A method and system for intelligent advertising delivery management based on data feature matching

Abstract

The present invention discloses an intelligent advertising delivery management method and system based on data feature matching, which relates to the field of intelligent advertising technology. By building a cloud data platform and a multi-channel management platform, integrating data from different sources, and building a preference prediction model, advertisements can better match the interests and preferences of target audiences, and the preference degree index can be combined with the number of operations on the commodity category in the current cycle and the historical cycle, so that the influence of the data of the historical cycle on the current user's commodity category preference gradually weakens over time, and can more comprehensively reflect the user's preference change trend and avoid the deviation caused by relying on historical cycle data. The method can accurately reflect the user's current preference for the category of commodities, thereby improving the accuracy and effect of advertising delivery. The method realizes continuous improvement of advertising delivery by collecting effect data after advertising delivery and feeding the data back to the user database.

Description

Technical Field

[0001] The present invention relates to the field of intelligent advertising technology, and in particular to an advertising intelligent delivery management method and system based on data feature matching. Background Art

[0002] With the rapid development of Internet technology, Internet advertising has gradually become a mainstream advertising media. Compared with traditional advertising media, Internet advertising has the advantages of wide coverage, strong initiative and enthusiasm, relatively low cost, high cost performance and strong interactivity. Therefore, Internet advertising is becoming more and more popular among companies and merchants.

[0003] Existing advertising delivery methods based on information flow usually adopt an on-demand allocation method based on a single media. This model lacks the function of controlling the frequency of advertising delivery to a single user. Multiple exposure of users to the same advertisement or the same type of advertisement may cause "advertising fatigue", which will reduce the click-through rate and conversion rate to a certain extent. In addition, the data from different online channels on each user end exist in isolation, making it difficult to conduct comprehensive analysis and optimization, and it is impossible to form a complete user feature data set, resulting in poor accuracy and effectiveness of advertising delivery.

[0004] In order to solve this problem, we propose an advertising intelligent delivery management method and system based on data feature matching. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an advertisement intelligent delivery management method and system based on data feature matching, which can effectively solve the problems in the background technology.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is: an advertisement intelligent delivery management method and system based on data feature matching, comprising:

[0007] Constructing a cloud data platform, wherein the cloud data platform is provided with an advertisement database and a user database, wherein the advertisement database is used to store a variety of original advertisement data;

[0008] After obtaining user authorization, it interacts with each user terminal and collects basic user information and network status information of each user terminal;

[0009] Set a data embedding point for each user terminal, collect the user feature data set and current timestamp of the corresponding user terminal in the current period at a certain frequency through the data embedding point, and upload the collected user feature data set and corresponding timestamp of each user terminal to the user database;

[0010] Build a multi-channel management platform, build a preference prediction model based on the user feature data set and timestamp collected from each user end, and obtain the user's preference index for each product category;

[0011] The multi-channel management platform obtains the real-time geographic location information and online channel type data of each user terminal, combines the real-time geographic location information and preference index of each user terminal, calculates the real-time matching degree between the user and the advertisement, and obtains the advertisement matching degree data;

[0012] Obtaining available advertisement delivery requests of each user terminal, where the available advertisement delivery requests include available advertisement delivery methods and corresponding available advertisement areas;

[0013] An advertisement bidding mechanism is set based on available advertisement delivery requests, so that advertisement publishers corresponding to each advertisement original data bid for the available advertisement area through the advertisement bidding mechanism to obtain bidding result data;

[0014] Build a comprehensive scoring model to generate advertising matching score data for each user end based on advertising matching data, advertising fatigue index and bidding result score, prioritize each advertisement according to the advertising matching score data, and select pre-delivered advertisements according to the priority ranking;

[0015] Based on the online channel type data of each user terminal, the delivery method of pre-delivered advertisements to each user terminal is preset to deliver the advertisements;

[0016] Collect the advertisement delivery frequency data, upload the delivery frequency data to the multi-channel management platform, collect the advertisement effect data after the advertisement is delivered, upload the advertisement effect data to the user database, and continuously optimize the subsequent advertisement delivery to each user end.

[0017] As a preferred embodiment, the cloud data platform is constructed, and the cloud data platform is provided with an advertisement database and a user database, the advertisement database is used to store a variety of original advertisement data; after obtaining user authorization, interacting with each user terminal, collecting basic user information and network status information of each user terminal; setting a data embedding point for each user terminal, collecting a user feature data set and a current timestamp of the corresponding user terminal in the current period at a certain frequency through the data embedding point, and uploading the collected user feature data set and the corresponding timestamp of each user terminal to the user database, specifically including:

[0018] The original advertisement data includes basic advertisement content, advertisement publisher and delivery effect data;

[0019] The user database is used to store basic user information, network status information and user feature data sets of each user terminal, and classify and organize the data;

[0020] The data embedding points include data collection embedding points and data integration embedding points;

[0021] Data collection points are set in various software on the user side to capture user interaction behaviors and usage records, and record the current timestamp;

[0022] The data integration tracking point is used to receive the data collected by the data collection tracking point on the server side, clean the collected data, filter out invalid or abnormal data records, fill in missing data fields, convert the data into a unified format, merge it into a user feature data set and store it in the user database for further analysis and optimization;

[0023] The user feature data set includes browsing data, search data, click data and consumption data;

[0024] Among them, browsing data includes the number of views and the categories of browsed goods, search data includes the number of searches and the categories of searched goods, click record data includes the number of clicks and the categories of clicked goods, and consumption data includes the number of consumptions and the categories of consumed goods.

