A positioning advertising delivery management system and method based on big data

Abstract

The present invention discloses a positioning advertising delivery management system and method based on big data, belonging to the field of big data technology. The present invention obtains the geographical scope of the expected advertising delivery, obtains the delivery point information of all advertising delivery points within the geographical scope, and combines it with the relevant information of the advertising delivery points to form an advertising delivery point set; obtains the related product information of the advertising delivery, and uses principal component analysis to reduce the dimension of the related product information; based on the results of feature extraction and analysis, the customers are grouped into several customer groups, and a customer portrait model is established in combination with the proportion; feature engineering is performed on the advertising delivery point set, and a crowd portrait model of different advertising delivery points is established; the customer portrait model and the crowd portrait model are compared to obtain the best advertising delivery point; the correlation is evaluated according to the competitive advertising information and weather information of the best advertising delivery point; the best delivery time period is determined and combined with the best advertising delivery point to generate a positioning advertising delivery management plan.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a big data-based positioning advertising delivery management system and method. Background Art

[0002] With the development and application of big data technology, advertising management is facing more and more challenges and opportunities. Traditional advertising management methods often face problems such as insufficient information acquisition, difficulty in evaluating advertising effects, and difficulty in selecting advertising delivery points.

[0003] First, traditional advertising management often relies on limited data sources, making it difficult to obtain comprehensive and accurate information. Traditional advertising management mainly relies on market research and personal experience, lacking sufficient data support. Advertisers often find it difficult to obtain key market information, such as target audience characteristics, competitor intelligence, etc., resulting in a lack of targeting and effectiveness in advertising. Secondly, the traditional advertising effect evaluation method is relatively simple, and often can only evaluate the effect of advertising through some simple indicators, making it difficult to objectively and accurately evaluate the actual impact and return of advertising. This makes it difficult for advertisers to understand the real impact and return on investment of advertising, and difficult to make reasonable decisions. In addition, the traditional selection of advertising points often relies on experience and intuition, lacking a scientific and systematic method to select the best advertising points. The traditional selection of advertising points usually only takes into account factors such as geographical location, while ignoring important factors such as the characteristics of the target audience, competition, and environmental factors, resulting in uncertainty and instability in the effect of advertising. Summary of the invention

[0004] The purpose of the present invention is to provide a big data-based positioning advertising delivery management system and method to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A method for managing positioning advertising delivery based on big data, the method comprising the following steps:

[0007] S100, obtaining the geographical range where the advertiser plans to place advertisements, obtaining the advertisement placement point information of all advertisement placement points within the geographical range, wherein the advertisement placement point information includes the name and longitude and latitude coordinates, and constitutes an advertisement placement point set; obtaining corresponding traffic information, crowd portrait information, crowd behavior data, competitive advertisement information, and weather information according to the information of each advertisement placement point in the advertisement placement point set, and adding the obtained information to the advertisement placement point set;

[0008] S200, obtaining information about related products of advertisements provided by advertisers, wherein different information is obtained depending on whether the related products are new products, and principal component analysis is used to reduce the dimension of the related product information;

[0009] S300, based on the results of principal component analysis, using K-means clustering to group customers into several customer groups with similar characteristics according to related product information, and obtaining the proportion of corresponding customer groups to establish a customer portrait model;

[0010] S400, feature engineering is performed on the set of advertising placement points, and crowd portrait models of different advertising placement points are established using K-means clustering; the customer portrait model and the crowd portrait model are compared by calculating a comprehensive similarity index to obtain the best advertising placement point;

[0011] S500. Based on the competitive advertising information and weather information of the best advertising delivery point, use the Pearson correlation coefficient to evaluate the correlation between the competitive advertising information and weather information and the advertising effect respectively; use the regression analysis model to determine the time period when the competitive advertising and weather have the least impact on the advertising delivery, and combine it with the best advertising delivery point to generate a targeted advertising delivery management plan.

[0012] According to step S100, the presence of human body is sensed by ultrasonic wave through the crowd flow detector installed at the advertising point to obtain crowd flow information; the face position and area in the image are detected by using the YOLO algorithm through the camera installed at the advertising point, and the identity, age, gender, clothing and appearance of the face are recognized in combination with the face recognition model DeepFace, and the crowd portrait information is merged; the behavior and action trajectory of the crowd are obtained through the camera installed at the advertising point through the behavior recognition algorithm based on recurrent neural network and the trajectory tracking algorithm based on Kalman filter, and the crowd behavior data is merged; the competitive advertising information is obtained through the field survey of personnel and input into the system to form the competitive advertising information; the weather information is obtained by accessing the meteorological data API provided by the meteorological department; the obtained information is added to the advertising point set, and the advertising point set is classified according to different advertising points.

