electronic products Ttype_8, the browsing time of maternal and child products Ttype_9, and the browsing time of other goods Ttype_10.

[0008] Advertisement display data: the number of advertisement displays received by the user N21, the number of times the user clicks on advertisements N22, and the number of times the user skips advertisements N23.

[0009] User consumption data: the total consumption of push commodities by the user in the recent month N31, the total consumption of purchased commodities N32, and the number of good reviews for push commodities N33.

[0010] Step 2: Preprocess the behavior data; Collect and record user behavior data, delete abnormal data that exceeds the maximum preset extreme value, merge duplicate records, and convert data in different formats into a unified structure.

[0011] Further, for each user, generate a uniquely corresponding behavior data matrix

[0012] Store the user list and its uniquely corresponding behavior data matrix in the database.

[0013] Step 3: Behavior data analysis; Through the formula Calculate the activity parameter E1 of each user. Among them, c1, c2, and c3 are preset weight coefficients, N11 and N12 respectively represent the number of newly added products in the favorites and the number of ad clicks; Ttype_p represents the browsing time of the user on various product pages in the recent month.

[0014] Mark users with activity parameter E1 higher than the preset maximum threshold E1Max as high-activity users; mark users with activity parameter E1 higher than the minimum threshold E1Min and lower than the maximum threshold E1Max as general-activity users; mark users with activity parameter E1 lower than the minimum threshold E1Min as low-activity users.

[0015] Through the formula Calculate the ad acceptance parameter E2 of each user, where w1 and w2 are preset influencing factors.

[0016] Mark users with ad acceptance parameter E2 higher than the preset maximum threshold E2Max as high-acceptance users; mark users with ad acceptance parameter E2 higher than the minimum threshold E2Min and lower than the maximum threshold E2Max as general-acceptance users; mark users with ad acceptance parameter lower than the minimum threshold E2Min as low-acceptance users.

[0017] Calculate the consumption level parameter E3 of each user through the formula, where w3, w4, and w5 are preset influencing factors.

[0018] Mark users with consumption level parameter E3 higher than the preset maximum threshold E3Max as high-consumption-level users; mark users with consumption level parameter E3 higher than the minimum threshold E3Min and lower than the maximum threshold E3Max as general-consumption-level users; mark users with consumption level parameter lower than the minimum threshold E3Min as low-consumption-level users.

[0019] Step 4: Similar behavior user marking and push optimization: Extract the user portraits of each user by analyzing the user gender code, user age code, user IP code, the types of product pages browsed by the user in the past month, and the corresponding browsing time T encoded by different user IDs. Use the decision tree machine learning algorithm to cluster users, and divide users into different interest groups through K-means clustering. Extract the user behavior data matrix H of all users in each interest group, and calculate the average matrix of all data matrices H as the behavior centroid of the interest group.

[0020] For each interest group, use the collaborative filtering algorithm to recommend advertisements based on the behavioral similarity between users.

[0021] The basic assumption of collaborative filtering is that if two users are similar in their past browsing behaviors and shopping habits, they may also be similar in their future browsing behaviors and shopping habits.

[0022] The weighted Euclidean distance formula for collaborative filtering is: , where i and j are the user ID codes of different users and i≠j. Among them, α(i, j) is the intimacy parameter between user i and j, and r1, r2, and r3 are preset weight factors. Among them, p1 is the number of common friends between user i and j, p2 is the number of the same products collected by user i and j, and p3 is the number of product page sharing and forwarding between user i and j. Among them, λ1, λ2, λ3, and λ4 are preset circle factors, kp = k1, k2,..., k10 is a set of preset browsing time influence coefficients.

[0023] Mark the user ID codes i and j of users with the similarity parameter COV(i, j) greater than the preset threshold as highly similar users.

[0024] For highly similar users i and j, whenever one of the users i purchases a certain product, push the same type of product to j.

[0025] Step Five: Advertising Effect Analysis; Obtain the total advertising cost B1, the average traffic revenue B2 obtained through the click-through browsing of all users, the total number of advertisements B3, and the total profit B4 of the advertising push platform. Through the formula Calculate the advertising revenue index Q1, advertising adaptation index Q2, and shopping willingness index Q3 of each user.

[0026] Extract the activity levels of each user, including low activity, medium activity, and high activity. Extract the acceptance levels of each user, including low acceptance, medium acceptance, and high acceptance. Extract the consumption levels of each user, including low consumption, medium consumption, and high consumption. Construct a three-dimensional classification interval, including 27 grouping areas, and each grouping area corresponds to a set of preset feature vectors (U1, U2, U3).

[0027] Calculate the advertising placement evaluation vectors (CVR1, CVR2, CVR3) of each user through the formula (CVR1, CVR2, CVR3) = (U1×Q1, U2×Q2, U3×Q3).

