Advertisement placement management system based on differentiated analysis of locations

Abstract

The present invention relates to the field of advertising delivery technology, and specifically to an advertising delivery location management system based on differentiated analysis locations, which includes a data collection module, a behavior pattern recognition module, a preference analysis module, a non-geographic advertising mode module, and a comparison and decision module, wherein: the data collection module collects user data at different geographical locations; the behavior pattern recognition module uses a clustering algorithm to recognize user behavior patterns based on user data at the same geographical location; the preference analysis module uses a support vector machine algorithm to recognize geographical advertising patterns based on user data at the same geographical location; the non-geographic advertising mode module uses a random clustering algorithm to predict the advertising mode corresponding to the user behavior mode, that is, the non-geographic advertising mode, based on the user behavior mode; the comparison and decision module compares the geographical advertising mode with the non-geographic advertising mode, and sets corresponding advertising content or optimizes the support vector machine algorithm according to the comparison result.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement placement, and in particular to an advertisement placement location management system based on differentiated analysis locations. Background Art

[0002] Traditional advertising delivery systems often face many disadvantages. First, in terms of data collection, traditional systems may lack accurate classification of user behavior data and geographic location annotation, resulting in poor data quality for subsequent analysis and difficulty in accurately grasping users' actual needs and behavior trends.

[0003] In addition, traditional systems often use simple statistical methods to analyze behavioral patterns, which lack pertinence and precision and cannot effectively identify users' specific behavioral patterns, resulting in inaccurate advertising targeting and poor results. In terms of advertising pattern recognition, traditional systems can often only simply judge the type of advertisement based on geographic location, lacking in-depth mining and understanding of user behavior, and cannot fully match advertising content with users' actual interests, reducing the relevance and attractiveness of advertisements.

[0004] Finally, in terms of advertising content generation, traditional systems often use fixed templates or simple rules and are unable to customize the content based on multi-dimensional information such as user behavior and geographic location, resulting in advertising content that lacks appeal and targeting, affecting advertising conversion rates and user satisfaction.

[0005] Therefore, traditional systems have many shortcomings in data collection, behavior pattern analysis, advertising pattern recognition and advertising content generation, and it is difficult to meet the advertising industry's needs for precise targeting and personalized marketing. There is an urgent need to improve the accuracy and effectiveness of advertising through an advertising placement location management system based on differentiated analysis locations. Summary of the invention

[0006] The purpose of the present invention is to provide an advertisement placement location management system based on differentiated analysis locations to solve the problems raised in the above background technology.

[0007] To achieve the above-mentioned purpose, the present invention provides the following technical solution: an advertising placement location management system based on differentiated analysis location, characterized in that it includes a data collection module, a behavior pattern recognition module, a preference analysis module, a non-geographic advertising model module and a comparison and decision module, wherein:

[0008] The data collection module collects user data at different geographical locations, wherein the user data is classified according to different geographical locations;

[0009] The behavior pattern recognition module recognizes user behavior patterns using a clustering algorithm based on user data at the same geographical location;

[0010] The preference analysis module uses a support vector machine algorithm to identify geographic advertising patterns based on user data in the same geographic location;

[0011] The non-geographic advertising mode module predicts the advertising mode corresponding to the user behavior mode using a random Senli algorithm according to the user behavior mode as the non-geographic advertising mode;

[0012] The comparison and decision module compares the geographic advertising pattern with the non-geographic advertising pattern, wherein:

[0013] If they are consistent, setting corresponding advertising content based on the user behavior pattern and the non-geographic advertising pattern, as well as the corresponding geographic location;

[0014] If there is inconsistency, the support vector machine algorithm in the preference analysis module is optimized and the optimized geographic advertising model is output. The optimized geographic advertising model is compared with the geographic advertising model. If there is still inconsistency, the corresponding advertising content is set based on the user behavior model and the geographic advertising model.