[0013] The preprocessed image is input into the YOLO algorithm for processing. The YOLO algorithm is a real-time target detection algorithm based on deep learning. It can accurately detect the position and area of ​​the face in the image. Through the YOLO algorithm, all faces in the image can be quickly detected and located. The face area detected by the YOLO algorithm is extracted and then input into the face recognition model DeepFace for recognition. DeepFace is a face recognition model based on deep learning that can accurately recognize the identity, age, gender, and appearance features of the face.

[0014] The facial information obtained by DeepFace recognition, such as identity, age, gender, clothing, appearance and other features, is associated with the location and area of ​​the face to form a complete crowd portrait information, which can be saved in the database.

[0015] Step S200 includes:

[0016] S201. If the associated product provided by the advertiser for advertising is a new product, the associated product information obtained includes product features, market competition, target audience characteristics and expected sales data;

[0017] S202: If the associated product provided by the advertiser for advertising is an old product, the associated product information obtained includes historical sales data, customer survey data, social media data, and market competition conditions;

[0018] S203. According to the associated product information, a data set X containing n samples and m features is obtained, where represents the jth eigenvalue of the i-th sample; the data in the dataset X needs to be standardized so that the mean of each feature is 0 and the variance is 1. The standardization formula is as follows:

[0019] ;

[0020] Among them, the mean of the jth feature is , the standard deviation is , is the standardized value of the jth feature of the i-th sample;

[0021] S204, the standardized data set Calculate the covariance matrix , whose elements are:

[0022] ;

[0023] in, is the normalized value of the i-th feature of the k-th sample, is the normalized value of the jth feature of the kth sample;

[0024] S205, covariance matrix Perform eigenvalue decomposition to obtain eigenvalues ​​and corresponding eigenvectors;

[0025] S206, performing principal component analysis, selecting eigenvectors corresponding to several most important eigenvalues, and forming a feature space after dimensionality reduction;

[0026] S207: Project the original data X to the new principal component space to obtain the data matrix Y after dimensionality reduction.

[0027] According to step S206, when performing principal component analysis, the eigenvalue represents the variance explained by each principal component; the proportion of each eigenvalue to the total variance is calculated, and then the number of principal components to be retained is selected according to a set threshold, and the threshold is the percentage of the total variance retained, which is set by the user; finally, the principal component that can explain most of the variance is selected to be retained.

[0028] Step S300 includes:

[0029] S301, randomly initialize r cluster centers, randomly select r samples from the data matrix Y as initial cluster centers, and the initial cluster centers will be used as initial customer group feature vectors;

[0030] S302, for each customer group feature vector, calculate its distance from each cluster center, and assign it to the customer group represented by the cluster center closest to it, so as to group customers into customer groups with similar features;

[0031] S303, for each customer group, recalculate the mean of its sample feature vector, and use the mean as the new cluster center to update the feature vector of the customer group;

[0032] S304, repeating steps S302-S303 until the cluster center of the customer group no longer changes and the clustering converges, thereby obtaining a clustering result of the customer group based on the associated product information;

[0033] S305. Count the proportion of each customer group based on the clustering results.

[0034] According to steps S301-S304, the goal of K-means clustering is to minimize the sum of the squares of the distances between each sample and its cluster center, which is expressed by the following formula:

[0035] ;

[0036] Among them, J is the objective function of K-means clustering; is the indicator function, indicating that the sample Is it a cluster center? ; K-means clustering updates the cluster centers by iteration and , until convergence;

[0037] According to step S305, the proportion of each customer group is the number of customers in each customer group divided by the total number of customers.

[0038] Step S400 includes:

[0039] S401, using principal component analysis, performing feature engineering based on the traffic information, crowd portrait information and crowd behavior data in the set of advertising placement points;

[0040] S402, using K-means clustering to establish crowd portrait models for different advertising delivery points;

[0041] S403, using the established customer portrait model and the crowd portrait model of each advertisement delivery point to respectively describe crowd characteristics and customer characteristics of different advertisement delivery points;

[0042] S404. For each advertisement delivery point, the similarity between its crowd portrait model and the customer portrait model is calculated using cosine similarity, Euclidean distance and Jaccard similarity coefficient, and the average of the three similarity calculation results is taken as a comprehensive similarity index;

[0043] S405: sort each advertisement placement point by the proportion of its customer group, where a higher proportion means that the advertisement placement point covers more customer groups;

[0044] S406, evaluating each advertising placement point based on the comprehensive similarity index and customer group ranking of each advertising placement point, using a weighted average algorithm to weight the similarity index and the proportion to obtain a score for each advertising placement point;

[0045] S407: Select the advertisement delivery point with the highest score as the best advertisement delivery point.