[0028] When CVR1 is greater than the preset threshold, it is determined that the click and purchase benefits obtained by advertising to this user are relatively high; otherwise, it is determined that the click and purchase benefits obtained by advertising to this user are relatively low; When CVR2 is greater than the preset threshold, it is determined that the user's interest in the advertised placement is relatively low; otherwise, it is determined that the user's interest in the advertised placement is relatively high; When CVR3 is greater than the preset threshold, it is determined that the user's willingness to shop through advertisement push is relatively high; otherwise, it is determined that the user's willingness to shop through advertisement push is relatively low.

[0029] In a second aspect, the present invention provides a mobile advertising effect analysis system based on user behavior, including a data scraping module, a data preprocessing module, a user analysis module, and an advertisement optimization and analysis module.

[0030] The data scraping module real-time scrapes relevant data of user behavior and advertisement placement through the SDA interface and API interface of the mobile advertising platform application: Obtain user-related information, including user ID coding, user gender coding, user age coding, and user IP coding; Where the user identity coding is UID; Where the user gender coding is UGE and the value of 1 represents female; the value of -1 represents male; the value of 0 represents not noted; Where the user age coding is UAE and the value of 0 represents 18 - 24 years old; the value of 1 represents 25 - 34 years old; the value of 2 represents 35 - 44 years old; the value of 3 represents 45 - 54 years old; the value of 4 represents 55 years old and above; Where the user IP coding is UIE and is in the integer format after conversion of the IPv4 format; Obtain user behavior data, including the number of newly added items in the user's favorites in the past month N11, the number of times the user clicks on advertisements in the past month N12, the type coding of the user's commodity browsing pages in the past month, and the corresponding browsing time; Among them, the browsing product page type is page_type, and page_type_1 represents sports products; page_type_2 represents cosmetics; page_type_3 represents daily necessities; page_type_4 represents cultural products; page_type_5 represents clothing products; page_type_6 represents luxury goods; page_type_7 represents industrial production accessories; page_type_8 represents electronic products; page_type_9 represents mother and baby products; page_type_10 represents other products.

[0031] The browsing time T includes the browsing time Ttype_1 of sports products, the browsing time Ttype_2 of cosmetics, the browsing time Ttype_3 of daily necessities, the browsing time Ttype_4 of cultural products, the browsing time Ttype_5 of clothing products, the browsing time Ttype_6 of luxury goods, the browsing time Ttype_7 of industrial production accessories, the browsing time Ttype_8 of electronic products, the browsing time Ttype_9 of mother and baby products, and the browsing time Ttype_10 of other products.

[0032] Advertising display data: the number of advertising displays N21 received by the user, the number of user clicks on the advertisement N22, and the number of user skips of the advertisement N23.

[0033] User consumption data: the total consumption N31 of the pushed products by the user in the recent month, the total consumption N32 of the purchased products, and the number of positive reviews N33 of the pushed products.

[0034] The data preprocessing module collects and records the user behavior data, deletes the abnormal data exceeding the maximum preset extreme value, merges the duplicate records, and converts the data in different formats into a unified structure.

[0035] Furthermore, for each user, a unique corresponding behavior data matrix is generated The user list and its unique corresponding behavior data matrix are stored in the database.

[0036] The user analysis module uses the formula to calculate the activity parameter E1 of each user. Among them, c1, c2, and c3 are preset weight coefficients, N11 and N12 respectively represent the number of newly added products in the favorites and the number of advertisement clicks; Ttype_p represents the browsing time of the user on various product pages in the recent month.

[0037] The users with the activity parameter E1 higher than the preset maximum threshold E1Max are marked as high-activity users; the users with the activity parameter E1 higher than the minimum threshold E1Min and lower than the maximum threshold E1Max are marked as general-activity users; the users with the activity parameter E1 lower than the minimum threshold E1Min are marked as low-activity users.

[0038] Calculate the advertisement acceptance parameter E2 for each user through a formula, where w1 and w2 are preset influence factors.

[0039] Mark users with an advertisement acceptance parameter E2 higher than the preset maximum threshold E2Max as high-acceptance users; mark users with an advertisement acceptance parameter E2 higher than the minimum threshold E2Min and lower than the maximum threshold E2Max as medium-acceptance users; mark users with an advertisement acceptance parameter lower than the minimum threshold E2Min as low-acceptance users.

[0040] Through the formula Calculate the consumption level parameter E3 for each user, where w3, w4, and w5 are preset influence factors.

[0041] Mark users with a consumption level parameter E3 higher than the preset maximum threshold E3Max as high-consumption-level users; mark users with a consumption level parameter E3 higher than the minimum threshold E3Min and lower than the maximum threshold E3Max as medium-consumption-level users; mark users with a consumption level parameter lower than the minimum threshold E3Min as low-consumption-level users.

[0042] Obtain the total advertisement placement cost B1, the average traffic revenue B2 obtained through the click-through browsing of all users, the total number of advertisement placements B3, and the total profit B4 of the advertisement push platform. Through the formula Calculate the advertisement revenue index Q1, the advertisement adaptability index Q2, and the shopping willingness index Q3 for each user.

