Advertisement putting management method and system based on big data comprehensive analysis

Abstract

The invention discloses an advertisement putting management method and system based on big data comprehensive analysis, and relates to the field of advertisement putting management.The method comprises the steps that feature items of purchase preference influence factors are extracted based on historical big data records fed back by advertisement putting, and a preference training sample list is established; on the basis of a machine learning algorithm model, purchase preferences are judged and predicted; historical records of different commodities purchased by each user are analyzed, and a purchased commodity flow set is established; on the basis of an Apriori algorithm, purchasing a preference judgment result according to a machine learning algorithm model, and recommending other advertisements associated with the preference advertisement; and updating and optimizing machine learning algorithm model parameters and a purchased commodity flow set according to an actual result fed back by advertisement putting. The system and the method have the advantages that intelligent delivery and effective delivery of the advertisement are realized, so that the advertisement delivery is more in line with the requirements of the public, and the acceptance of browsing the advertisement is increased.

Description

Technical Field

[0001] The present invention relates to the field of advertising delivery management, and in particular to an advertising delivery management method and system based on comprehensive analysis of big data. Background Art

[0002] The purpose of advertising is to let more groups know about the products produced by the company. It is one of the important channels for businesses to communicate with consumers in the market. Indiscriminate advertising will only cause visual fatigue in most groups, and create disgust for unnecessary product ads. It will also make the groups that really need the product lack more choices. Therefore, it is very necessary to accurately recommend related products and related product advertisements by analyzing the group's preference for different products.

[0003] Although existing technologies also use association rules to analyze product associations, the analysis of product associations based on association rules is based on the premise that users purchase products. Through the establishment of a previous associated product library, related associated products are pushed. This processing method is difficult to effectively analyze the preferences of the group, and it is impossible to provide accurate advertising during the user's browsing process. It lacks effective intelligent advertising delivery capabilities, cannot better serve the public, and is not conducive to the acceptance of advertisements by the public. Summary of the invention

[0004] In order to solve the above technical problems, a method and system for advertising delivery management based on comprehensive analysis of big data are provided. This technical solution solves the problem that the processing method proposed in the above background technology is difficult to effectively analyze the preferences of the group, cannot provide accurate advertising delivery during the user browsing process, lacks effective intelligent advertising delivery capabilities, cannot better serve the public, and is not conducive to the acceptance of advertisements by the public.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] An advertising delivery management method based on comprehensive analysis of big data, comprising:

[0007] Based on the historical big data records of advertising feedback, extract the characteristic items of factors affecting purchase preferences and establish a list of preference training samples;

[0008] According to the preference training sample list, based on the machine learning algorithm model, the purchase preference is judged and predicted;

[0009] Based on the historical big data records of advertising feedback, analyze the purchase history of each user for different products and establish a purchase flow set;

[0010] Based on the Apriori algorithm, other ads related to the preferred ads are recommended according to the purchase preference judgment results of the machine learning algorithm model;

[0011] Based on the actual results of advertising feedback, the machine learning algorithm model parameters and the purchase product flow set are updated and optimized.

[0012] Preferably, the extracting of characteristic items of factors influencing purchase preferences based on historical big data records of advertisement delivery feedback and establishing a preference training sample list specifically includes:

[0013] Determine the characteristic items of factors influencing purchase preference based on the events related to the advertisement viewing behavior, wherein the events related to the advertisement viewing behavior include: advertisement webpage browsing records, purchase records, number of searches for similar products, advertisement dwelling time, and number of clicks on reviews for similar products;

[0014] Based on the historical big data records of advertising feedback, the historical data of the characteristic items of factors affecting purchase preferences are extracted to establish a preference training sample list.