[0046] According to step S500, the Pearson correlation coefficient is used to evaluate the correlation between the competitive advertising information and the weather information and the advertising effect, wherein the advertising effect is specifically quantified by the average value of each conversion of the advertisement; for the competitive advertising information, the Pearson correlation coefficient The calculation formula is as follows:

[0047] ;

[0048] in, represents the sample value of the competing advertisement information, A sample value representing the advertising effect. represents the mean of the data of competing advertising information, The mean of the data representing the advertising effect, Represents the product of the standard deviations between competing advertising information A and advertising effectiveness B.

[0049] According to step S500, a multivariate linear regression model is used to describe the relationship between competitive advertising information, weather information and advertising effect. The model is expressed as:

[0050]

[0051] Among them, B is the advertising effect, A is the competitive advertising information, and A' is the weather information. is the error term, , and are the coefficients of the multiple linear regression model, is the correlation between competitive advertising information and advertising effectiveness, is the correlation between weather information and advertising effectiveness, and Need to be obtained in the training of the multivariate linear regression model;

[0052] Divide the set of advertising points into training set and test set, select features for training from the training set, use the training set data to train the linear regression model, and find the best coefficients through the square loss function , minimize the error between the predicted value and the actual value; use the test set data to evaluate the trained model, use the mean square error to calculate the evaluation result, and tune the multivariate linear regression model according to the evaluation result;

[0053] The established regression model is used to analyze the impact of competitive advertising information and weather information on advertising effectiveness, and to determine the time period with the least impact; the determined time period with the least impact is combined with the optimal advertising delivery point to generate a targeted advertising delivery management plan.

[0054] A positioning advertising delivery management system based on big data, comprising:

[0055] Geographic information acquisition module: including: geographic range acquisition unit and advertisement placement point information acquisition unit; wherein the geographic range acquisition unit is responsible for acquiring geographic range information of the expected advertisement placement from the advertiser, and the advertisement placement point information acquisition unit is responsible for acquiring the names and latitude and longitude coordinate information of all advertisement placement points within the geographic range;

[0056] Data acquisition and integration module: including: a traffic information acquisition unit, a crowd portrait information acquisition unit, a crowd behavior data acquisition unit, a competitive advertising information acquisition unit, a weather information acquisition unit and an information integration unit; wherein the traffic information acquisition unit is responsible for acquiring the traffic information of each advertising delivery point from relevant data sources, the crowd portrait information acquisition unit is responsible for acquiring the crowd portrait information of each advertising delivery point, the crowd behavior data acquisition unit is responsible for acquiring the crowd behavior data of each advertising delivery point, the competitive advertising information acquisition unit is responsible for acquiring relevant information of competitive advertising, the weather information acquisition unit is responsible for acquiring the weather information of the area where each advertising delivery point is located, and the information integration unit is responsible for integrating the various types of information acquired into the set of advertising delivery points;

[0057] Related product information processing module: including: related product information acquisition unit, new product information processing unit, old product information processing unit and principal component analysis unit; wherein the related product information acquisition unit is responsible for acquiring related product information provided by advertisers for advertisement placement, the new product information processing unit performs corresponding processing and analysis based on the new product information, the old product information processing unit performs corresponding processing and analysis based on the old product information, and the principal component analysis unit is responsible for performing principal component analysis on the related product information, reducing the dimension and performing feature extraction and analysis;

[0058] Customer clustering module: including: a first K-means clustering unit and a customer portrait model building unit; wherein the first K-means clustering unit uses K-means clustering to group customers into groups with similar characteristics and determine their proportions, and the customer portrait model building unit is responsible for building a customer portrait model according to the processing results of the first K-means clustering unit;

[0059] Advertisement delivery point module: including: an advertisement delivery point feature extraction unit, a second K-means clustering unit and an optimal advertisement delivery point acquisition unit; wherein the advertisement delivery point feature extraction unit extracts and analyzes features of data in the advertisement delivery point set, the second K-means clustering unit uses K-means clustering to group advertisement delivery points and establish crowd portrait models for different advertisement delivery points; the optimal advertisement delivery point acquisition unit compares the customer portrait model with the crowd portrait model by calculating a comprehensive similarity index to obtain the optimal advertisement delivery point;