[0043] Extract the activity levels of each user, including low activity, medium activity, and high activity. Extract the acceptance levels of each user, including low acceptance, medium acceptance, and high acceptance. Extract the consumption levels of each user, including low consumption, medium consumption, and high consumption. Construct a three-dimensional classification interval, including 27 grouping areas, and each grouping area corresponds to a set of preset feature vectors (U1, U2, U3).

[0044] Calculate the advertisement placement evaluation vector (CVR1, CVR2, CVR3) for each user through the formula (CVR1, CVR2, CVR3) = (U1 × Q1, U2 × Q2, U3 × Q3).

[0045] When CVR1 is greater than the preset threshold, it is determined that the click and purchase revenues obtained from advertising to this user are high; otherwise, it is determined that the click and purchase revenues obtained from advertising to this user are low; When CVR2 is greater than the preset threshold, it is determined that the user's interest in the placed advertisement is low; otherwise, it is determined that the user's interest in the placed advertisement is high; When CVR3 is greater than a preset threshold, it is determined that the user has a relatively high willingness to shop through advertisement push; otherwise, it is determined that the user has a relatively low willingness to shop through advertisement push.

[0046] The advertisement optimization and analysis module extracts the user portraits of each user by extracting the user gender code, user age code, user IP code, the type code of the user's browsed product pages in the recent month, and the corresponding browsing time T encoded by different user IDs. The specific process is as follows: Use the decision tree machine learning algorithm to cluster users, and divide users into different interest groups through K-means clustering. Extract the user behavior data matrix H of all users in each interest group, and calculate the average matrix of all data matrices H as the behavior centroid of this interest group.

[0047] For each interest group, use the collaborative filtering algorithm to recommend advertisements based on the behavior similarity between users.

[0048] The basic assumption of collaborative filtering is that if two users are similar in their past browsing behaviors and shopping habits, they may also be similar in their future browsing behaviors and shopping habits.

[0049] The weighted Euclidean distance formula of collaborative filtering is: , where i and j are the user ID codes of different users and i≠j. Among them, α(i, j) is the intimacy parameter between user i and j, and r1, r2, and r3 are preset weight factors. Among them, p1 is the number of common friends between user i and j, p2 is the number of the same products collected by user i and j, and p3 is the number of sharing and forwarding of product pages between user i and j. Among them, λ1, λ2, λ3, and λ4 are preset circle factors, kp = k1, k2,..., k10 are a set of preset browsing time influence coefficients.

[0050] Mark the user ID codes i and j of users with a similarity parameter COV(i, j) greater than the preset threshold as highly similar users.

[0051] For highly similar users i and j, whenever one of the users i purchases a certain product, the same type of product will be pushed to j.

[0052] Compared with the prior art, the beneficial effects of the present invention are: 1. Through the SDA interface and API interface of the mobile advertising platform, the present invention captures in real time the detailed data of user behavior and advertisement delivery, including user basic information, user behavior data, advertisement display data, etc., providing a rich and accurate data basis for subsequent precise analysis; by calculating activity parameters, advertisement acceptance parameters, and consumption level parameters, users are segmented into user groups with different activity levels, acceptance levels, and consumption levels, achieving precise characterization and classification of user behavior, which helps advertisers more accurately target users.

[0053] 2. The present invention uses a collaborative filtering algorithm to recommend advertisements based on the behavioral similarity between users, which can discover potential user needs, improve the accuracy and conversion rate of advertisements, and reduce advertising costs at the same time.

[0054] 3. By calculating key indicators such as advertisement revenue index, advertisement adaptation index, and shopping intention index, the present invention comprehensively evaluates the effect of advertisement delivery, provides a scientific decision-making basis for advertisers, helps optimize advertisement delivery strategies, and improves advertisement effects. Through precise advertisement recommendations and optimized advertisement delivery strategies, it is possible to reduce users' resistance to irrelevant advertisements, improve users' acceptance and satisfaction of advertisements, and thus enhance the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the accompanying drawings: Figure 1 is the method flow chart of the present invention; Figure 2 is the three-dimensional classification interval schematic diagram of the present invention; Figure 3 is the system block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0057] Please refer to Figure 1 as shown, a method for analyzing the effect of mobile advertisements based on user behavior includes the following steps: Step 1. Capturing user behavior data; Capture in real time the relevant data of user behavior and advertisement delivery through the SDA interface and API interface applied by the mobile advertising platform: Obtain user-related information, including user ID code, user gender code, user age code, and user IP code; Among them, the user identity code is UID; Among them, the user gender code is UGE, and a value of 1 represents female; a value of -1 represents male; a value of 0 represents not noted; Among them, the user age code is UAE, and a value of 0 represents 18 - 24 years old; a value of 1 represents 25 - 34 years old; a value of 2 represents 35 - 44 years old; a value of 3 represents 45 - 54 years old; a value of 4 represents 55 years old and above; Among them, the user IP code is UIE and is in the integer format after conversion of the IPv4 format; Obtain user behavior data, including the number of newly added items in the user's favorites in the recent month N11, the number of times of clicking on advertisements in the recent month N12, the browsing time corresponding to the browsing commodity page type code of the user in the recent month; Among them, the browsing commodity page type is page_type, and page_type_1 represents sports goods; page_type_2 represents cosmetics; page_type_3 represents daily necessities; page_type_4 represents cultural goods; page_type_5 represents clothing goods; page_type_6 represents luxury goods; page_type_7 represents industrial production accessories; page_type_8 represents electronic products; page_type_9 represents maternal and child products; page_type_10 represents other goods.