[0015] Preferably, the determining and predicting the purchase preference based on the preference training sample list and the machine learning algorithm model specifically includes:

[0016] According to the historical data of the characteristic items of the factors influencing the purchase preference, an n×m input matrix of the characteristic items of the factors influencing the purchase preference is established;

[0017] According to the number of characteristic items of factors affecting purchase preference, a 1×n parameter matrix θ is established, θ=θ1θ2θ3…θ i …θ n ], where θ i is the parameter value corresponding to the i-th feature item, and n is the number of feature items of factors influencing purchase preference;

[0018] According to the characteristic items of factors influencing purchase preference, a linear regression expression of purchase preference is constructed;

[0019] According to the gradient descent algorithm, the parameter update expression of the characteristic items of the factors affecting purchase preference is established;

[0020] Based on the large amount of data in the preferred training sample list, the machine learning algorithm model is learned and trained, and the parameters in the model are adjusted according to the actual purchase results, and the preferred training sample list is optimized;

[0021] By inputting the real-time value of the characteristic item of the factors influencing the purchase preference into the trained model, it is judged and predicted whether there is a purchase preference. If so, the advertisement of products with the same nature is placed to the customer. If not, the advertisement of products with the same nature is not placed.

[0022] The input matrix expression of the characteristic item n×m of the purchase preference influencing factor is:

[0023]

[0024] Where x is the n×m input matrix of the factors influencing purchase preference, nm is the historical data of the mth purchase preference influencing factor feature item of the nth feature item, n is the number of purchase preference influencing factor feature items, and m is the number of sample groups of the historical data of the purchase preference influencing factor feature items;

[0025] The purchase preference linear regression expression is:

[0026] f(x j )=θ j x j +δ

[0027] In the formula, f(x j ) is the purchase preference result corresponding to the historical data of the characteristic items of the purchase preference influencing factors of the jth group, θ j is the characteristic item parameter matrix corresponding to the historical data of the characteristic items of the j-th group of purchase preference influencing factors, x j is the historical data of the characteristic items of the factors affecting the purchase preference of the jth group, δ is a 1×m error constant term matrix, and the values ​​in the matrix are all the same;

[0028] The parameter update expression of the characteristic item of the purchase preference influencing factor is:

[0029]

[0030] In the formula, is the updated parameter value of the i-th feature item corresponding to the historical data of the j-th group of purchase preference influencing factor feature items, is the parameter value of the i-th feature item corresponding to the historical data of the feature item of the j-th group of purchase preference influencing factors, β is the learning rate, m is the number of sample groups of the historical data of the feature item of the purchase preference influencing factors, f is the actual purchase preference result corresponding to the historical data of the feature item of the j-th group of purchase preference influencing factors, is the historical data of the j-th group of purchase preference influencing factor feature items for the i-th feature item.

[0031] Preferably, the analysis of the historical records of each user's purchase of different commodities based on the historical big data records of advertisement delivery feedback and the establishment of a commodity purchase flow set specifically includes:

[0032] Based on the historical big data records of advertising feedback, obtain the historical records of each user purchasing different products within a period of time;

[0033] According to the attributes of different commodities purchased by each user, a commodity purchase flow set is established, wherein the commodity attribute refers to the type of item.

[0034] Preferably, the method of recommending other advertisements associated with the preferred advertisement based on the Apriori algorithm and the purchase preference determination result of the machine learning algorithm model specifically includes:

[0035] Based on the purchase flow of goods, calculate the support for users to purchase a certain product or multiple different products;

[0036] Based on the calculation results of the support for the user to purchase a certain product or multiple different products, calculate the confidence of the user purchasing multiple different products at the same time;

[0037] Calculate the lift of the two sets based on the confidence values ​​of users purchasing multiple different products at the same time;

[0038] According to the values ​​of confidence and lift, it is determined whether the lift value is greater than 1. If so, the association between the products is convincing; otherwise, the association between the products is not convincing.