[0060] Positioning advertising delivery management module: including a Pearson correlation coefficient calculation unit, a regression analysis modeling unit and a positioning advertising delivery management plan generation unit; wherein, the Pearson correlation coefficient calculation unit is responsible for calculating the correlation between competing advertising information and weather information and advertising effects, the regression analysis modeling unit is responsible for establishing a mathematical model between competing advertising information and weather information and advertising effects, and determining the delivery time period with the least impact, and the positioning advertising delivery management plan generation unit is responsible for combining the time period with the least impact of competing advertising and weather on advertising delivery with the optimal advertising delivery point to generate a positioning advertising delivery management plan.

[0061] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0062] The present invention uses big data technology and data analysis methods, comprehensively considers multiple factors such as geographic information, crowd portraits, competitive advertising information and weather information, and can accurately select the best advertising placement point. Compared with the traditional method of selecting advertising placement points based on experience and intuition, the present invention can determine the most suitable placement location more scientifically and systematically, thereby improving the advertising effect.

[0063] The present invention extracts and analyzes the features of related product information, uses methods such as K-means clustering to group customers into groups with similar features, and establishes a customer portrait model. Compared with traditional market research and personal experience, the present invention can more accurately understand the characteristics and behaviors of the target customer group and provide a more targeted strategy for advertising.

[0064] The present invention analyzes competitive advertising information and weather information, and uses methods such as Pearson correlation coefficient and regression analysis to establish a mathematical model, which can objectively and accurately evaluate the advertising effect and determine the best time period for advertising. Compared with the traditional simple evaluation method, the present invention can more comprehensively consider the factors affecting the advertising effect, thereby optimizing the advertising delivery strategy and improving the efficiency of advertising delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0066] Figure 1 It is a schematic diagram of the steps of a method for managing positioning advertising based on big data of the present invention;

[0067] Figure 2 It is a system structure diagram of a positioning advertising delivery management system based on big data of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of 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.

[0069] See also Figure 1 and Figure 2 , the present invention provides a technical solution:

[0070] According to one embodiment of the present invention, Figure 1 A step diagram of a method for managing positioning advertising delivery based on big data is shown in the figure. A method for managing positioning advertising delivery based on big data comprises the following steps:

[0071] S100, obtaining the geographical range where the advertiser plans to place advertisements, obtaining the advertisement placement point information of all advertisement placement points within the geographical range, wherein the advertisement placement point information includes the name and longitude and latitude coordinates, and constitutes an advertisement placement point set; obtaining corresponding traffic information, crowd portrait information, crowd behavior data, competitive advertisement information, and weather information according to the information of each advertisement placement point in the advertisement placement point set, and adding the obtained information to the advertisement placement point set;

[0072] S200, obtaining information about related products of advertisements provided by advertisers, wherein different information is obtained depending on whether the related products are new products, and principal component analysis is used to reduce the dimension of the related product information;

[0073] S300, based on the results of principal component analysis, using K-means clustering to group customers into several customer groups with similar characteristics according to related product information, and obtaining the proportion of corresponding customer groups to establish a customer portrait model;

[0074] S400, feature engineering is performed on the set of advertising placement points, and crowd portrait models of different advertising placement points are established using K-means clustering; the customer portrait model and the crowd portrait model are compared by calculating a comprehensive similarity index to obtain the best advertising placement point;

[0075] S500. Based on the competitive advertising information and weather information of the best advertising delivery point, use the Pearson correlation coefficient to evaluate the correlation between the competitive advertising information and weather information and the advertising effect respectively; use the regression analysis model to determine the time period when the competitive advertising and weather have the least impact on the advertising delivery, and combine it with the best advertising delivery point to generate a targeted advertising delivery management plan.

[0076] According to step S100, the presence of human body is sensed by ultrasonic wave through the crowd flow detector installed at the advertising point to obtain crowd flow information; the face position and area in the image are detected by using the YOLO algorithm through the camera installed at the advertising point, and the identity, age, gender, clothing and appearance of the face are recognized in combination with the face recognition model DeepFace, and the crowd portrait information is merged; the behavior and action trajectory of the crowd are obtained through the camera installed at the advertising point through the behavior recognition algorithm based on recurrent neural network and the trajectory tracking algorithm based on Kalman filter, and the crowd behavior data is merged; the competitive advertising information is obtained through the field survey of personnel and input into the system to form the competitive advertising information; the weather information is obtained by accessing the meteorological data API provided by the meteorological department; the obtained information is added to the advertising point set, and the advertising point set is classified according to different advertising points.