[0025] Preferably, the multi-channel management platform is constructed to construct a preference prediction model based on the user feature data set and timestamp collected from each user terminal to obtain the user's preference index for each commodity category, specifically including:

[0026] Multi-channel management platform is used to integrate and manage data from various user terminals;

[0027] According to the basic information of users in the user database, each user is associated with the corresponding user feature data set and timestamp;

[0028] Extracting browsed product category information and corresponding browse count information from the browse data, extracting searched product category information and corresponding search count information from the search data, extracting clicked product category information and corresponding click count information from the click record data, and extracting consumed product category information and corresponding consumption count information from the consumption data;

[0029] The collected data is divided into different periods according to the timestamp. Each timestamp corresponds to a data set of a period. The duration of each period is the same and each period is recorded in chronological order. Cycle, Cycle… cycle, classify and integrate the above data according to the commodity category information of the data in each data set, calculate the user's preference for the category of commodities in the current cycle according to the preference prediction model, and obtain the preference index;

[0030] The preference index is:

[0031] ;

[0032] In the formula, Indicates user In the Product categories degree of preference, Indicated in Product categories within a cycle degree of preference, Indicated in Product categories within a cycle degree of preference, Indicates user In the Product categories within a cycle Number of views, Indicates user In the Product categories within a cycle Number of searches, Indicates user In the Product categories within a cycle Number of clicks, Indicates user In the Product categories within a cycle The number of consumptions, Indicates user In the Product categories within a cycle Number of views, Indicates user In the Product categories within a cycle Number of searches, Indicates user In the Product categories within a cycle Number of clicks, Indicates user In the Product categories within a cycle The number of consumptions, , , , are weight parameters, is a constant between 0 and 1. Represents the ratio of weight reduction for each forward cycle.

[0033] Preferably, the real-time geographic location information and online channel type data of each user terminal are obtained, and the real-time matching degree between the user and the advertisement is calculated in combination with the real-time geographic location information and preference index of each user terminal to obtain the advertisement matching degree data, specifically including:

[0034] Obtain online channel historical transaction data based on online channel type data;

[0035] According to the historical transaction data of the online channel, based on the named entity recognition of the transaction goods, the transaction tag information is obtained, and the transaction tag information represents each category of the transaction goods;

[0036] Perform named entity recognition on the original advertisement data to obtain advertisement label information, where the advertisement label information indicates each advertisement type, and each advertisement in the original advertisement data is correspondingly provided with at least one advertisement label information;

[0037] Product categories Ad Type Match and obtain the advertising index based on the transaction tag information and the advertising tag information;

[0038] ;

[0039] In the formula, For product category The advertising delivery index of the corresponding advertising type, For product category The number of tags in the transaction tag information, is the total number of labels of the advertising label information, For the Total number of transaction tags in online channels;

[0040] Obtain the location information of the advertiser based on the original advertisement data;

[0041] According to the advertising merchant location information and the user's real-time geographic location information, the inverse of the distance between the advertising merchant location and the user's terminal location is used as the distance influence coefficient;

[0042] Obtain the advertising matching degree based on the advertising placement index, location influence coefficient and preference index;

[0043] The calculation formula for advertisement matching is:

[0044] ;

[0045] In the formula, For the Ad matching degree of ads, For product categories The corresponding Ad delivery index of ad type for ads, is the distance between the advertising merchant location and the user's location, For the The preference index of the product category corresponding to the advertisement.

[0046] Preferably, the comprehensive scoring model is constructed to generate advertisement matching scoring data of each user terminal based on advertisement matching data, advertisement fatigue index and bidding result scoring, and to prioritize each advertisement according to the advertisement matching scoring data, and to select pre-delivered advertisements according to the priority ranking, specifically including:

[0047] Obtain the bidding result data provided by the advertisement publisher, and calculate the bidding result score of the advertisement based on the bidding result data and the delivery effect data corresponding to the advertisement;

[0048] The delivery effect data includes the historical click-through rate and historical conversion rate of the advertisement;

[0049] Obtain the advertising delivery data of the advertisement in the current cycle, and calculate the advertising fatigue index of the advertisement based on the delivery frequency data;

[0050] Advertisement delivery data includes the frequency of ad delivery, user reporting frequency and total browsing time in the current cycle;

[0051] Based on the current cycle’s ad matching data, ad delivery data, and bidding result scores, obtain ad matching score indicators for each type of ad on each user end;

[0052] The priority of advertisements is sorted based on the advertisement matching score index data of each user end. Ads with higher advertisement matching scores have higher priority.

[0053] A scoring threshold is set according to the delivery demand, and ads above the scoring threshold are considered as pre-delivery ads;

[0054] The bidding result score is:

[0055] ;

[0056] In the formula, Indicates the bidding result score. Indicates the historical click-through rate of the advertisement. Indicates the historical conversion rate of the ad. Indicates the advertising bidding price data. , , They respectively represent the weight parameters corresponding to the historical click-through rate of advertisements, the historical conversion rate of advertisements and the bidding price data of advertisements;

[0057] The advertising fatigue index is:

[0058] ;

[0059] In the formula, represents the advertising fatigue index, Indicates the frequency of advertising delivery. Indicates the frequency of user reports. Indicates the ad viewing time;

[0060] The advertisement matching scoring index is:

[0061] ;

[0062] In the formula, Indicates user In the The cycle Matching score data for ads, Indicates the ad matching degree. Indicates the bidding result score. represents the advertising fatigue index, , , They represent the weight parameters corresponding to the advertising matching degree, delivery frequency and bidding result score respectively. Indicates the offset.