[0039] Based on the Apriori algorithm results under big data, a product relevance prompt library is established. According to the purchase preference judgment results of the machine learning algorithm model, the related products of the product in the product relevance prompt library are retrieved, and other advertisements related to the preferred advertisements are recommended;

[0040] The support expression for the user to purchase a certain product or multiple different products is:

[0041]

[0042] Among them, Sup(X) is the support of the user purchasing a certain product or multiple different products, X is the set of the user purchasing a certain product or multiple different products, B is the number of groups purchasing a certain product or multiple different products in the purchase flow set, and D is the total number of groups in the purchase flow set;

[0043] The confidence expression for calculating the user's simultaneous purchase of multiple different commodities is:

[0044]

[0045] Where Con(X=>Y) is the confidence that the user purchases multiple different products at the same time, and Sup(XY) is the support that the user purchases products that satisfy both X and Y sets.

[0046] The expression for calculating the lifting degree of the two sets is:

[0047]

[0048] Preferably, the updating and optimizing of the machine learning algorithm model parameters and the purchase commodity flow set according to the actual results of the advertisement delivery feedback specifically includes:

[0049] By comparing the actual results of advertising feedback with the judgment and prediction results of the machine learning algorithm model, we can determine whether the purchase preference is accurate. If so, it means that the machine learning algorithm model is running normally. If not, the machine learning algorithm model parameters and training sets need to be updated and optimized.

[0050] By comparing the actual results of advertising delivery feedback with the results of the Apriori algorithm, it is determined whether the recommendations and other advertisements associated with the preferred advertisements are accurate. If so, it means that the purchase product flow set is relatively comprehensive. If not, the purchase product flow set needs to be updated and optimized.

[0051] Furthermore, the present solution proposes an advertising delivery management system based on comprehensive analysis of big data, which is used to implement the above-mentioned advertising delivery management method based on comprehensive analysis of big data, including:

[0052] A purchase preference judgment module, which is used to extract characteristic items of purchase preference influencing factors based on historical big data records of advertising delivery feedback, and establish a preference training sample list; and judge and predict purchase preferences based on the preference training sample list and a machine learning algorithm model;

[0053] The module for recommending related products is used to analyze the historical records of each user's purchase of different products based on the historical big data records of advertising feedback, and to establish a set of purchased product flow records; based on the Apriori algorithm, according to the purchase preference judgment results of the machine learning algorithm model, other advertisements associated with the preferred advertisements are recommended;

[0054] The feedback and optimization module is used to update and optimize the machine learning algorithm model parameters and the purchase commodity flow set according to the actual results of the advertisement delivery feedback.

[0055] Preferably, the purchase preference determination module includes:

[0056] Establishing a sample unit, wherein the sample unit is used to extract feature items of factors affecting purchase preferences based on historical big data records of advertising delivery feedback, and establish a preference training sample list;

[0057] A purchase preference determination unit is used to determine and predict purchase preferences based on a preference training sample list and a machine learning algorithm model.

[0058] Preferably, the related product recommendation module includes:

[0059] Establishing a flow set unit, the flow set establishing unit is used to analyze the historical records of each user purchasing different commodities based on the historical big data records of advertisement delivery feedback, and establish a commodity purchase flow set;

[0060] The recommended associated advertisement unit is used to recommend other advertisements associated with the preferred advertisements based on the Apriori algorithm and the purchase preference determination result of the machine learning algorithm model.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] Through big data analysis, the feature items of factors affecting purchase preferences are extracted. Based on the machine learning algorithm model, the feature items of factors affecting purchase preferences and a large amount of data on purchase results are trained and learned. Therefore, by receiving the data of the feature items of factors affecting purchase preferences, the purchase preferences are judged and predicted. According to the results of judgment and prediction, based on the association rules of the Apriori algorithm, other related products of credible services are pushed to the products with purchase preferences. By effectively analyzing the purchase preferences of the group, accurate advertising is given during the user's browsing process, realizing intelligent and effective delivery of advertisements, making the advertising delivery more in line with the needs of the public and increasing the acceptance of browsed advertisements. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 This is a flow chart of an advertising delivery management method based on comprehensive analysis of big data according to the present invention;