[0077] Step S200 includes:

[0078] S201. If the associated product provided by the advertiser for advertising is a new product, the associated product information obtained includes product features, market competition, target audience characteristics and expected sales data;

[0079] In this embodiment, the new product provided by the advertiser is a smart bracelet. Some of the related product information we can obtain is as follows: Product features: The smart bracelet has functions such as heart rate monitoring, sleep monitoring, and step counting; the design is fashionable and lightweight, suitable for wearing during sports; it has functions such as waterproof and dustproof, and is suitable for various outdoor activities; Market competition: Competitor A's smart bracelet is priced lower, but has fewer functions, and is mainly aimed at price-sensitive consumers; Competitor B's smart bracelet has more functions, but is more expensive, and is mainly aimed at consumers who pursue quality and technology; Target audience characteristics: The age group is mainly concentrated between 25 and 40 years old, and they are young people who pay attention to health and fashion; they prefer outdoor sports and fitness exercises, and pay attention to their own health status and sports data; Expected sales data: The sales volume in the first month after listing is expected to be 10,000 units. It is expected that after market promotion and promotion activities, the sales volume will gradually increase and reach a stable level.

[0080] S202: If the associated product provided by the advertiser for advertising is an old product, the associated product information obtained includes historical sales data, customer survey data, social media data, and market competition conditions;

[0081] S203. According to the associated product information, a data set X containing n samples and m features is obtained, where represents the jth eigenvalue of the i-th sample; the data in the dataset X needs to be standardized so that the mean of each feature is 0 and the variance is 1. The standardization formula is as follows:

[0082] ;

[0083] Among them, the mean of the jth feature is , the standard deviation is , is the standardized value of the jth feature of the i-th sample;

[0084] S204, the standardized data set Calculate the covariance matrix , whose elements are:

[0085] ;

[0086] in, is the normalized value of the i-th feature of the k-th sample, is the normalized value of the jth feature of the kth sample;

[0087] S205, covariance matrix Perform eigenvalue decomposition to obtain eigenvalues ​​and corresponding eigenvectors;

[0088] S206, performing principal component analysis, selecting eigenvectors corresponding to several most important eigenvalues, and forming a feature space after dimensionality reduction;

[0089] S207: Project the original data X to the new principal component space to obtain the data matrix Y after dimensionality reduction.

[0090] According to step S206, when performing principal component analysis, the eigenvalue represents the variance explained by each principal component; the proportion of each eigenvalue in the total variance is calculated, and then the number of principal components to be retained is selected according to the set threshold, and the threshold is the percentage of the total variance retained. The threshold we set is 90% of the total variance; finally, the principal component that can explain most of the variance is selected.

[0091] Step S300 includes:

[0092] S301, randomly initialize 5 cluster centers, randomly select 5 samples from the data matrix Y as initial cluster centers, and the initial cluster centers will be used as initial customer group feature vectors;

[0093] S302, for each customer group feature vector, calculate its distance from each cluster center, and assign it to the customer group represented by the cluster center closest to it, so as to group customers into customer groups with similar features;

[0094] S303, for each customer group, recalculate the mean of its sample feature vector, and use the mean as the new cluster center to update the feature vector of the customer group;

[0095] S304, repeating steps S302-S303 until the cluster center of the customer group no longer changes and the clustering converges, thereby obtaining a clustering result of the customer group based on the associated product information;

[0096] S305. Count the proportion of each customer group based on the clustering results.

[0097] According to steps S301-S304, the goal of K-means clustering is to minimize the sum of the squares of the distances between each sample and its cluster center, which is expressed by the following formula:

[0098] ;

[0099] Among them, J is the objective function of K-means clustering; is the indicator function, indicating that the sample Is it a cluster center? ; K-means clustering updates the cluster centers by iteration and , until convergence;

[0100] According to step S305, the proportion of each customer group is the number of customers in each customer group divided by the total number of customers. The total number of customers is 1000, and now we have divided it into 5 customer groups, and the number of customers in each group is 200, 250, 150, 300 and 100 respectively. Then the proportion of each customer group is:

[0101] Customer group 1 accounts for 20%, customer group 2 accounts for 25%, customer group 3 accounts for 15%, customer group 4 accounts for 30%, and customer group 5 accounts for 10%.