[0058] The browsing time T includes the browsing time of sports goods Ttype_1, the browsing time of cosmetics Ttype_2, the browsing time of daily necessities Ttype_3, the browsing time of cultural goods Ttype_4, the browsing time of clothing goods Ttype_5, the browsing time of luxury goods Ttype_6, the browsing time of industrial production accessories Ttype_7, the browsing time of electronic products Ttype_8, the browsing time of maternal and child products Ttype_9, and the browsing time of other goods Ttype_10.

[0059] Advertisement display data: the number of advertisement displays received by the user N21, the number of times the user clicks on advertisements N22, and the number of times the user skips advertisements N23.

[0060] User consumption data: the total consumption of the pushed goods by the user in the recent month N31, the total consumption of the purchased goods N32, and the number of good reviews for the pushed goods N33.

[0061] Step Two: Pre - process the behavior data; Collect and record user behavior data, delete abnormal data that exceeds the maximum preset extreme value, merge duplicate records, and convert data in different formats into a unified structure.

[0062] Furthermore, for each user, generate a uniquely corresponding behavior data matrix

[0063] Store the user list and its uniquely corresponding behavior data matrix into the database.

[0064] Step 3: Behavior data analysis; Through the formula Calculate the activity parameter E1 of each user. Among them, c1, c2, and c3 are preset weight coefficients, N11 and N12 respectively represent the number of newly added products in the favorites and the number of advertisement clicks; Ttype_p represents the browsing time of users on various product pages in the recent month.

[0065] Mark users with activity parameter E1 higher than the preset maximum threshold E1Max as high-activity users; mark users with activity parameter E1 higher than the minimum threshold E1Min and lower than the maximum threshold E1Max as general-activity users; mark users with activity parameter E1 lower than the minimum threshold E1Min as low-activity users.

[0066] Through the formula Calculate the advertisement acceptance parameter E2 of each user, where w1 and w2 are preset influencing factors.

[0067] Mark users with advertisement acceptance parameter E2 higher than the preset maximum threshold E2Max as high-acceptance users; mark users with advertisement acceptance parameter E2 higher than the minimum threshold E2Min and lower than the maximum threshold E2Max as general-acceptance users; mark users with advertisement acceptance parameter lower than the minimum threshold E2Min as low-acceptance users.

[0068] Through the formula Calculate the consumption level parameter E3 of each user, where w3, w4, and w5 are preset influencing factors.

[0069] Mark users with consumption level parameter E3 higher than the preset maximum threshold E3Max as high-consumption-level users; mark users with consumption level parameter E3 higher than the minimum threshold E3Min and lower than the maximum threshold E3Max as general-consumption-level users; mark users with consumption level parameter lower than the minimum threshold E3Min as low-consumption-level users.

[0070] Step 4: Similar behavior user marking and push optimization: For each interest group, use the collaborative filtering algorithm to perform advertisement recommendations based on the behavior similarity between users.

[0071] The basic assumption of collaborative filtering is that if two users are similar in their past browsing behaviors and shopping habits, then they may also be similar in their future browsing behaviors and shopping habits.

[0072] The weighted Euclidean distance formula for collaborative filtering is as follows: , where i and j are the user ID codes of different users and i ≠ j. Among them, α(i, j) is the intimacy parameter between user i and j, and r1, r2, and r3 are preset weight factors. Among them, p1 is the number of common friends between user i and j, p2 is the number of the same products collected by user i and j, and p3 is the number of sharing and forwarding of product pages between user i and j. Among them, λ1, λ2, λ3, and λ4 are preset circle factors, kp = k1, k2,..., k10 is a set of preset browsing time influence coefficients.

[0073] Mark the user ID codes i and j of users with the similarity parameter COV(i, j) greater than the preset threshold as highly similar users.

[0074] For highly similar users i and j, whenever one of the users i purchases a certain product, push the same type of product to j.

[0075] Step Five: Analysis of advertising effects; Obtain the total advertising cost B1, the average traffic revenue B2 obtained through the click and browse of all users, the total number of advertising placements B3, and the total profit B4 of the advertising push platform. Through the formula Calculate the advertising revenue index Q1, the advertising adaptation index Q2, and the shopping willingness index Q3 of each user.

[0076] Please refer to Figure 2 As shown, extract the activity levels of each user, including low activity, general activity, and high activity. Extract the acceptance levels of each user, including low acceptance, general acceptance, and high acceptance. Extract the consumption levels of each user, including low consumption, general consumption, and high consumption. Construct a three-dimensional classification interval, including 27 grouped areas, and each grouped area corresponds to a set of preset feature vectors (U1, U2, U3).