[0064] Figure 2 A flow chart of determining and predicting purchase preferences based on a preference training sample list and a machine learning algorithm model according to the present invention;

[0065] Figure 3 This is a flowchart of the present invention that recommends other advertisements associated with the preferred advertisements based on the Apriori algorithm and the purchase preference determination results of the machine learning algorithm model. DETAILED DESCRIPTION

[0066] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0067] Reference Figure 1 As shown, an advertising delivery management method based on comprehensive analysis of big data includes:

[0068] Based on the historical big data records of advertising feedback, extract the characteristic items of factors affecting purchase preferences and establish a list of preference training samples;

[0069] According to the preference training sample list, based on the machine learning algorithm model, the purchase preference is judged and predicted;

[0070] Based on the historical big data records of advertising feedback, analyze the purchase history of each user for different products and establish a purchase flow set;

[0071] Based on the Apriori algorithm, other ads related to the preferred ads are recommended according to the purchase preference judgment results of the machine learning algorithm model;

[0072] Based on the actual results of advertising feedback, the machine learning algorithm model parameters and the purchase product flow set are updated and optimized.

[0073] It can be understood that this solution is to extract the feature items of factors affecting purchase preferences through big data analysis, and train and learn the feature items of factors affecting purchase preferences and a large amount of data on purchase results based on the machine learning algorithm model, so as to judge and predict their purchase preferences by receiving the data of the feature items of factors affecting purchase preferences, and according to the results of judgment and prediction, based on the association rules of the Apriori algorithm, push other related products of credible services to the products with purchase preferences, and through the dynamic and static analysis of purchase preferences and related products, the accuracy and intelligence of advertising can be effectively improved.

[0074] Reference Figure 2 As shown, the determination and prediction of purchase preferences based on the preference training sample list and the machine learning algorithm model specifically include:

[0075] According to the historical data of the characteristic items of the factors influencing the purchase preference, an n×m input matrix of the characteristic items of the factors influencing the purchase preference is established;

[0076] According to the number of characteristic items of factors affecting purchase preference, a 1×n parameter matrix θ is established, θ=θ1θ2θ3…θ i …θ n ], where θ i is the parameter value corresponding to the i-th feature item, and n is the number of feature items of factors influencing purchase preference;

[0077] According to the characteristic items of factors influencing purchase preference, a linear regression expression of purchase preference is constructed;

[0078] According to the gradient descent algorithm, the parameter update expression of the characteristic items of the factors affecting purchase preference is established;

[0079] Based on the large amount of data in the preferred training sample list, the machine learning algorithm model is learned and trained, and the parameters in the model are adjusted according to the actual purchase results, and the preferred training sample list is optimized;

[0080] By inputting the real-time value of the characteristic item of the factors influencing the purchase preference into the trained model, it is judged and predicted whether there is a purchase preference. If so, the advertisement of products with the same nature is placed to the customer. If not, the advertisement of products with the same nature is not placed.

[0081] The input matrix expression of the characteristic item n×m of the purchase preference influencing factor is:

[0082]

[0083] Where x is the n×m input matrix of the factors influencing purchase preference, nm is the historical data of the mth purchase preference influencing factor feature item of the nth feature item, n is the number of purchase preference influencing factor feature items, and m is the number of sample groups of the historical data of the purchase preference influencing factor feature items;

[0084] The purchase preference linear regression expression is:

[0085] f(x j )=θ j x j +δ

[0086] In the formula, f(x j ) is the purchase preference result corresponding to the historical data of the characteristic items of the purchase preference influencing factors of the jth group, θ j is the characteristic item parameter matrix corresponding to the historical data of the characteristic items of the j-th group of purchase preference influencing factors, x j is the historical data of the characteristic items of the factors affecting the purchase preference of the jth group, δ is a 1×m error constant term matrix, and the values ​​in the matrix are all the same;