[0102] Step S400 includes:

[0103] S401, using principal component analysis, performing feature engineering based on the traffic information, crowd portrait information and crowd behavior data in the set of advertising placement points;

[0104] S402, using K-means clustering to establish crowd portrait models for different advertising delivery points;

[0105] S403, using the established customer portrait model and the crowd portrait model of each advertisement delivery point to respectively describe crowd characteristics and customer characteristics of different advertisement delivery points;

[0106] S404. For each advertisement delivery point, the similarity between its crowd portrait model and the customer portrait model is calculated using cosine similarity, Euclidean distance and Jaccard similarity coefficient, and the average of the three similarity calculation results is taken as a comprehensive similarity index;

[0107] S405: sort each advertisement placement point by the proportion of its customer group, where a higher proportion means that the advertisement placement point covers more customer groups;

[0108] S406, evaluating each advertising placement point based on the comprehensive similarity index and customer group ranking of each advertising placement point, using a weighted average algorithm to weight the similarity index and the proportion to obtain a score for each advertising placement point;

[0109] S407: Select the advertisement delivery point with the highest score as the best advertisement delivery point.

[0110] According to step S500, the Pearson correlation coefficient is used to evaluate the correlation between the competitive advertising information and the weather information and the advertising effect, wherein the advertising effect is specifically quantified by the average value of each conversion of the advertisement; for the competitive advertising information, the Pearson correlation coefficient The calculation formula is as follows:

[0111] ;

[0112] in, represents the sample value of the competing advertisement information, A sample value representing the advertising effect. represents the mean of the data of competing advertising information, The mean of the data representing the advertising effect, Represents the product of the standard deviations between competing advertising information A and advertising effectiveness B.

[0113] According to step S500, a multivariate linear regression model is used to describe the relationship between competitive advertising information, weather information and advertising effect. The model is expressed as:

[0114]

[0115] Among them, B is the advertising effect, A is the competitive advertising information, and A' is the weather information. is the error term, , and are the coefficients of the multiple linear regression model, is the correlation between competitive advertising information and advertising effectiveness, is the correlation between weather information and advertising effectiveness, and Need to be obtained in the training of the multivariate linear regression model;

[0116] Divide the set of advertising points into training set and test set, select features for training from the training set, use the training set data to train the linear regression model, and find the best coefficients through the square loss function , minimize the error between the predicted value and the actual value; use the test set data to evaluate the trained model, use the mean square error to calculate the evaluation result, and tune the multivariate linear regression model according to the evaluation result;

[0117] The established regression model is used to analyze the impact of competitive advertising information and weather information on advertising effectiveness, and to determine the time period with the least impact; the determined time period with the least impact is combined with the optimal advertising delivery point to generate a targeted advertising delivery management plan.

[0118] According to another embodiment of the present invention, Figure 2 As shown in the system structure diagram of a positioning advertising delivery management system based on big data, a positioning advertising delivery management system based on big data includes:

[0119] Geographic information acquisition module: including: geographic range acquisition unit and advertisement placement point information acquisition unit; wherein the geographic range acquisition unit is responsible for acquiring geographic range information of the expected advertisement placement from the advertiser, and the advertisement placement point information acquisition unit is responsible for acquiring the names and latitude and longitude coordinate information of all advertisement placement points within the geographic range;

[0120] Data acquisition and integration module: including: a traffic information acquisition unit, a crowd portrait information acquisition unit, a crowd behavior data acquisition unit, a competitive advertising information acquisition unit, a weather information acquisition unit and an information integration unit; wherein the traffic information acquisition unit is responsible for acquiring the traffic information of each advertising delivery point from relevant data sources, the crowd portrait information acquisition unit is responsible for acquiring the crowd portrait information of each advertising delivery point, the crowd behavior data acquisition unit is responsible for acquiring the crowd behavior data of each advertising delivery point, the competitive advertising information acquisition unit is responsible for acquiring relevant information of competitive advertising, the weather information acquisition unit is responsible for acquiring the weather information of the area where each advertising delivery point is located, and the information integration unit is responsible for integrating the various types of information acquired into the set of advertising delivery points;

[0121] Related product information processing module: including: related product information acquisition unit, new product information processing unit, old product information processing unit and principal component analysis unit; wherein the related product information acquisition unit is responsible for acquiring related product information provided by advertisers for advertisement placement, the new product information processing unit performs corresponding processing and analysis based on the new product information, the old product information processing unit performs corresponding processing and analysis based on the old product information, and the principal component analysis unit is responsible for performing principal component analysis on the related product information, reducing the dimension and performing feature extraction and analysis;