[0063] Preferably, the method of presetting a delivery method of pre-delivering advertisements to each user terminal based on the online channel type data of each user terminal and delivering the advertisements specifically includes:

[0064] Determine the current channel type of each user terminal through the collected online channel type data;

[0065] Adjust and preset the delivery method of the advertisement according to the type of current online channel;

[0066] The pre-delivered ads are delivered in sequence according to the priority, and the delivery effect of the ads is monitored in real time.

[0067] Preferably, the method collects the advertisement delivery frequency data, uploads the delivery frequency data to the multi-channel management platform, collects the advertisement effect data after the advertisement delivery, uploads the advertisement effect data to the user database, and continuously optimizes the subsequent advertisement delivery of each user terminal, specifically including:

[0068] Collect the advertising data of the online channel in the current cycle;

[0069] Upload the advertising data of the online channel in the current cycle to the multi-channel management platform;

[0070] When it is detected that the advertising delivery frequency of the online channel has changed in the current cycle, the latest delivery frequency data is uploaded to the multi-channel management platform and the original advertising delivery data is overwritten. At the end of the current cycle, the latest advertising delivery data is ensured to be retained in the multi-channel management platform;

[0071] Collect advertising effectiveness data of the online channel in the current cycle;

[0072] The advertising effect data includes browsing data, click data and consumption data of the advertisement;

[0073] The above data will be merged into the user database for subsequent data analysis on the user side and continuous optimization of advertising delivery.

[0074] Furthermore, the present invention also proposes an advertisement intelligent delivery management system based on data feature matching, which is used to implement any of the above management methods, including:

[0075] The cloud data platform module is used to provide an infrastructure for data storage and management, and is used to store basic information about advertisements, publisher information, delivery effectiveness data, and basic information about users, network status information, and user feature data sets, and to organize and classify the data;

[0076] The data collection module is used to obtain user authorization, ensure the legitimacy of data collection, collect various data from the user end, clean and filter the data, fill in missing fields, convert the format, and merge them into user feature data sets and store them in the user database;

[0077] The multi-channel management module is used to integrate and manage data from various user terminals, build a user preference prediction model, and calculate the user's preference for different product categories;

[0078] The real-time matching module is used to obtain the real-time geographic location information and online channel type data of the user end in real time, and calculate the advertising matching degree between the user and the advertisement;

[0079] Advertisement bidding module, used to obtain bidding result data from advertisement publishers, and obtain bidding result scores through bidding result data and delivery effect data;

[0080] Comprehensive scoring module, which combines the advertisement matching data, advertisement fatigue index and bidding result scoring to generate advertisement matching scores;

[0081] Advertisement delivery module, used to deliver advertisements according to priority and monitor delivery effects in real time;

[0082] The data feedback module is used to collect data after advertising is launched, conduct effect evaluation and continuous optimization.

[0083] Preferably, the data acquisition module specifically includes:

[0084] User authorization management unit, used to interact with each user terminal after obtaining user authorization;

[0085] The data tracking unit is used to set data collection tracking points in various software on the user side to capture the user's interactive behavior and usage records, and record the current timestamp;

[0086] The data integration unit is used to set up data integration points on the server side, to receive data collected by data collection points, to clean, filter, fill in missing fields, convert the format of the collected data, and merge them into user feature data sets, and store them in the user database.

[0087] Preferably, the comprehensive scoring module specifically includes:

[0088] The advertisement delivery data collection unit is used to obtain the advertisement delivery data of the advertisement delivered in the current cycle;

[0089] A bidding result score calculation unit, used to calculate the bidding result score of the advertisement based on the bidding result data and the delivery effect data;

[0090] A fatigue index calculation module is used to collect the advertising delivery data in the current cycle and calculate the advertising fatigue index of the advertisement based on the delivery frequency data;

[0091] An advertisement matching score calculation unit, used to calculate an advertisement matching score by combining advertisement matching data, advertisement delivery data and bidding result score;

[0092] The priority sorting unit is used to prioritize advertisements according to advertisement matching scores and filter out pre-delivery advertisements according to score thresholds.

[0093] Compared with the prior art, the present invention provides an advertisement intelligent delivery management method and system based on data feature matching, which has the following beneficial effects:

[0094] 1. By building a cloud data platform and a multi-channel management platform, data from different sources can be integrated. By building a preference prediction model, advertisements can better match the interests and preferences of the target audience. The preference index can be combined with the number of operations on the product category in the current cycle and the historical cycle, so that the impact of the historical cycle data on the current user's product category preference will gradually weaken over time. It can more comprehensively reflect the user's preference change trend and avoid the deviation caused by relying on historical cycle data. This advertising delivery method can accurately reflect the user's current preference for this category of products, thereby improving the accuracy and effectiveness of advertising delivery.

[0095] 2. By constructing a comprehensive scoring model, the advertising matching score data of each user end is generated based on the advertising matching data, advertising fatigue index and bidding result score, and each advertisement is prioritized according to the advertising matching score data. Pre-delivered advertisements are screened out according to the priority ranking, and the advertising fatigue index is used as a negative indicator of the advertising matching score. The number of times an advertisement is delivered can be limited, effectively reducing the excessive exposure of a single advertisement to the user and avoiding the user's "ad fatigue" for specific advertisements.

[0096] 3. This advertising delivery method collects effect data after advertising delivery, such as advertising browsing data, click data and consumption data, and feeds this data back to the user database for continuous optimization of advertising delivery strategies and continuous improvement of advertising delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 The present invention is a flow chart of an advertising intelligent delivery management method based on data feature matching.