[0087] The parameter update expression of the characteristic item of the purchase preference influencing factor is:

[0088]

[0089] In the formula, is the updated parameter value of the i-th feature item corresponding to the historical data of the j-th group of purchase preference influencing factor feature items, is the parameter value of the i-th feature item corresponding to the historical data of the feature item of the j-th group of purchase preference influencing factors, β is the learning rate, m is the number of sample groups of the historical data of the feature item of the purchase preference influencing factors, f is the actual purchase preference result corresponding to the historical data of the feature item of the j-th group of purchase preference influencing factors, is the historical data of the j-th group of purchase preference influencing factor feature items for the i-th feature item.

[0090] It is understandable that by analyzing the linear relationship between advertising web page browsing records, purchase records, search times for similar products, ad dwell time, number of clicks on similar product reviews and purchase preferences, the gradient descent algorithm is used to reduce the error between the calculation results of the linear regression equation and the actual results, thereby maximizing the credibility of the calculation results of the linear regression equation, thereby establishing a machine learning algorithm model of the characteristic items of factors influencing purchase preferences and purchase preferences. Through real-time dynamic analysis of the model, the purchase preference is judged and predicted using the real-time received characteristic item values ​​of factors influencing purchase preferences, thereby achieving precise delivery of advertisements with purchase preferences.

[0091] Reference Figure 3 As shown, based on the Apriori algorithm, according to the purchase preference determination result of the machine learning algorithm model, the recommendation of other advertisements associated with the preferred advertisement specifically includes:

[0092] Based on the purchase flow of goods, calculate the support for users to purchase a certain product or multiple different products;

[0093] Based on the calculation results of the support for the user to purchase a certain product or multiple different products, calculate the confidence of the user purchasing multiple different products at the same time;

[0094] Calculate the lift of the two sets based on the confidence values ​​of users purchasing multiple different products at the same time;

[0095] According to the values ​​of confidence and lift, it is determined whether the lift value is greater than 1. If so, the association between the products is convincing; otherwise, the association between the products is not convincing.

[0096] Based on the Apriori algorithm results under big data, a product relevance prompt library is established. According to the purchase preference judgment results of the machine learning algorithm model, the related products of the product in the product relevance prompt library are retrieved, and other advertisements related to the preferred advertisements are recommended;

[0097] The support expression for the user to purchase a certain product or multiple different products is:

[0098]

[0099] Among them, Sup(X) is the support of the user purchasing a certain product or multiple different products, X is the set of the user purchasing a certain product or multiple different products, B is the number of groups purchasing a certain product or multiple different products in the purchase flow set, and D is the total number of groups in the purchase flow set;

[0100] The confidence expression for calculating the user's simultaneous purchase of multiple different commodities is:

[0101]

[0102] Where Con(X=>Y) is the confidence that the user purchases multiple different products at the same time, and Sup(XY) is the support that the user purchases products that satisfy both X and Y sets.

[0103] The expression for calculating the lifting degree of the two sets is:

[0104]

[0105] It can be understood that the iterative process of screening out product relevance based on the Apriori algorithm is as follows:

[0106] a. Calculate the support of all individual products with different attributes, and filter out the user purchase history records with a support greater than the minimum support, where the minimum support is a value derived from big data;

[0107] b. Based on the screening results in step a, calculate the support of all two products with different attributes, and filter out the user purchase history records with a support greater than the minimum support;

[0108] c. According to the screening results of step b, based on the theorem of the Apriori algorithm, pre-prune the screening results of step b to screen out the user purchase history records that meet the theorem, wherein the theorem of the Apriori algorithm is: if X is not k-1 frequent items, then it must not be a frequent k-item set. In this solution, the frequent item set refers to the set of products whose purchase support is greater than the minimum support;

[0109] d. Based on the screening results in step c, calculate the support of all three products with different attributes, and filter out the user purchase history records with a support greater than the minimum support;