[0077] Calculate the advertising placement evaluation vector (CVR1, CVR2, CVR3) of each user through the formula (CVR1, CVR2, CVR3) = (U1 × Q1, U2 × Q2, U3 × Q3).

[0078] When CVR1 is greater than the preset threshold, it is determined that the click and purchase revenues obtained from advertising this user are high; otherwise, it is determined that the click and purchase revenues obtained from advertising this user are low; When CVR2 is greater than the preset threshold, it is determined that the user's interest in the placed advertisement is low; otherwise, it is determined that the user's interest in the placed advertisement is high; When CVR3 is greater than the preset threshold, it is determined that the user's willingness to shop through advertising push is high; otherwise, it is determined that the user's willingness to shop through advertising push is low.

[0079] Please refer to Figure 3 As shown, a mobile advertising effect analysis system based on user behavior includes a data scraping module, a data preprocessing module, a user analysis module, and an advertising optimization and analysis module.

[0080] The data scraping module real-time scrapes relevant data on user behavior and advertising placement through the SDA interface and API interface of the mobile advertising platform application: Obtain user-related information, including user ID encoding, user gender encoding, user age encoding, and user IP encoding; Among them, the user identity encoding is UID; Among them, the user gender encoding is UGE and the value of 1 represents female; the value of -1 represents male; the value of 0 represents not noted; Among them, the user age encoding is UAE and the value of 0 represents 18 - 24 years old; the value of 1 represents 25 - 34 years old; the value of 2 represents 35 - 44 years old; the value of 3 represents 45 - 54 years old; the value of 4 represents 55 years old and above; Among them, the user IP encoding is UIE and it is the integer format after conversion of the IPv4 format; Obtain user behavior data, including the number of newly added items in the user's favorites in the recent month N11, the number of times the user clicks on advertisements in the recent month N12, the browsing time of the user on different types of product pages in the recent month, and the corresponding browsing time; Among them, the types of product pages are page_type, where page_type_1 represents sports products; page_type_2 represents cosmetics; page_type_3 represents daily necessities; page_type_4 represents cultural products; page_type_5 represents clothing products; page_type_6 represents luxury goods; page_type_7 represents industrial production accessories; page_type_8 represents electronic products; page_type_9 represents maternal and child products; page_type_10 represents other products.

[0081] The browsing time T includes the browsing time of sports products Ttype_1, the browsing time of cosmetics Ttype_2, the browsing time of daily necessities Ttype_3, the browsing time of cultural products Ttype_4, the browsing time of clothing products Ttype_5, the browsing time of luxury goods Ttype_6, the browsing time of industrial production accessories Ttype_7, the browsing time of electronic products Ttype_8, the browsing time of maternal and child products Ttype_9, and the browsing time of other products Ttype_10.

[0082] Advertising display data: the number of times the user receives an advertisement display N21, the number of times the user clicks on an advertisement N22, and the number of times the user skips an advertisement N23.

[0083] User consumption data: The total consumption N31 of the pushed products, the total consumption N32 of the purchased products, and the number of positive reviews N33 of the pushed products by the user in the recent month.

[0084] The data preprocessing module collects and records user behavior data, deletes abnormal data exceeding the maximum preset extreme value, merges duplicate records, and converts data in different formats into a unified structure.

[0085] Furthermore, for each user, a unique corresponding behavior data matrix is generated The user list and its unique corresponding behavior data matrix are stored in the database.

[0086] The user analysis module calculates the activity parameter E1 of each user through the formula where c1, c2, and c3 are preset weight coefficients, N11 and N12 respectively represent the number of newly added products in the favorites and the number of ad clicks; Ttype_p represents the browsing time of the user on various product pages in the recent month.

[0087] Users with the activity parameter E1 higher than the preset maximum threshold E1Max are marked as high-activity users; users with the activity parameter E1 higher than the minimum threshold E1Min and lower than the maximum threshold E1Max are marked as general-activity users; users with the activity parameter E1 lower than the minimum threshold E1Min are marked as low-activity users.

[0088] Calculate the advertisement acceptance parameter E2 of each user through the formula where w1 and w2 are preset influence factors.

[0089] Users with the advertisement acceptance parameter E2 higher than the preset maximum threshold E2Max are marked as high-acceptance users; users with the advertisement acceptance parameter E2 higher than the minimum threshold E2Min and lower than the maximum threshold E2Max are marked as general-acceptance users; users with the advertisement acceptance parameter lower than the minimum threshold E2Min are marked as low-acceptance users.

[0090] Calculate the consumption level parameter E3 of each user through the formula where w3, w4, and w5 are preset influence factors.

[0091] Users with the consumption level parameter E3 higher than the preset maximum threshold E3Max are marked as high-consumption-level users; users with the consumption level parameter E3 higher than the minimum threshold E3Min and lower than the maximum threshold E3Max are marked as general-consumption-level users; users with the consumption level parameter lower than the minimum threshold E3Min are marked as low-consumption-level users.