[0098] Figure 2 The present invention is a flowchart for obtaining advertisement matching data.

[0099] Figure 3 The present invention is a framework diagram of an intelligent advertising delivery management system based on data feature matching. DETAILED DESCRIPTION

[0100] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the present invention will be further described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0101] In order to address the shortcomings of the existing technology, such as Figure 1 , 2 As shown, the present invention provides an advertisement intelligent delivery management method based on data feature matching, comprising:

[0102] Build a cloud data platform, which is equipped with an advertising database and a user database. The advertising database is used to store a variety of original advertising data;

[0103] After obtaining user authorization, it interacts with each user terminal and collects basic user information and network status information of each user terminal;

[0104] Set a data embedding point for each user terminal, collect the user feature data set and current timestamp of the corresponding user terminal in the current period at a certain frequency through the data embedding point, and upload the collected user feature data set and corresponding timestamp of each user terminal to the user database;

[0105] Build a multi-channel management platform, build a preference prediction model based on the user feature data set and timestamp collected from each user end, and obtain the user's preference index for each product category;

[0106] The multi-channel management platform obtains the real-time geographic location information and online channel type data of each user terminal, combines the real-time geographic location information and preference index of each user terminal, calculates the real-time matching degree between the user and the advertisement, and obtains the advertisement matching degree data;

[0107] Obtaining available advertisement delivery requests of each user terminal, where the available advertisement delivery requests include available advertisement delivery methods and corresponding available advertisement areas;

[0108] An advertisement bidding mechanism is set based on available advertisement delivery requests, so that advertisement publishers corresponding to each advertisement original data bid for the available advertisement area through the advertisement bidding mechanism to obtain bidding result data;

[0109] Build a comprehensive scoring model to generate advertising matching score data for each user end based on advertising matching data, advertising fatigue index and bidding result score, prioritize each advertisement according to the advertising matching score data, and select pre-delivered advertisements according to the priority ranking;

[0110] Based on the online channel type data of each user terminal, the delivery method of pre-delivered advertisements to each user terminal is preset to deliver the advertisements;

[0111] Collect the advertisement delivery frequency data, upload the delivery frequency data to the multi-channel management platform, collect the advertisement effect data after the advertisement is delivered, upload the advertisement effect data to the user database, and continuously optimize the subsequent advertisement delivery to each user end.

[0112] Specifically, a cloud data platform is constructed, and the cloud data platform is provided with an advertising database and a user database. The advertising database is used to store various original advertising data; after obtaining user authorization, the platform interacts with each user terminal to collect basic user information and network status information of each user terminal; a data embedding point is set for each user terminal, and the user feature data set and the current timestamp of the corresponding user terminal in the current period are collected at a certain frequency through the data embedding point, and the collected user feature data set and the corresponding timestamp of each user terminal are uploaded to the user database, including:

[0113] The original advertising data includes basic advertising content, advertising publishers and delivery effect data;

[0114] The user database is used to store basic user information, network status information and user feature data sets of each user terminal, and classify and organize the data;

[0115] Data embedding includes data collection embedding and data integration embedding;

[0116] Data collection points are set in various software on the user side to capture user interaction behaviors and usage records, and record the current timestamp;

[0117] The data integration tracking point is used to receive the data collected by the data collection tracking point on the server side, clean the collected data, filter out invalid or abnormal data records, fill in missing data fields, convert the data into a unified format, merge it into a user feature data set and store it in the user database for further analysis and optimization;

[0118] User feature data sets include browsing data, search data, click data, and consumption data;

[0119] Among them, browsing data includes the number of views and the categories of browsed goods, search data includes the number of searches and the categories of searched goods, click record data includes the number of clicks and the categories of clicked goods, and consumption data includes the number of consumptions and the categories of consumed goods.

[0120] Specifically, a multi-channel management platform is constructed, and a preference prediction model is constructed based on the user feature data set and timestamp collected from each user terminal to obtain the user's preference index for each product category, including:

[0121] Multi-channel management platform is used to integrate and manage data from various user terminals;

[0122] According to the basic information of users in the user database, each user is associated with the corresponding user feature data set and timestamp;

[0123] Extracting browsed product category information and corresponding browse count information from the browse data, extracting searched product category information and corresponding search count information from the search data, extracting clicked product category information and corresponding click count information from the click record data, and extracting consumed product category information and corresponding consumption count information from the consumption data;

[0124] The collected data is divided into different periods according to the timestamp. Each timestamp corresponds to a data set of a period. The duration of each period is the same and each period is recorded in chronological order. Cycle, Cycle… cycle, classify and integrate the above data according to the commodity category information of the data in each data set, calculate the user's preference for the category of commodities in the current cycle according to the preference prediction model, and obtain the preference index;

[0125] The preference index is:

[0126] ;

[0127] In the formula, Indicates user In the Product categories degree of preference, Indicated in Product categories within a cycle degree of preference, Indicated in Product categories within a cycle degree of preference, Indicates user In the Product categories within a cycle Number of views, Indicates user In the Product categories within a cycle Number of searches, Indicates user In the Product categories within a cycle Number of clicks, Indicates user In the Product categories within a cycle The number of consumptions, Indicates user In the Product categories within a cycle Number of views, Indicates user In the Product categories within a cycle Number of searches, Indicates user In the Product categories within a cycle Number of clicks, Indicates user In the Product categories within a cycle The number of consumptions, , , , are weight parameters, is a constant between 0 and 1. Represents the ratio of weight reduction for each forward cycle.