[0110] e. Iterate in sequence until the maximum number of different commodity attributes purchased by the user is reached;

[0111] f. According to all the purchase commodity sets that meet the conditions after iteration, take out the true subsets of all sets, that is, the different combinations of all commodities excluding their own combinations;

[0112] g. Calculate the confidence and lift values ​​based on the true subsets of all sets, and judge whether the association between the products is convincing by comparing the lift value with 1. If it is greater than 1, the association between the products is convincing; if it is not greater than 1, the association between the products is not convincing;

[0113] Through big data analysis of advertised purchased products, we screen out all products with a lift value greater than 1, and establish a product relevance prompt library. By receiving product information with purchase preferences, we deliver related products based on the product relevance prompt library.

[0114] Further, with the same inventive concept as the advertising delivery management method based on comprehensive analysis of big data, this solution also proposes an advertising delivery management system based on comprehensive analysis of big data, including:

[0115] A purchase preference judgment module, which is used to extract characteristic items of purchase preference influencing factors based on historical big data records of advertising delivery feedback, and establish a preference training sample list; and judge and predict purchase preferences based on the preference training sample list and a machine learning algorithm model;

[0116] The module for recommending related products is used to analyze the historical records of each user's purchase of different products based on the historical big data records of advertising feedback, and to establish a set of purchased product flow records; based on the Apriori algorithm, according to the purchase preference judgment results of the machine learning algorithm model, other advertisements associated with the preferred advertisements are recommended;

[0117] A feedback and optimization module, which is used to update and optimize the machine learning algorithm model parameters and the purchase commodity flow set according to the actual results of the advertisement delivery feedback;

[0118] The purchase preference judgment module includes:

[0119] Establishing a sample unit, wherein the sample unit is used to extract feature items of factors affecting purchase preferences based on historical big data records of advertising delivery feedback, and establish a preference training sample list;

[0120] A purchase preference determination unit, the purchase preference determination unit is used to determine and predict purchase preferences based on a preference training sample list and a machine learning algorithm model;

[0121] The related product recommendation module includes:

[0122] Establishing a flow set unit, the flow set establishing unit is used to analyze the historical records of each user purchasing different commodities based on the historical big data records of advertisement delivery feedback, and establish a commodity purchase flow set;

[0123] The recommended associated advertisement unit is used to recommend other advertisements associated with the preferred advertisements based on the Apriori algorithm and the purchase preference determination result of the machine learning algorithm model.

[0124] To sum up, the advantages of the present invention are: through big data analysis, feature items of factors affecting purchase preferences are extracted, and based on a machine learning algorithm model, a large amount of data on the feature items of factors affecting purchase preferences and purchase results are trained and learned, so that by receiving the data of the feature items of factors affecting purchase preferences, the purchase preferences are judged and predicted, and according to the results of the judgment and prediction, based on the association rules of the Apriori algorithm, other related products of credible services are pushed to the goods with purchase preferences, and by effectively analyzing the purchase preferences of the group, accurate advertising is given during the user's browsing process, thereby realizing intelligent and effective delivery of advertisements, making the advertising delivery more in line with the needs of the public and increasing the acceptance of browsed advertisements.

[0125] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. An advertising delivery management method based on comprehensive analysis of big data, characterized in that: include: Based on the historical big data records of advertising feedback, extract the characteristic items of factors affecting purchase preferences and establish a list of preference training samples; According to the preference training sample list, based on the machine learning algorithm model, the purchase preference is judged and predicted; Based on the historical big data records of advertising feedback, analyze the purchase history of each user for different products and establish a purchase flow set; Based on the Apriori algorithm, other ads related to the preferred ads are recommended according to the purchase preference judgment results of the machine learning algorithm model; Based on the actual results of advertising feedback, the machine learning algorithm model parameters and the purchase product flow set are updated and optimized.