[0122] Customer clustering module: including: a first K-means clustering unit and a customer portrait model building unit; wherein the first K-means clustering unit uses K-means clustering to group customers into groups with similar characteristics and determine their proportions, and the customer portrait model building unit is responsible for building a customer portrait model according to the processing results of the first K-means clustering unit;

[0123] Advertisement delivery point module: including: an advertisement delivery point feature extraction unit, a second K-means clustering unit and an optimal advertisement delivery point acquisition unit; wherein the advertisement delivery point feature extraction unit extracts and analyzes features of data in the advertisement delivery point set, the second K-means clustering unit uses K-means clustering to group advertisement delivery points and establish crowd portrait models for different advertisement delivery points; the optimal advertisement delivery point acquisition unit compares the customer portrait model with the crowd portrait model by calculating a comprehensive similarity index to obtain the optimal advertisement delivery point;

[0124] Positioning advertising delivery management module: including a Pearson correlation coefficient calculation unit, a regression analysis modeling unit and a positioning advertising delivery management plan generation unit; wherein, the Pearson correlation coefficient calculation unit is responsible for calculating the correlation between competing advertising information and weather information and advertising effects, the regression analysis modeling unit is responsible for establishing a mathematical model between competing advertising information and weather information and advertising effects, and determining the delivery time period with the least impact, and the positioning advertising delivery management plan generation unit is responsible for combining the time period with the least impact of competing advertising and weather on advertising delivery with the optimal advertising delivery point to generate a positioning advertising delivery management plan.

[0125] In this embodiment, we draw the following conclusions through Pearson correlation coefficient calculation and regression analysis: the correlation between competitive advertising information and advertising effect is low, with a correlation coefficient of 0.2; the correlation between weather information and advertising effect is high, with a correlation coefficient of 0.8. According to regression analysis, we found that weather information has a more significant impact on advertising effect. Therefore, the positioning advertising delivery management plan we obtained is as follows: choose to deliver advertising during time periods with good weather, such as sunny days or warm seasons; combine competitive advertising information to avoid time periods with more competitive advertising to reduce competitive pressure; deliver advertising during time periods with high user activity, such as evenings or weekends; and give the location of the best advertising delivery point.

[0126] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0127] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for managing location-based advertising based on big data, characterized in that: The method comprises the following steps: S100, obtaining the geographical range where the advertiser plans to place advertisements, obtaining the advertisement placement point information of all advertisement placement points within the geographical range, wherein the advertisement placement point information includes the name and longitude and latitude coordinates, and constitutes an advertisement placement point set; obtaining corresponding traffic information, crowd portrait information, crowd behavior data, competitive advertisement information, and weather information according to the information of each advertisement placement point in the advertisement placement point set, and adding the obtained information to the advertisement placement point set; S200, obtaining information about related products of advertisements provided by advertisers, wherein different information is obtained depending on whether the related products are new products, and principal component analysis is used to reduce the dimension of the related product information; S201. If the associated product provided by the advertiser for advertising is a new product, the associated product information obtained includes product features, market competition, target audience characteristics and expected sales data; S202: If the associated product provided by the advertiser for advertising is an old product, the associated product information obtained includes historical sales data, customer survey data, social media data, and market competition conditions; S300, based on the results of principal component analysis, using K-means clustering to group customers into several customer groups with similar characteristics according to related product information, and obtaining the proportion of corresponding customer groups to establish a customer portrait model; S400, feature engineering is performed on the set of advertising placement points, and crowd portrait models of different advertising placement points are established using K-means clustering; the customer portrait model and the crowd portrait model are compared by calculating a comprehensive similarity index to obtain the best advertising placement point; Step S400 includes: S401, using principal component analysis, performing feature engineering based on the traffic information, crowd portrait information and crowd behavior data in the set of advertising placement points; S402, using K-means clustering to establish crowd portrait models for different advertising delivery points; S403, using the established customer portrait model and the crowd portrait model of each advertisement delivery point to respectively describe crowd characteristics and customer characteristics of different advertisement delivery points; S404. For each advertisement delivery point, the similarity between its crowd portrait model and the customer portrait model is calculated using cosine similarity, Euclidean distance and Jaccard similarity coefficient, and the average of the three similarity calculation results is taken as a comprehensive similarity index; S405: sort each advertisement placement point by the proportion of its customer group, where a higher proportion means that the advertisement placement point covers more customer groups; S406, evaluating each advertising placement point based on the comprehensive similarity index and customer group ranking of each advertising placement point, using a weighted average algorithm to weight the similarity index and the proportion to obtain a score for each advertising placement point; S407, selecting the advertisement delivery point with the highest score as the best advertisement delivery point; S500. Based on the competitive advertising information and weather information of the best advertising delivery point, use the Pearson correlation coefficient to evaluate the correlation between the competitive advertising information and weather information and the advertising effect respectively; use the regression analysis model to determine the time period when the competitive advertising and weather have the least impact on the advertising delivery, and combine it with the best advertising delivery point to generate a targeted advertising delivery management plan.