[0092] Obtain the total cost of advertising placement B1, the average traffic revenue B2 obtained through the clicks and views of all users, the total number of advertising placements B3, and the total profit B4 of the advertising push platform. Through the formula Calculate the advertising revenue index Q1, the advertising adaptation index Q2, and the shopping intention index Q3 for each user.

[0093] Extract the activity levels of each user, including low activity, medium activity, and high activity. Extract the acceptance levels of each user, including low acceptance, medium acceptance, and high acceptance. Extract the consumption levels of each user, including low consumption, medium consumption, and high consumption. Construct a three-dimensional classification interval, including 27 grouping areas, and each grouping area corresponds to a set of preset feature vectors (U1, U2, U3).

[0094] Calculate the advertising placement evaluation vector (CVR1, CVR2, CVR3) for each user through the formula (CVR1, CVR2, CVR3) = (U1×Q1, U2×Q2, U3×Q3).

[0095] When CVR1 is greater than the preset threshold, it is determined that the click and purchase revenues obtained from advertising this user are relatively high; otherwise, it is determined that the click and purchase revenues obtained from advertising this user are relatively low; When CVR2 is greater than the preset threshold, it is determined that the user has a relatively low interest in the placed advertisement; otherwise, it is determined that the user has a relatively high interest in the placed advertisement; When CVR3 is greater than the preset threshold, it is determined that the user has a relatively high willingness to shop through the advertisement push; otherwise, it is determined that the user has a relatively low willingness to shop through the advertisement push.

[0096] For each interest group, the advertisement optimization and analysis module uses the collaborative filtering algorithm to perform advertisement recommendations based on the behavioral similarity between users.

[0097] The basic assumption of collaborative filtering is that if two users are similar in their past browsing behaviors and shopping habits, they may also be similar in their future browsing behaviors and shopping habits.

[0098] The weighted Euclidean distance formula for collaborative filtering is: , where i and j are the user ID codes of different users and i≠j. Among them, α(i, j) is the intimacy parameter between user i and j, and r1, r2, and r3 are preset weight factors. Among them, p1 is the number of common friends between user i and j, p2 is the number of the same products collected by user i and j, and p3 is the number of sharing and forwarding of product pages between user i and j. Among them, λ1, λ2, λ3, and λ4 are preset circle factors, kp = k1, k2,..., k10 is a set of preset browsing time influence coefficients.

[0099] Encode the user IDs i and j of users with a similarity parameter COV(i, j) greater than a preset threshold as highly similar users.

[0100] For highly similar users i and j, whenever one of the users i purchases a certain product, push the same type of product to j.

[0101] It should be understood that the terms "comprising" and "including" used in the specification and claims of this disclosure indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0102] It should also be understood that the terms used in this disclosure specification are merely for the purpose of describing specific embodiments and are not intended to limit this disclosure. As used in this disclosure specification and claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms. It should further be understood that the term "and / or" used in this disclosure specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations; The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to only the specific embodiments. Obviously, many modifications and variations can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A method for analyzing the effect of mobile advertising based on user behavior, characterized in that: The following steps are involved: Step 1: Capture user behavior data. Use the SDA interface and API interface of the mobile advertising platform application to capture user behavior and advertising delivery related data in real time, including user ID code, user gender code, user age code, user IP code, the number of new items added to the user's favorites in the past month, the number of ad clicks in the past month, the type code of the product page browsed by the user in the past month and the corresponding browsing time, ad display data, and user consumption data; Step 2: Behavior data preprocessing: collect and record user behavior data, delete abnormal data, merge duplicate records, convert data in different formats into a unified structure, and generate a unique corresponding behavior data matrix for each user and store it in the database; Step 3: Behavioral data analysis: Calculate the activity parameters, advertising acceptance parameters and consumption level parameters of each user through formulas, and mark the users as high, average and low activity users, high, average and low acceptance users, and high, average and low consumption level users according to preset thresholds; Step 4: Tagging and optimizing push notifications for users with similar behaviors. Use collaborative filtering algorithms to recommend ads based on the behavioral similarity between users. Users with similarity parameters greater than a preset threshold are marked as high-similarity users, and product push notifications are optimized for high-similarity users. Step 5: Advertising effectiveness analysis, obtain the relevant cost, revenue and quantity data of advertising, calculate the advertising revenue index, advertising suitability index and shopping intention index of each user, and construct a three-dimensional classification interval based on the user's activity level, acceptance level and consumption level. Calculate the advertising evaluation vector of each user through the formula, and judge the advertising effect based on the advertising evaluation vector.