2. The advertising delivery management method based on comprehensive analysis of big data according to claim 1, characterized in that: The extracting of feature items of factors influencing purchase preferences based on historical big data records of advertising delivery feedback and establishing a preference training sample list specifically include: Determine the characteristic items of factors influencing purchase preference based on the events related to the advertisement viewing behavior, wherein the events related to the advertisement viewing behavior include: advertisement webpage browsing records, purchase records, number of searches for similar products, advertisement dwelling time, and number of clicks on reviews for similar products; Based on the historical big data records of advertising feedback, the historical data of the characteristic items of factors affecting purchase preferences are extracted to establish a preference training sample list.

3. The advertising delivery management method based on comprehensive analysis of big data according to claim 2, characterized in that: The determination and prediction of purchase preferences based on the preference training sample list and the machine learning algorithm model specifically include: According to the historical data of the characteristic items of the factors influencing the purchase preference, an n×m input matrix of the characteristic items of the factors influencing the purchase preference is established; According to the number of characteristic items of factors affecting purchase preference, a 1×n parameter matrix θ is established, θ=θ1θ2θ3…θ i …θ n ], where θ i is the parameter value corresponding to the i-th feature item, and n is the number of feature items of factors influencing purchase preference; According to the characteristic items of factors influencing purchase preference, a linear regression expression of purchase preference is constructed; According to the gradient descent algorithm, the parameter update expression of the characteristic items of the factors affecting purchase preference is established; Based on the large amount of data in the preferred training sample list, the machine learning algorithm model is learned and trained, and the parameters in the model are adjusted according to the actual purchase results, and the preferred training sample list is optimized; By inputting the real-time value of the characteristic item of the factors influencing the purchase preference into the trained model, it is judged and predicted whether there is a purchase preference. If so, the product advertisement with the same nature is placed to the customer. If not, the product advertisement with the same nature is not placed. The input matrix expression of the characteristic item n×m of the purchase preference influencing factor is: Where z is the n×m input matrix of the factors influencing purchase preference, x nm is the historical data of the mth purchase preference influencing factor feature item of the nth feature item, n is the number of purchase preference influencing factor feature items, and m is the number of sample groups of the historical data of the purchase preference influencing factor feature items; The purchase preference linear regression expression is: f(x j )=θ j x j +d In the formula, f(x j ) is the purchase preference result corresponding to the historical data of the characteristic items of the purchase preference influencing factors of the jth group, θ j is the characteristic item parameter matrix corresponding to the historical data of the characteristic items of the j-th group of purchase preference influencing factors, x j is the historical data of the characteristic items of the factors affecting the purchase preference of the jth group, δ is a 1×m error constant term matrix, and the values ​​in the matrix are all the same; The parameter update expression of the characteristic item of the purchase preference influencing factor is: In the formula, is the updated parameter value of the i-th feature item corresponding to the historical data of the j-th group of purchase preference influencing factor feature items, is the parameter value of the i-th feature item corresponding to the historical data of the feature item of the j-th group of purchase preference influencing factors, β is the learning rate, m is the number of sample groups of the historical data of the feature item of the purchase preference influencing factors, f is the actual purchase preference result corresponding to the historical data of the feature item of the j-th group of purchase preference influencing factors, is the historical data of the j-th group of purchase preference influencing factor feature items for the i-th feature item.

4. The advertising delivery management method based on comprehensive analysis of big data according to claim 3 is characterized in that: The historical big data records based on the feedback from advertisement delivery are used to analyze the historical records of each user purchasing different commodities and establish a commodity purchase flow set, which specifically includes: Based on the historical big data records of advertising feedback, obtain the historical records of each user purchasing different products within a period of time; According to the attributes of different commodities purchased by each user, a commodity purchase flow set is established, wherein the commodity attribute refers to the type of item.