[0128] Specifically, the real-time geographic location information and online channel type data of each user terminal are obtained, and the real-time matching degree between the user and the advertisement is calculated by combining the real-time geographic location information and preference index of each user terminal to obtain advertisement matching degree data, including:

[0129] Obtain online channel historical transaction data based on online channel type data;

[0130] According to the historical transaction data of online channels, based on the named entity recognition of transaction commodities, the transaction label information is obtained, and the transaction label information represents the categories of various commodities in the transaction. It should be noted that the transaction label information can be divided into food, accommodation services, clothing, medical care, education and training, etc.;

[0131] Perform named entity recognition on the original advertisement data to obtain advertisement label information, where the advertisement label information indicates each advertisement type. Each advertisement in the original advertisement data is correspondingly provided with at least one advertisement label information, and the advertisement label information may correspond to the transaction label information;

[0132] Product categories Ad Type Match and obtain the advertising index based on the transaction tag information and the advertising tag information;

[0133] ;

[0134] In the formula, For product category The advertising delivery index of the corresponding advertising type, For product category The number of tags in the transaction tag information, is the total number of labels of the advertising label information, For the Total number of transaction tags in online channels;

[0135] Obtain the location information of the advertiser based on the original advertisement data;

[0136] According to the advertising merchant location information and the user's real-time geographic location information, the inverse of the distance between the advertising merchant location and the user's terminal location is used as the distance influence coefficient;

[0137] Obtain the advertising matching degree based on the advertising placement index, location influence coefficient and preference index;

[0138] The calculation formula for advertisement matching is:

[0139] ;

[0140] In the formula, For the Ad matching degree of ads, For product categories The corresponding Ad delivery index of ad type for ads, is the distance between the advertising merchant location and the user's location, For the The preference index of the product category corresponding to the advertisement.

[0141] Specifically, a comprehensive scoring model is constructed to generate advertising matching scoring data for each user end based on advertising matching data, advertising fatigue index, and bidding result scores, and to prioritize each advertisement according to the advertising matching scoring data, and to select pre-delivered advertisements according to the priority ranking, including:

[0142] Obtain the bidding result data provided by the advertisement publisher, and calculate the bidding result score of the advertisement based on the bidding result data and the delivery effect data corresponding to the advertisement;

[0143] The delivery effect data includes the historical click-through rate and historical conversion rate of the advertisement;

[0144] Obtain the advertising delivery data of the advertisement in the current cycle, and calculate the advertising fatigue index of the advertisement based on the delivery frequency data;

[0145] Advertisement delivery data includes the frequency of ad delivery, user reporting frequency and total browsing time in the current cycle;

[0146] Based on the current cycle’s ad matching data, ad delivery data, and bidding result scores, obtain ad matching score indicators for each type of ad on each user end;

[0147] The priority of advertisements is sorted based on the advertisement matching score index data of each user end. Ads with higher advertisement matching scores have higher priority.

[0148] A scoring threshold is set according to the delivery demand, and ads above the scoring threshold are considered as pre-delivery ads;

[0149] The bidding result score is;

[0150] ;

[0151] In the formula, Indicates the bidding result score. Indicates the historical click-through rate of the advertisement. Indicates the historical conversion rate of the ad. Indicates the advertising bidding price data. , , They respectively represent the weight parameters corresponding to the historical click-through rate of advertisements, the historical conversion rate of advertisements and the bidding price data of advertisements;

[0152] The advertising fatigue index is:

[0153] ;

[0154] In the formula, represents the advertising fatigue index, Indicates the frequency of advertising delivery. Indicates the frequency of user reports. Indicates the ad viewing time;

[0155] Ad matching score indicators are:

[0156] ;

[0157] In the formula, Indicates user In the The cycle Matching score data for ads, Indicates the ad matching degree. Indicates the bidding result score. represents the advertising fatigue index, , , They represent the weight parameters corresponding to the advertising matching degree, delivery frequency and bidding result score respectively. Indicates the offset.

[0158] Specifically, based on the online channel type data of each user terminal, a delivery method for pre-delivering advertisements to each user terminal is preset, and advertisements are delivered, including:

[0159] Determine the current channel type of each user terminal through the collected online channel type data;

[0160] Adjust and preset the delivery method of the advertisement according to the type of current online channel;

[0161] The pre-delivered ads are delivered in order according to the priority, and the delivery effect of the ads is monitored in real time;

[0162] It should be noted that by building a preference prediction model, the user's preference for the category of goods in the current cycle is calculated to obtain a preference index. Based on the user's preference index, personalized advertising recommendations are provided to reduce the interference of irrelevant advertisements. Through the advertising bidding mechanism, fair competition for advertising positions is ensured. The advertising matching score is generated by comprehensively considering the matching degree, frequency of advertising and bidding results. At the same time, the advertising fatigue index, as a negative indicator of the advertising matching score, can limit the number of advertising runs, effectively reduce the excessive exposure of a single advertisement to the user, and avoid the user's "advertising fatigue" for specific advertisements.

[0163] And the types of online channels at least include Internet advertising platforms, social media, and search engines; when the current user's online channel type is an Internet advertising platform, banner ads can be displayed at the top, bottom, or sidebar of the website or application; when the current user's online channel type is social media, ads can be inserted into the information flow and displayed mixed with other content; when the current user's online channel type is a search engine, keywords can be displayed in the user's search bar for advertising delivery, and the way ads are displayed can be adjusted according to different online channel types to ensure the display effect of ads in different scenarios.