2. A method for analyzing mobile advertising effects based on user behavior according to claim 1, characterized in that: The data related to user behavior and advertising delivery described in step 1 include: The user identity is encoded as UID; The user gender is coded as UGE, and the UGE value is 1 for female, -1 for male, and 0 for unremarked. The user age code is UAE and the value 0 represents 18-24 years old; the value 1 represents 25-34 years old; the value 2 represents 35-44 years old; the value 3 represents 45-54 years old; the value 4 represents 55 years old and above; User IP code UIE is in integer format after conversion to IPv4 format; User behavior data, including the number of new items added to the user's favorites in the past month N11, the number of advertisements clicked in the past month N12, and the type codes of the product pages browsed by users in the past month and the corresponding browsing time; The browsing product page type is page_type and page_type_1 represents sports products; page_type_2 represents cosmetics; page_type_3 represents daily necessities; page_type_4 represents cultural products; page_type_5 represents clothing products; page_type_6 represents luxury goods; page_type_7 represents industrial production accessories; page_type_8 represents electronic products; page_type_9 represents maternal and child products; page_type_10 represents other products; Browsing time T includes sports goods browsing time Ttype_1, cosmetics browsing time Ttype_2, daily necessities browsing time Ttype_3, cultural goods browsing time Ttype_4, clothing goods browsing time Ttype_5, luxury goods browsing time Ttype_6, industrial production accessories browsing time Ttype_7, electronic products browsing time Ttype_8, ​​maternal and child products browsing time Ttype_9, and other goods browsing time Ttype_10; Ad display data: the number of ad displays received by the user N21, the number of ad clicks by the user N22, and the number of ad skips by the user N23; User consumption data: the user’s total consumption of pushed products in the past month N31, total consumption of purchased products N32, and the number of positive reviews for pushed products N33.

3. The method for analyzing the effect of mobile advertising based on user behavior according to claim 1, characterized in that: The behavior data preprocessing described in step 2 also includes deleting abnormal data that exceeds a maximum preset extreme value.

4. The method for analyzing the effect of mobile advertising based on user behavior according to claim 1, characterized in that: The specific process of behavioral data analysis and user classification described in step 3 is as follows: By formula Calculate the activity parameter E1 of each user; where c1, c2 and c3 are preset weight coefficients, N11 and N12 represent the number of new items added to favorites and the number of ad clicks respectively; Ttype_p represents the browsing time of users on various product pages in the past month; Users whose activity parameter E1 is higher than the preset maximum threshold E1Max are marked as highly active users; users whose activity parameter E1 is higher than the minimum threshold E1Min and lower than the maximum threshold E1Max are marked as generally active users; users whose activity parameter E1 is lower than the minimum threshold E1Min are marked as low active users; By formula Calculate the advertisement acceptance parameter E2 of each user, where w1 and w2 are preset influencing factors; Mark users whose advertising acceptance parameter E2 is higher than the preset maximum threshold E2Max as high acceptance users; mark users whose advertising acceptance parameter E2 is higher than the minimum threshold E2Min and lower than the maximum threshold E2Max as average acceptance users; mark users whose advertising acceptance parameter is lower than the minimum threshold E2Min as low acceptance users; By formula Calculate the consumption level parameter E3 of each user, where w3, w4 and w5 are preset influencing factors; Users whose consumption level parameter E3 is higher than the preset maximum threshold E3Max are marked as high consumption level users; users whose consumption level parameter E3 is higher than the minimum threshold E3Min and lower than the maximum threshold E3Max are marked as general consumption level users; users whose consumption level parameter is lower than the minimum threshold E3Min are marked as low consumption level users.

5. The method for analyzing the effect of mobile advertising based on user behavior according to claim 1, characterized in that: The specific process of similar behavior user tagging and push optimization is as follows: By extracting the user gender code, user age code, user IP code, and the type code of the product page browsed by the user in the past month and the corresponding browsing time T from different user ID codes, the user portrait of each user is extracted, and the users are grouped using a decision tree-based machine learning algorithm. The users are divided into different interest groups through K-means clustering; the user behavior data matrix H of all users in each interest group is extracted, and the average value matrix of all data matrices H is calculated as the behavior centroid of the interest group; Use collaborative filtering algorithms to recommend ads based on the similarity of user behaviors; The weighted Euclidean distance formula for collaborative filtering is: Where i and j are user ID codes of different users and i≠j; where α(i, j) is the intimacy parameter of users i and j and r1, r2 and r3 are preset weight factors, where p1 is the number of common friends of users i and j, p2 is the number of the same products collected by users i and j, and p3 is the number of product page sharing and forwarding between users i and j; where λ1, λ2, λ3 and λ4 are preset circle factors, and kp=k1, k2, ..., k10 is a set of preset browsing time influence coefficients; Mark the user ID codes i and j of users whose similarity parameter COV (i, j) is greater than a preset threshold as high-similarity users; For users i and j with high similarity, whenever user i purchases a product, the same type of product will be pushed to j.

6. The method for analyzing the effect of mobile advertising based on user behavior according to claim 1, characterized in that: The specific process of calculating the advertising revenue index, advertising suitability index and shopping intention index is as follows: Get the total cost of advertising B1, the average traffic revenue obtained through all users' clicks and browsing B2, the total number of advertisements B3 and the total profit of the advertising push platform B4; through the formula Calculate each user's advertising revenue index Q1, advertising suitability index Q2, and shopping intention index Q3; The advertising revenue index, advertising suitability index and shopping intention index are weightedly calculated according to the characteristic vector of the three-dimensional classification interval.