5. The advertising delivery management method based on comprehensive analysis of big data according to claim 4 is characterized in that: The Apriori algorithm is used to determine the purchase preference based on the machine learning algorithm model, and the recommendation of other advertisements associated with the preferred advertisements specifically includes: Based on the purchase flow of goods, calculate the support for users to purchase a certain product or multiple different products; Based on the calculation results of the support for the user to purchase a certain product or multiple different products, calculate the confidence of the user purchasing multiple different products at the same time; Calculate the lift of the two sets based on the confidence values ​​of users purchasing multiple different products at the same time; According to the values ​​of confidence and lift, it is determined whether the lift value is greater than 1. If so, the association between the products is convincing; otherwise, the association between the products is not convincing. Based on the Apriori algorithm results under big data, a product relevance prompt library is established. According to the purchase preference judgment results of the machine learning algorithm model, the related products of the product in the product relevance prompt library are retrieved, and other advertisements related to the preferred advertisements are recommended; The support expression for the user to purchase a certain product or multiple different products is: Among them, Sup(X) is the support of the user purchasing a certain product or multiple different products, X is the set of the user purchasing a certain product or multiple different products, B is the number of groups purchasing a certain product or multiple different products in the purchase flow set, and D is the total number of groups in the purchase flow set; The confidence expression for calculating the user's simultaneous purchase of multiple different commodities is: Where Con(X=>Y) is the confidence that the user purchases multiple different products at the same time, and Sup(XY) is the support that the user purchases products that satisfy both X and Y sets. The expression for calculating the lifting degree of the two sets is:

6. The advertising delivery management method based on comprehensive analysis of big data according to claim 5, characterized in that: The updating and optimization of the machine learning algorithm model parameters and the purchase commodity flow set according to the actual results of the advertisement delivery feedback specifically includes: By comparing the actual results of advertising feedback with the judgment and prediction results of the machine learning algorithm model, we can determine whether the purchase preference is accurate. If so, it means that the machine learning algorithm model is running normally. If not, the machine learning algorithm model parameters and training sets need to be updated and optimized. By comparing the actual results of advertising delivery feedback with the results of the Apriori algorithm, it is determined whether the recommendations and other advertisements associated with the preferred advertisements are accurate. If so, it means that the purchase product flow set is relatively comprehensive. If not, the purchase product flow set needs to be updated and optimized.

7. An advertising delivery management system based on comprehensive analysis of big data, used to implement any one of the advertising delivery management methods based on comprehensive analysis of big data as claimed in claims 1-6, characterized in that: include: A purchase preference judgment module, which is used to extract characteristic items of purchase preference influencing factors based on historical big data records of advertising delivery feedback, and establish a preference training sample list; and judge and predict purchase preferences based on the preference training sample list and a machine learning algorithm model; The module for recommending related products is used to analyze the historical records of each user's purchase of different products based on the historical big data records of advertising feedback, and to establish a set of purchased product flow records; based on the Apriori algorithm, according to the purchase preference judgment results of the machine learning algorithm model, other advertisements associated with the preferred advertisements are recommended; The feedback and optimization module is used to update and optimize the machine learning algorithm model parameters and the purchase commodity flow set according to the actual results of the advertisement delivery feedback.

8. The advertising delivery management system based on comprehensive analysis of big data according to claim 7, characterized in that: The purchase preference judgment module specifically includes: Establishing a sample unit, wherein the sample unit is used to extract feature items of factors affecting purchase preferences based on historical big data records of advertising delivery feedback, and establish a preference training sample list; A purchase preference determination unit is used to determine and predict purchase preferences based on a preference training sample list and a machine learning algorithm model.

9. The advertising delivery management system based on comprehensive analysis of big data according to claim 8, characterized in that: The related product recommendation module specifically includes: Establishing a flow set unit, the flow set establishing unit is used to analyze the historical records of each user purchasing different commodities based on the historical big data records of advertisement delivery feedback, and establish a commodity purchase flow set; The recommended associated advertisement unit is used to recommend other advertisements associated with the preferred advertisements based on the Apriori algorithm and the purchase preference determination result of the machine learning algorithm model.

Images