[0164] Specifically, collect the advertisement delivery frequency data, upload the delivery frequency data to the multi-channel management platform, collect the advertisement effect data after the advertisement delivery, upload the advertisement effect data to the user database, and continuously optimize the subsequent advertisement delivery of each user end, including:

[0165] Collect the advertising data of the online channel in the current cycle;

[0166] Upload the advertising data of the online channel in the current cycle to the multi-channel management platform;

[0167] When it is detected that the advertising delivery frequency of the online channel has changed in the current cycle, the latest delivery frequency data is uploaded to the multi-channel management platform and the original advertising delivery data is overwritten. At the end of the current cycle, the latest advertising delivery data is ensured to be retained in the multi-channel management platform;

[0168] Collect advertising effectiveness data of the online channel in the current cycle;

[0169] Advertising effect data includes browsing data, click data and consumption data of the advertisement;

[0170] The above data is merged into the user database for subsequent data analysis of the user end. This method forms a closed-loop data feedback mechanism by continuously collecting advertising frequency and effect data to continuously optimize advertising strategies and improve the overall effect of advertising.

[0171] Further, such as Figure 3 As shown, the present invention also proposes an advertising intelligent delivery management system based on data feature matching, which is used to implement any of the above management methods, including:

[0172] The cloud data platform module is used to provide an infrastructure for data storage and management, and is used to store basic information about advertisements, publisher information, delivery effectiveness data, and basic information about users, network status information, and user feature data sets, and to organize and classify the data;

[0173] The data collection module is used to obtain user authorization, ensure the legitimacy of data collection, collect various data from the user end, clean and filter the data, fill in missing fields, convert the format, and merge them into user feature data sets and store them in the user database;

[0174] The multi-channel management module is used to integrate and manage data from various user terminals, build a user preference prediction model, and calculate the user's preference for different product categories;

[0175] The real-time matching module is used to obtain the real-time geographic location information and online channel type data of the user end in real time, and calculate the advertising matching degree between the user and the advertisement;

[0176] Advertisement bidding module, used to obtain bidding result data from advertisement publishers, and obtain bidding result scores through bidding result data and delivery effect data;

[0177] Comprehensive scoring module, which combines the advertisement matching data, advertisement fatigue index and bidding result scoring to generate advertisement matching scores;

[0178] Advertisement delivery module, used to deliver advertisements according to priority and monitor delivery effects in real time;

[0179] The data feedback module is used to collect data after advertising is launched, conduct effect evaluation and continuous optimization.

[0180] Specifically, the data acquisition module includes:

[0181] User authorization management unit, used to interact with each user terminal after obtaining user authorization;

[0182] The data tracking unit is used to set data collection tracking points in various software on the user side to capture the user's interactive behavior and usage records, and record the current timestamp;

[0183] The data integration unit is used to set up data integration points on the server side, to receive data collected by data collection points, to clean, filter, fill in missing fields, convert the format of the collected data, and merge them into user feature data sets, and store them in the user database.

[0184] Specifically, the comprehensive scoring module includes:

[0185] The advertisement delivery data collection unit is used to obtain the advertisement delivery data of the advertisement delivered in the current cycle;

[0186] A bidding result score calculation unit, used to calculate the bidding result score of the advertisement based on the bidding result data and the delivery effect data;

[0187] A fatigue index calculation module is used to collect the advertising delivery data in the current cycle and calculate the advertising fatigue index of the advertisement based on the delivery frequency data;

[0188] An advertisement matching score calculation unit, used to calculate an advertisement matching score by combining advertisement matching data, advertisement delivery data and bidding result score;

[0189] The priority sorting unit is used to prioritize advertisements according to advertisement matching scores and filter out pre-delivery advertisements according to score thresholds.

[0190] In summary, the advantages of the present invention are:

[0191] This method generates an ad matching score by comprehensively evaluating the ad matching data, ad fatigue index and bidding result score, provides personalized ad recommendations, reduces interference from irrelevant ads, ensures the accuracy of ad delivery, and introduces the ad fatigue index as part of the ad matching score, which helps limit overexposure of ads and avoids user boredom. At the same time, it can dynamically adjust the frequency of ad delivery and balance ad effect and user experience. The user database can collect ad effect data after ad delivery to form a data feedback mechanism to continuously optimize the ad delivery strategy.