[0015] As a further improvement of the technical solution, the behavior pattern recognition module selects user data at the same geographical location, uses a clustering algorithm to recognize user behavior patterns, and implements clustering by assigning data points to K clusters and maximizing the similarity of data points within the clusters. The steps are as follows:

[0016] Input user feature vector and the number of clusters. User feature vector is represented as ,in is the user feature vector, and the number of clusters is ;

[0017] Random Initialization Cluster Centers , and each user feature vector Assign it to the cluster to which the nearest cluster center belongs, update the cluster center of each cluster to the average value of all data points in the cluster, and repeat the process until the cluster center no longer changes, where the subscript Represents the The index of the user feature vector, The range is from 1 to ;

[0018] Output clustering results, by dividing users into clusters, each cluster represents a user behavior pattern; based on the clustering results, different user behavior patterns are identified.

[0019] As a further improvement of the technical solution, the preference analysis module uses a support vector machine algorithm to identify geographic advertising patterns based on user data in the same geographic location, specifically including:

[0020] A radial basis function is selected to handle nonlinear problems, and labeled support vector machine training data is used. The support vector machine training data includes user feature vectors and corresponding advertising mode labels. The support vector machine model includes a regularization parameter C. During the training process, the radial basis function and the regularization parameter C are adjusted according to the support vector machine training data, and finally a decision boundary is determined, wherein the decision boundary is a boundary that determines the classification of data points, and data points of different categories are separated according to the training data;

[0021] Input the user data of the same geographic location into the trained support vector machine model, which will classify each user data into a specific advertising mode based on the learned decision boundary;

[0022] The classification results are counted, the number of users of each advertising mode is recorded, and the advertising mode with the largest number is identified as the geographic advertising mode.

[0023] As a further improvement of the technical solution, the non-geographic advertising mode module predicts the advertising mode corresponding to the user behavior mode using the random Senli algorithm according to the user behavior mode, as the non-geographic advertising mode, specifically including:

[0024] According to the user behavior pattern, the corresponding user consumption records are extracted, and the random forest algorithm will predict the corresponding advertising pattern based on the user consumption records;

[0025] Collect user consumption records and corresponding known advertising patterns from public databases and use them as labeled training datasets;

[0026] Training of the random forest model: Initialize the number of trees, tree depth and other parameters to optimize the model performance; Random Forest makes decisions by building multiple decision trees and uses a voting mechanism to determine the final classification result. The training data is fed into the random forest model for training. Each decision tree For an observation The classification result is , where each tree is constructed as follows:

[0027] Randomly select samples from the training data, randomly select a certain number of features at each decision point, build each tree independently without pruning, and the final classification result is determined by the prediction results of all trees through majority voting;

[0028] The prediction results of random forest are: ,in It is the majority function, which means that the most advertising mode is selected as the final result, and the non-geographic advertising mode is obtained.

[0029] As a further improvement of the technical solution, the comparison and decision module includes a comparison unit, and the comparison unit compares the geographic advertising mode with the non-geographic advertising mode. The comparison process is as follows:

[0030] Feature vector construction: Convert geographic advertising patterns and non-geographic advertising patterns into feature vectors and ;

[0031] Calculation of cosine similarity: Cosine similarity is defined by calculating the ratio of the dot product of two vectors and the product of their moduli. For two feature vectors and , the cosine similarity is expressed as:

[0032] , where the dot product It means that the elements of the vector are multiplied and summed. and are the Euclidean norm of the vector respectively;

[0033] Similarity threshold setting: Set a threshold to determine whether two advertising modes are consistent;

[0034] Comparison logic: If the cosine similarity is greater than or equal to the threshold, the two advertising models are considered to be consistent; otherwise, the comparison is inconsistent.