7. The method for analyzing the effect of mobile advertising based on user behavior according to claim 6, characterized in that: The characteristic vector and weighted calculation are specifically as follows: Extract the activity level of each user, including low activity, general activity and high activity; extract the acceptance level of each user, including low acceptance, general acceptance and high acceptance; extract the consumption level of each user, including low consumption, general consumption and high consumption; construct a three-dimensional classification interval, including 27 grouping areas, each grouping area corresponds to a set of preset feature vectors (U1, U2, U3); Calculate the advertising evaluation vector (CVR1, CVR2, CVR3) of each user through the formula (CVR1, CVR2, CVR3) = (U1×Q1, U2×Q2, U3×Q3); When CVR1 is greater than a preset threshold, it is determined that the click and purchase revenue obtained by placing advertisements to the user is high; otherwise, it is determined that the click and purchase revenue obtained by placing advertisements to the user is low; When CVR2 is greater than a preset threshold, it is determined that the user has a low interest in advertising; otherwise, it is determined that the user has a high interest in advertising; When CVR3 is greater than a preset threshold, it is determined that the user has a high willingness to shop through advertising push; otherwise, it is determined that the user has a low willingness to shop through advertising push.

8. A mobile advertising effect analysis system based on user behavior, characterized in that: It includes data capture module, data preprocessing module, user analysis module and advertising optimization and analysis module; The data capture module captures user behavior and advertising delivery related data in real time through the SDA interface and API interface of the mobile advertising platform application, including user ID code, user gender code, user age code and user IP code; The data preprocessing module collects and records user behavior data, deletes abnormal data that exceeds the maximum preset extreme value, merges duplicate records, and converts data in different formats into a unified structure; For each user, generate a unique corresponding behavior data matrix Store the user list and its unique corresponding behavior data matrix into the database; The user analysis module uses the formula Calculate the activity parameter E1 of each user; where c1, c2 and c3 are preset weight coefficients, N11 and N12 represent the number of new items added to favorites and the number of ad clicks respectively; Ttype_p represents the browsing time of users on various product pages in the past month; Users whose activity parameter E1 is higher than the preset maximum threshold E1Max are marked as highly active users; users whose activity parameter E1 is higher than the minimum threshold E1Min and lower than the maximum threshold E1Max are marked as generally active users; users whose activity parameter E1 is lower than the minimum threshold E1Min are marked as low active users; By formula Calculate the advertisement acceptance parameter E2 of each user, where w1 and w2 are preset influencing factors; Mark users whose advertising acceptance parameter E2 is higher than the preset maximum threshold E2Max as high acceptance users; mark users whose advertising acceptance parameter E2 is higher than the minimum threshold E2Min and lower than the maximum threshold E2Max as average acceptance users; mark users whose advertising acceptance parameter is lower than the minimum threshold E2Min as low acceptance users; By formula Calculate the consumption level parameter E3 of each user, where w3, w4 and w5 are preset influencing factors; Users whose consumption level parameter E3 is higher than the preset maximum threshold E3Max are marked as high consumption level users; users whose consumption level parameter E3 is higher than the minimum threshold E3Min and lower than the maximum threshold E3Max are marked as average consumption level users; users whose consumption level parameter is lower than the minimum threshold E3Min are marked as low consumption level users; Get the total cost of advertising B1, the average traffic revenue obtained through all users' clicks and browsing B2, the total number of advertisements B3 and the total profit of the advertising push platform B4; through the formula Calculate each user's advertising revenue index Q1, advertising suitability index Q2, and shopping intention index Q3; Extract the activity level of each user, including low activity, general activity and high activity; extract the acceptance level of each user, including low acceptance, general acceptance and high acceptance; extract the consumption level of each user, including low consumption, general consumption and high consumption; construct a three-dimensional classification interval, including 27 grouping areas, each grouping area corresponds to a set of preset feature vectors (U1, U2, U3); Calculate the advertising evaluation vector (CVR1, CVR2, CVR3) of each user through the formula (CVR1, CVR2, CVR3) = (U1×Q1, U2×Q2, U3×Q3); When CVR1 is greater than a preset threshold, it is determined that the click and purchase revenue obtained by placing advertisements to the user is high; otherwise, it is determined that the click and purchase revenue obtained by placing advertisements to the user is low; When CVR2 is greater than a preset threshold, it is determined that the user has a low interest in advertising; otherwise, it is determined that the user has a high interest in advertising; When CVR3 is greater than a preset threshold, it is determined that the user has a high willingness to shop through advertising push; otherwise, it is determined that the user has a low willingness to shop through advertising push; The advertising optimization and analysis module uses collaborative filtering algorithms to recommend advertisements based on the behavioral similarities between users for each interest group. The similarity parameter COV (i, j) between each user i and j is calculated using the weighted Euclidean distance formula of collaborative filtering. Mark the user ID codes i and j of users whose similarity parameter COV (i, j) is greater than a preset threshold as high-similarity users; For users i and j with high similarity, whenever user i purchases a product, the same type of product will be pushed to j.

Images