[0192] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. An intelligent advertising delivery management method based on data feature matching, characterized in that: include: Constructing a cloud data platform, wherein the cloud data platform is provided with an advertisement database and a user database, wherein the advertisement database is used to store a variety of original advertisement data; After obtaining user authorization, it interacts with each user terminal and collects basic user information and network status information of each user terminal; Set a data embedding point for each user terminal, collect the user feature data set and current timestamp of the corresponding user terminal in the current period at a certain frequency through the data embedding point, and upload the collected user feature data set and corresponding timestamp of each user terminal to the user database; Build a multi-channel management platform, build a preference prediction model based on the user feature data set and timestamp collected from each user end, and obtain the user's preference index for each product category; The multi-channel management platform obtains the real-time geographic location information and online channel type data of each user terminal, combines the real-time geographic location information and preference index of each user terminal, calculates the real-time matching degree between the user and the advertisement, and obtains the advertisement matching degree data; Obtaining available advertisement delivery requests of each user terminal, where the available advertisement delivery requests include available advertisement delivery methods and corresponding available advertisement areas; An advertisement bidding mechanism is set based on available advertisement delivery requests, so that advertisement publishers corresponding to each advertisement original data bid for the available advertisement area through the advertisement bidding mechanism to obtain bidding result data; Build a comprehensive scoring model to generate advertising matching score data for each user end based on advertising matching data, advertising fatigue index and bidding result score, prioritize each advertisement according to the advertising matching score data, and select pre-delivered advertisements according to the priority ranking; Based on the online channel type data of each user terminal, the delivery method of pre-delivered advertisements to each user terminal is preset to deliver the advertisements; Collect the frequency data of the advertisement, upload the frequency data to the multi-channel management platform, collect the advertising effect data after the advertisement is released, upload the advertising effect data to the user database, and continuously optimize the subsequent advertisement release for each user end; The multi-channel management platform is constructed to construct a preference prediction model based on the user feature data set and timestamp collected from each user terminal to obtain the user's preference index for each commodity category, specifically including: Multi-channel management platform is used to integrate and manage data from various user terminals; According to the basic information of users in the user database, each user is associated with the corresponding user feature data set and timestamp; Extracting browsed product category information and corresponding browse count information from the browse data, extracting searched product category information and corresponding search count information from the search data, extracting clicked product category information and corresponding click count information from the click record data, and extracting consumed product category information and corresponding consumption count information from the consumption data; The collected data is divided into different periods according to the timestamp. Each timestamp corresponds to a data set of a period. The duration of each period is the same and each period is recorded in chronological order. Cycle, Cycle… cycle, classify and integrate the above data according to the commodity category information of the data in each data set, calculate the user's preference for the category of commodities in the current cycle according to the preference prediction model, and obtain the preference index; The preference index is: ； In the formula, Indicates user In the Product categories degree of preference, Indicated in Product categories within a cycle degree of preference, Indicated in Product categories within a cycle degree of preference, Indicates user In the Product categories within a cycle Number of views, Indicates user In the Product categories within a cycle Number of searches, Indicates user In the Product categories within a cycle Number of clicks, Indicates user In the Product categories within a cycle The number of consumptions, Indicates user In the Product categories within a cycle Number of views, Indicates user In the Product categories within a cycle Number of searches, Indicates user In the Product categories within a cycle Number of clicks, Indicates user In the Product categories within a cycle The number of consumptions, , , , are weight parameters, is a constant between 0 and 1. Represents the proportion of weight reduction in each forward cycle; The acquisition of the real-time geographic location information and online channel type data of each user terminal, combined with the real-time geographic location information and preference index of each user terminal, calculates the real-time matching degree between the user and the advertisement, and obtains the advertisement matching degree data, specifically includes: Obtain online channel historical transaction data based on online channel type data; According to the historical transaction data of the online channel, based on the named entity recognition of the transaction goods, the transaction tag information is obtained, and the transaction tag information represents each category of the transaction goods; Perform named entity recognition on the original advertisement data to obtain advertisement label information, where the advertisement label information indicates each advertisement type, and each advertisement in the original advertisement data is correspondingly provided with at least one advertisement label information; Product categories Ad Type Match and obtain the advertising index based on the transaction tag information and the advertising tag information; ； In the formula, For product category The advertising delivery index of the corresponding advertising type, For product category The number of tags in the transaction tag information, is the total number of labels of the advertising label information, For the Total number of transaction tags in online channels; Obtain the location information of the advertiser based on the original advertisement data; According to the advertising merchant location information and the user's real-time geographic location information, the inverse of the distance between the advertising merchant location and the user's terminal location is used as the distance influence coefficient; Obtain the advertising matching degree based on the advertising placement index, location influence coefficient and preference index; The calculation formula for advertisement matching is: ； In the formula, For the Ad matching degree of ads, For product categories The corresponding Ad delivery index of ad type for ads, is the distance between the advertising merchant location and the user's location, For the The preference index of the product category corresponding to the advertisement.

2. The method for intelligent advertising delivery management based on data feature matching according to claim 1, characterized in that: The cloud data platform is constructed, and the cloud data platform is provided with an advertisement database and a user database, and the advertisement database is used to store a variety of original advertisement data; after obtaining user authorization, interact with each user terminal, and collect basic user information and network status information of each user terminal; set a data embedding point for each user terminal, and collect the user feature data set and the current timestamp of the corresponding user terminal in the current period at a certain frequency through the data embedding point, and upload the collected user feature data set and the corresponding timestamp of each user terminal to the user database, specifically including: The original advertisement data includes basic advertisement content, advertisement publisher and delivery effect data; The user database is used to store basic user information, network status information and user feature data sets of each user terminal, and classify and organize the data; The data embedding points include data collection embedding points and data integration embedding points; Data collection points are set in various software on the user side to capture user interaction behaviors and usage records, and record the current timestamp; The data integration tracking point is used to receive the data collected by the data collection tracking point on the server side, clean the collected data, filter out invalid or abnormal data records, fill in missing data fields, convert the data into a unified format, merge it into a user feature data set and store it in the user database for further analysis and optimization; The user feature data set includes browsing data, search data, click data and consumption data; Among them, browsing data includes the number of views and the categories of browsed goods, search data includes the number of searches and the categories of searched goods, click record data includes the number of clicks and the categories of clicked goods, and consumption data includes the number of consumptions and the categories of consumed goods.