[0035] As a further improvement of the present technical solution, the comparison and decision module includes a decision unit, and the decision unit makes a decision according to the comparison result in the comparison unit, wherein:

[0036] If the comparison result is consistent, then the corresponding advertisement content is set based on the user behavior pattern in the behavior pattern recognition module and the non-geographic advertisement pattern in the non-geographic advertisement pattern module, as well as the geographic location selected by the behavior pattern recognition module;

[0037] If the comparison results are inconsistent, the support vector machine algorithm in the preference analysis module is optimized and the C parameter is adjusted by cross-validation.

[0038] As a further improvement of the technical solution, the decision unit sets corresponding advertising content based on the user behavior pattern and the non-geographic advertising pattern, as well as the geographic location, specifically including:

[0039] The user behavior pattern, non-geographic advertising pattern and geographic location are integrated into a comprehensive feature vector, and the formula is as follows:

[0040] ,in , and is a weight parameter used to adjust the influence of different features. is the feature fusion function, is the feature vector of the user behavior pattern, is the feature vector of the non-geographic advertising mode, is the feature vector of the geographic location;

[0041] Ad content generation: using functions According to the fused feature vector Generate advertising content, the formula is as follows:

[0042] ,in Including but not limited to a neural network algorithm model for generating targeted advertising copy based on an input feature vector.

[0043] As a further improvement of the technical solution, the decision unit adjusts the C parameter by cross-validation, specifically including:

[0044] Divide the training data set of the support vector machine: divide the data set into m mutually exclusive subsets, usually using m-fold cross-validation, each subset is called a fold, where m-1 folds are used as training sets and the remaining fold is used as a validation set;

[0045] Model training and validation: For each candidate value of C, use the data from m-1 folds to train the support vector machine model, validate it on the remaining fold, and calculate the performance indicators of the model on the validation set;

[0046] Performance evaluation: For each C, calculate its average performance index in m-fold cross validation;

[0047] Select the best parameter: Select C with the best performance index as the final parameter value.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. The advertising placement management system based on differentiated analysis location collects and classifies user data in different geographical locations, which provides accurate and specific user behavior information for subsequent analysis. It uses clustering algorithms to predict the behavior patterns of user data in the same geographical location, thereby identifying different user behavior patterns. This allows advertisers to better understand user behavior trends and design more personalized advertising strategies for them. It uses support vector machine algorithms to classify user data and identify geographic advertising patterns, further improving the relevance and attractiveness of advertising. It uses random forest algorithms to predict advertising patterns corresponding to user behavior patterns, which are not restricted by geographical locations, thereby providing more possibilities for subsequent verification processing. Such precise advertising positioning ensures that the advertising content is consistent with the actual interests and needs of users, improves the effectiveness of advertising, and also increases user satisfaction and loyalty.

[0050] 2. The advertising placement location management system based on differentiated analysis location is based on user behavior patterns, predicts the advertising pattern corresponding to the user behavior pattern, and as a non-geographic advertising pattern, identifies the consistency between the geographic advertising pattern and the non-geographic advertising pattern by comparing the two patterns. If the two patterns are consistent, it means that the advertising content is highly matched with the user's actual behavior and geographic location preferences, which may increase the click-through rate and conversion rate of the advertisement; if they are inconsistent, this comparison reveals the room for further analysis or adjustment to more accurately reach the target user group.

[0051] 3. The advertising placement management system based on differentiated analysis location is based on user behavior patterns and non-geographic advertising patterns, as well as corresponding geographic locations, to set corresponding advertising content. By integrating user behavior patterns, non-geographic advertising patterns and geographic locations into a comprehensive feature vector, targeted advertising copy is generated based on the integrated feature vector. The main advantage of this step is that it can create a highly personalized advertising experience. Based on user behavior patterns and non-geographic features as well as geographic locations, advertising content can be customized to meet the specific needs and preferences of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0053] Figure 2 It is a schematic diagram of the comparison and decision module unit of the present invention.

[0054] In the figure: 100, data collection module; 200, behavior pattern recognition module; 300, preference analysis module; 400, non-geographic advertising model module; 500, comparison and decision module; 501, comparison unit; 502, decision unit. DETAILED DESCRIPTION

[0055] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions 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.

[0056] See also Figure 1-2 The present invention provides a technical solution: an advertising placement location management system based on differentiated analysis locations, including a data collection module 100, a behavior pattern recognition module 200, a preference analysis module 300, a non-geographic advertising model module 400 and a comparison and decision module 500.

[0057] The data collection module 100 collects user data from different geographical locations and classifies the user data according to the geographical location. For example, assuming that user data from three geographical regions are collected: North America, Europe and Asia, the data of each user includes an identifier indicating its geographical location, that is, the data of users in North America includes the label "NA", the data of users in Europe includes the label "EU", and the data of users in Asia includes the label "AS"; the user data includes but is not limited to the history of browsing products, search queries, purchase behavior, page dwell time and geographical location.

[0058] The main function of the data collection module 100 is to obtain and classify user data in various geographical locations and provide basic data support. By accurately labeling and classifying user data, the module can ensure that subsequent analysis and decision-making are based on accurate and specific user behavior information.

[0059] The behavior pattern recognition module 200 selects user data at the same geographical location and uses a clustering algorithm to predict the user behavior pattern of the user data at the same geographical location, specifically including:

[0060] Preprocess the collected user data, including data cleaning, feature extraction, and data conversion, to ensure data quality and consistency, and convert the user data into feature vector representation;

[0061] Clustering is achieved by assigning data points to K clusters and maximizing the similarity of data points within the clusters. The steps are as follows:

[0062] Input user feature vector and the number of clusters. User feature vector is represented as ,in is the user feature vector, and the number of clusters is ;

[0063] Random Initialization Cluster Centers , and each user feature vector Assign it to the cluster to which the nearest cluster center belongs, update the cluster center of each cluster to the average value of all data points in the cluster, and repeat the process until the cluster center no longer changes, where the subscript Represents the The index of the user feature vector, The range is from 1 to ;

[0064] Output clustering results, that is, divide users into Clusters are formed, each cluster represents a user behavior pattern; based on the clustering results, different user behavior patterns are identified, such as shopping, dining, etc. These behavior patterns can provide important references for subsequent advertising targeting and personalized recommendations.

[0065] Through the above steps, the behavior pattern recognition module 200 can effectively use the clustering algorithm to identify the user behavior patterns of user data in the same geographical location, thereby providing a basis for advertising targeting and personalized recommendations. By analyzing the user data in the same geographical location through the clustering algorithm, specific user behavior patterns can be effectively identified. This process helps companies gain insight into the common behavior trends of user groups, such as shopping or dining habits, and allows advertisers to design more personalized advertising strategies based on these behavior patterns. By identifying and utilizing these behavior patterns, advertising content can be better connected with the actual interests and needs of users, thereby improving user engagement and advertising conversion rates.

[0066] The preference analysis module 300 uses a support vector machine algorithm to classify user data in the same geographical location to identify geographic advertising patterns, including:

[0067] Clean the collected user data, remove outliers and irrelevant features, and handle missing data; to ensure model performance, the data needs to be standardized, such as normalizing features such as age and income to eliminate the impact of different dimensions;

[0068] The radial basis function is selected to handle nonlinear problems, and the labeled support vector machine training data (including user feature vectors and corresponding advertising model labels) is used to train the support vector machine model. During the training process, the radial basis function and regularization parameter C are adjusted to optimize the model to achieve the best classification effect.

[0069] Input the user data of the same geographic location into the trained support vector machine model. The model will classify each user data into a specific advertising mode based on the learned decision boundary, for example, classifying users into "instant promotion advertising" or "long-term advertising";

[0070] The classification results are counted, the number of users of each advertising mode is recorded, and the advertising mode with the largest number is identified. For example, if "instant promotion advertising" has the most users, then this mode is the main advertising mode in this geographic location.