3. The method for intelligent advertising delivery management based on data feature matching according to claim 2, characterized in that: The comprehensive scoring model is constructed to generate the advertisement matching scoring data of each user terminal based on the advertisement matching data, advertisement fatigue index and bidding result score, and to prioritize each advertisement according to the advertisement matching scoring data, and to select the pre-delivered advertisements according to the priority ranking, specifically including: Obtain the bidding result data provided by the advertisement publisher, and calculate the bidding result score of the advertisement based on the bidding result data and the delivery effect data corresponding to the advertisement; The delivery effect data includes the historical click-through rate and historical conversion rate of the advertisement; Obtain the advertising delivery data of the advertisement in the current cycle, and calculate the advertising fatigue index of the advertisement based on the delivery frequency data; Advertisement delivery data includes the frequency of ad delivery, user reporting frequency and total browsing time in the current cycle; Based on the current cycle’s ad matching data, ad delivery data, and bidding result scores, obtain ad matching score indicators for each type of ad on each user end; The priority of advertisements is sorted based on the advertisement matching score index data of each user end. Ads with higher advertisement matching scores have higher priority. A scoring threshold is set according to the delivery demand, and ads above the scoring threshold are considered as pre-delivery ads; The bidding result score is: ； In the formula, Indicates the bidding result score. Indicates the historical click-through rate of the advertisement. Indicates the historical conversion rate of the ad. Indicates the advertising bidding price data. , , They respectively represent the weight parameters corresponding to the historical click-through rate of advertisements, the historical conversion rate of advertisements and the bidding price data of advertisements; The advertising fatigue index is: ； In the formula, represents the advertising fatigue index, Indicates the frequency of advertising delivery. Indicates the frequency of user reports. Indicates the ad viewing time; The advertisement matching scoring index is: ； In the formula, Indicates user In the The cycle Matching score data for ads, Indicates the ad matching degree. Indicates the bidding result score. represents the advertising fatigue index, , , They represent the weight parameters corresponding to the advertising matching degree, delivery frequency and bidding result score respectively. Indicates the offset.

4. The method for intelligent advertising delivery management based on data feature matching according to claim 3, characterized in that: The method of presetting the delivery method of pre-delivering advertisements to each user terminal based on the online channel type data of each user terminal and delivering the advertisements specifically includes: Determine the current channel type of each user terminal through the collected online channel type data; Adjust and preset the delivery method of the advertisement according to the type of current online channel; The pre-delivered ads are delivered in sequence according to the priority, and the delivery effect of the ads is monitored in real time.

5. The method for intelligent advertising delivery management based on data feature matching according to claim 4, characterized in that: The collecting of the advertisement delivery frequency data, uploading the delivery frequency data to the multi-channel management platform, collecting the advertisement effect data after the advertisement delivery, uploading the advertisement effect data to the user database, and continuously optimizing the subsequent advertisement delivery of each user end specifically include: Collect the advertising data of the online channel in the current cycle; Upload the advertising data of the online channel in the current cycle to the multi-channel management platform; When it is detected that the advertising delivery frequency of the online channel has changed in the current cycle, the latest delivery frequency data is uploaded to the multi-channel management platform and the original advertising delivery data is overwritten. At the end of the current cycle, the latest advertising delivery data is ensured to be retained in the multi-channel management platform; Collect the advertising effect data of the online channel in the current cycle, the advertising effect data includes browsing data, click data and consumption data of the advertisement, and merge the above data into the user database for subsequent data analysis on the user end and continuous optimization of advertising delivery.

6. An intelligent advertising delivery management system based on data feature matching, used to implement the management method according to any one of claims 1 to 5, characterized in that: include: The cloud data platform module is used to provide the infrastructure for data storage and management, and is used to store basic information of advertisements, publisher information, delivery effect data, and basic information of users, network status information, and user feature data sets, and to organize and classify the data; The data collection module is used to obtain user authorization, ensure the legitimacy of data collection, collect various data from the user end, clean and filter the data, fill in missing fields, convert the format, and merge them into user feature data sets and store them in the user database; The multi-channel management module is used to integrate and manage data from various user terminals, build a user preference prediction model, and calculate the user's preference for different product categories; The real-time matching module is used to obtain the real-time geographic location information and online channel type data of the user end in real time, and calculate the advertising matching degree between the user and the advertisement; Advertisement bidding module, used to obtain bidding result data from advertisement publishers, and obtain bidding result scores through bidding result data and delivery effect data; Comprehensive scoring module, which combines the advertisement matching data, advertisement fatigue index and bidding result scoring to generate advertisement matching scores; Advertisement delivery module, used to deliver advertisements according to priority and monitor delivery effects in real time; The data feedback module is used to collect data after advertising is launched, conduct effect evaluation and continuous optimization.

7. The advertising intelligent delivery management system based on data feature matching according to claim 6, characterized in that: The data acquisition module specifically includes: User authorization management unit, used to interact with each user terminal after obtaining user authorization; The data tracking unit is used to set data collection tracking points in various software on the user side to capture the user's interactive behavior and usage records, and record the current timestamp; The data integration unit is used to set up data integration points on the server side, to receive data collected by data collection points, to clean, filter, fill in missing fields, convert the format of the collected data, and merge them into user feature data sets, and store them in the user database.

8. The intelligent advertising delivery management system based on data feature matching according to claim 7, characterized in that: The comprehensive scoring module specifically includes: The advertisement delivery data collection unit is used to obtain the advertisement delivery data of the advertisement delivered in the current cycle; A bidding result score calculation unit, used to calculate the bidding result score of the advertisement based on the bidding result data and the delivery effect data; A fatigue index calculation module is used to collect the advertising delivery data in the current cycle and calculate the advertising fatigue index of the advertisement based on the delivery frequency data; An advertisement matching score calculation unit, used to calculate an advertisement matching score by combining advertisement matching data, advertisement delivery data and bidding result score; The priority sorting unit is used to prioritize advertisements according to advertisement matching scores and filter out pre-delivery advertisements according to score thresholds.

Images