[0071] The preference analysis module 300 utilizes a support vector machine algorithm to classify user data and identify different advertising patterns. By analyzing user preferences and associating them with advertising patterns, the module is able to more accurately locate the types of advertisements that users may be interested in. This precise preference matching ensures the relevance and attractiveness of advertising content, thereby improving the effectiveness of advertising delivery and increasing user satisfaction and loyalty.

[0072] The non-geographic advertising model module 400 predicts the advertising model corresponding to the user behavior model using the random forest algorithm according to the user behavior model predicted by the behavior model recognition module 200 as the non-geographic advertising model, which specifically includes:

[0073] Random forest prediction data preparation: According to user behavior patterns, extract corresponding user consumption records, such as shopping and dining activity records. The random forest algorithm will predict the corresponding advertising model (non-geographic advertising model) based on user consumption records;

[0074] Random forest training data preparation: collect user consumption records and corresponding known advertising patterns from public databases and use them as labeled training data sets, that is, the input variables are user consumption records and the output variables are known advertising patterns;

[0075] Training of the random forest model: Initialize the number of trees, tree depth and other parameters to optimize the model performance; Random Forest makes decisions by building multiple decision trees and uses a voting mechanism to determine the final classification result. The training data is fed into the random forest model for training. Each decision tree For an observation The classification result is , where each tree is constructed as follows:

[0076] Randomly select samples from the training data, randomly select a certain number of features at each decision point, build each tree independently without pruning, and the final classification result is determined by the prediction results of all trees through majority voting;

[0077] The prediction results of random forest are: ,in It is the majority function, indicating that the most selected advertising mode is the final result; the non-geographic advertising mode is obtained by predicting the prediction data using the trained model.

[0078] The non-geographic advertising model module 400 predicts the advertising model corresponding to the user behavior model through the random forest algorithm, which is not restricted by the geographical location and is used for subsequent verification processing.

[0079] The comparison unit 501 in the comparison and decision module 500 compares the geographic advertising mode with the non-geographic advertising mode. The comparison process is as follows:

[0080] Feature vector construction: Convert geographic advertising patterns and non-geographic advertising patterns into feature vectors and ;

[0081] Calculation of cosine similarity: Cosine similarity is defined by calculating the ratio of the dot product of two vectors and the product of their moduli. For two feature vectors and , the cosine similarity is expressed as:

[0082] , where the dot product It means that the elements of the vector are multiplied and summed. and They are the Euclidean norm (i.e., modulus) of the vector;

[0083] Similarity threshold setting: Set a threshold, such as 0.8, to determine whether two advertising modes are consistent;

[0084] Comparison logic: If the cosine similarity is greater than or equal to the threshold, the two advertising modes are considered to be consistent; otherwise, the comparison is inconsistent; the comparison result is sent to the decision unit 502.

[0085] The decision unit 502 in the comparison and decision module 500 makes a decision according to the comparison result of the comparison unit 501, specifically including:

[0086] If the comparison result in the comparison unit 501 is consistent, then based on the user behavior pattern in the behavior pattern recognition module 200 and the non-geographic advertising pattern in the non-geographic advertising pattern module 400, and the geographical location selected by the behavior pattern recognition module 200, the corresponding advertising content is set, specifically including:

[0087] The user behavior pattern, non-geographic advertising pattern and geographic location are integrated into a comprehensive feature vector, and the formula is as follows:

[0088] ,in , and is a weight parameter used to adjust the influence of different features. is the feature fusion function, is the feature vector of the user behavior pattern, is the feature vector of the non-geographic advertising mode, is the feature vector of the geographic location;

[0089] Ad content generation: using functions According to the fused feature vector Generate advertising content, the formula is as follows:

[0090] ,in Including but not limited to a neural network algorithm model for generating targeted advertising copy based on an input feature vector.

[0091] If the comparison result received by the decision unit 502 is inconsistent, the support vector machine algorithm in the preference analysis module 300 is optimized and the C parameter is adjusted by cross-validation, specifically including:

[0092] Divide the training data set of the support vector machine: divide the data set into m mutually exclusive subsets, usually using m-fold cross-validation, each subset is called a fold, where m-1 folds are used as training sets and the remaining fold is used as a validation set;

[0093] Model training and validation: For each candidate value of C, use the data from m-1 folds to train the support vector machine model, validate it on the remaining fold, and calculate the performance indicators of the model on the validation set, such as accuracy, F1 score, etc.

[0094] Performance evaluation: For each C, calculate its average performance index in m-fold cross validation;

[0095] Select the best parameter: Select C with the best performance index as the final parameter value.

[0096] After optimizing the support vector machine algorithm in the preference analysis module 300, the decision unit 502 outputs the optimized geographic advertising model, compares the optimized geographic advertising model with the geographic advertising model, and if there is still inconsistency, sets the corresponding advertising content based on the user behavior pattern and the geographic advertising model. The process of setting the advertising content is consistent with the above logic.

[0097] 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 by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An advertising placement location management system based on differentiated analysis of locations, characterized in that: The system comprises a data collection module (100), a behavior pattern recognition module (200), a preference analysis module (300), a non-geographic advertising model module (400) and a comparison and decision module (500), wherein: The data collection module (100) collects user data at different geographical locations, wherein the user data is classified according to different geographical locations; The behavior pattern recognition module (200) recognizes user behavior patterns using a clustering algorithm based on user data at the same geographical location; The preference analysis module (300) uses a support vector machine algorithm to identify geographic advertising patterns based on user data at the same geographic location; The non-geographic advertising model module (400) predicts the advertising model corresponding to the user behavior model using a random forest algorithm according to the user behavior model as the non-geographic advertising model; The comparison and decision module (500) compares the geographic advertising model with the non-geographic advertising model, wherein: If they are consistent, setting corresponding advertising content based on the user behavior pattern and the non-geographic advertising pattern, as well as the corresponding geographic location; If they are inconsistent, the support vector machine algorithm in the preference analysis module (300) is optimized and the optimized geographic advertising model is output, and the optimized geographic advertising model is compared with the geographic advertising model. If they are still inconsistent, corresponding advertising content is set based on the user behavior model and the geographic advertising model; The comparison and decision module (500) comprises a comparison unit (501), wherein the comparison unit (501) compares the geographic advertising mode with the non-geographic advertising mode, and the comparison process is as follows: Feature vector construction: converting geographic advertising patterns and non-geographic advertising patterns into feature vectors vg and vng; Calculation of cosine similarity: Cosine similarity is defined by calculating the ratio of the dot product of two vectors and the product of their moduli. For two feature vectors vg and vng, cosine similarity is expressed as: The dot product vg·vng represents the sum of the corresponding elements of the vector after multiplication, and |vg| and |vng| are the Euclidean norms of the vectors respectively; Similarity threshold setting: Set a threshold to determine whether two advertising modes are consistent; Comparison logic: If the cosine similarity is greater than or equal to the threshold, the two advertising models are considered to be consistent; otherwise, the comparison is inconsistent; The comparison and decision module (500) comprises a decision unit (502), wherein the decision unit (502) makes a decision according to the comparison result in the comparison unit (501), wherein: If the comparison result is consistent, then based on the user behavior pattern in the behavior pattern recognition module (200) and the non-geographic advertising pattern in the non-geographic advertising pattern module (400), and the geographical location selected by the behavior pattern recognition module (200), corresponding advertising content is set; If the comparison result is inconsistent, the support vector machine algorithm in the preference analysis module (300) is optimized and the C parameter is adjusted by cross-validation; The decision unit (502) sets corresponding advertising content based on the user behavior pattern and the non-geographic advertising pattern, as well as the geographic location, specifically including: The user behavior pattern, non-geographic advertising pattern and geographic location are integrated into a comprehensive feature vector, and the formula is as follows: Z = g(U, vng, L) = αU + βvng + γL, where α, β and γ are weight parameters used to adjust the influence of different features, g is the feature fusion function, U is the feature vector of user behavior pattern, vng is the feature vector of non-geographic advertising pattern, and L is the feature vector of geographic location; Advertisement content generation: Use the function H(Z) to generate advertisement content based on the fused feature vector Z. The formula is as follows: Ad Content = H(Z), where H includes but is not limited to a neural network algorithm model, which is used to generate targeted advertising copy based on the input feature vector; The decision unit (502) adjusts the C parameter by cross-validation, specifically including: Divide the training data set of the support vector machine: divide the data set into m mutually exclusive subsets, usually using m-fold cross-validation, each subset is called a fold, where m-1 folds are used as training sets and the remaining fold is used as a validation set; Model training and validation: For each candidate value of C, use the data from m-1 folds to train the support vector machine model, validate it on the remaining fold, and calculate the performance indicators of the model on the validation set; Performance evaluation: For each C, calculate its average performance index in m-fold cross validation; Select the best parameter: select C with the best performance index as the final parameter value; The preference analysis module (300) uses a support vector machine algorithm to identify geographic advertising patterns based on user data at the same geographic location, specifically including: A radial basis function is selected to handle nonlinear problems, and labeled support vector machine training data is used. The support vector machine training data includes user feature vectors and corresponding advertising mode labels. The support vector machine model includes a regularization parameter C. During the training process, the radial basis function and the regularization parameter C are adjusted according to the support vector machine training data, and finally a decision boundary is determined, wherein the decision boundary is a boundary that determines the classification of data points, and data points of different categories are separated according to the training data; Input the user data of the same geographic location into the trained support vector machine model, which will classify each user data into a specific advertising mode based on the learned decision boundary; Count the classification results, record the number of users of each advertising mode, and identify the advertising mode with the largest number as the geographic advertising mode; The non-geographic advertising model module (400) predicts the advertising model corresponding to the user behavior model using a random forest algorithm according to the user behavior model as the non-geographic advertising model, specifically including: According to the user behavior pattern, the corresponding user consumption records are extracted, and the random forest algorithm will predict the corresponding advertising pattern based on the user consumption records; Collect user consumption records and corresponding known advertising patterns from public databases and use them as labeled training datasets; Training of the random forest model: Initialize the number of trees, tree depth and other parameters to optimize the model performance; Random Forest makes decisions by constructing multiple decision trees and uses a voting mechanism to determine the final classification result. The training data is fed into the random forest model for training. Each decision tree T i The classification result for an observation x is T i (x), where each tree is constructed as follows: Randomly select samples from the training data, randomly select a certain number of features at each decision point, build each tree independently without pruning, and the final classification result is determined by the prediction results of all trees through majority voting; The prediction result of random forest is: RF(x)=mode(T1(x),T2(x),...,T n (x)), where mode is the mode function, indicating that the most popular advertising mode is selected as the final result, resulting in a non-geographic advertising mode.

2. The advertising placement location management system based on differentiated analysis location according to claim 1, characterized in that: The behavior pattern recognition module (200) selects user data at the same geographical location, uses a clustering algorithm to recognize user behavior patterns, and implements clustering by allocating data points to K clusters and maximizing the similarity of data points within the clusters. The steps are as follows: Input user feature vector and the number of clusters. User feature vector is represented as (X=x1,x2,...,x n ), where x1,x2,...,x n is the user feature vector, and the number of clusters is K; Randomly initialize K cluster centers (μ1,μ2,...,μ K ), and each user feature vector x i Assign it to the cluster to which the nearest cluster center belongs, update the cluster center of each cluster to the average value of all data points in the cluster, and repeat the process until the cluster center no longer changes, where the subscript i represents the index of the i-th user feature vector, and i ranges from 1 to n; Output clustering results, by dividing users into K clusters, each cluster represents a user behavior pattern; Based on the clustering results, different user behavior patterns are identified.

Images