2. The method for managing location-based advertising based on big data according to claim 1, characterized in that: According to step S100, the presence of human bodies is sensed by ultrasonic waves through the crowd flow detector installed at the advertising point to obtain crowd flow information; the face position and area in the image are detected by using the YOLO algorithm through the camera installed at the advertising point, and the identity, age, gender, clothing and appearance of the face are recognized in combination with the face recognition model DeepFace, and the crowd portrait information is combined; the behavior and action trajectory of the crowd are obtained through the camera installed at the advertising point through the behavior recognition algorithm based on the recurrent neural network and the trajectory tracking algorithm based on the Kalman filter, and the crowd behavior data is combined; Competitive advertising information is obtained through on-site surveys by personnel and input into the system to form competitive advertising information; weather information is obtained by accessing the meteorological data API provided by the meteorological department; the obtained information is added to the advertising delivery point set, and the advertising delivery point set is classified according to different advertising delivery points.

3. The method for managing location-based advertising based on big data according to claim 1, characterized in that: Step S200 includes: S203. According to the associated product information, a data set X containing n samples and m features is obtained, where represents the jth eigenvalue of the i-th sample; the data in the dataset X needs to be standardized so that the mean of each feature is 0 and the variance is 1. The standardization formula is as follows: ； Among them, the mean of the jth feature is , the standard deviation is , is the standardized value of the jth feature of the i-th sample; S204, the standardized data set Calculate the covariance matrix , whose elements are: ； in, is the normalized value of the i-th feature of the k-th sample, is the normalized value of the jth feature of the kth sample; S205, covariance matrix Perform eigenvalue decomposition to obtain eigenvalues ​​and corresponding eigenvectors; S206, performing principal component analysis, selecting eigenvectors corresponding to several most important eigenvalues, and forming a feature space after dimensionality reduction; S207: Project the original data X to the new principal component space to obtain the data matrix Y after dimensionality reduction.

4. The method for managing location-based advertising based on big data according to claim 3, characterized in that: According to step S206, when performing principal component analysis, the eigenvalue represents the variance explained by each principal component; Calculate the proportion of each eigenvalue to the total variance, and then select how many principal components to retain based on the set threshold. The threshold is the percentage of the total variance retained and is set by the user; Ultimately, the principal components that explain most of the variance are selected to be retained.

5. The method for managing location-based advertising based on big data according to claim 1, characterized in that: Step S300 includes: S301, randomly initialize r cluster centers, randomly select r samples from the data matrix Y as initial cluster centers, and the initial cluster centers will be used as initial customer group feature vectors; S302, for each customer group feature vector, calculate its distance from each cluster center, and assign it to the customer group represented by the cluster center closest to it, so as to group customers into customer groups with similar features; S303, for each customer group, recalculate the mean of its sample feature vector, and use the mean as the new cluster center to update the feature vector of the customer group; S304, repeating steps S302-S303 until the cluster center of the customer group no longer changes and the clustering converges, thereby obtaining a clustering result of the customer group based on the associated product information; S305. Count the proportion of each customer group based on the clustering results.

6. A method for managing location-based advertising based on big data according to claim 5, characterized in that: According to steps S301-S304, the goal of K-means clustering is to minimize the sum of the squares of the distances between each sample and its cluster center, which is expressed by the following formula: ； Among them, J is the objective function of K-means clustering; is the indicator function, indicating that the sample Is it a cluster center? ; K-means clustering updates the cluster centers by iteration and , until convergence; According to step S305, the proportion of each customer group is the number of customers in each customer group divided by the total number of customers.

7. The method for managing location-based advertising based on big data according to claim 1, characterized in that: According to step S500, the Pearson correlation coefficient is used to evaluate the correlation between the competitive advertising information and the weather information and the advertising effect, wherein the advertising effect is specifically quantified by the average value of each conversion of the advertisement; for the competitive advertising information, the Pearson correlation coefficient The calculation formula is as follows: ； in, represents the sample value of the competing advertisement information, A sample value representing the advertising effect. represents the mean of the data of competing advertising information, The mean of the data representing the advertising effect, Represents the product of the standard deviations between competing advertising information A and advertising effectiveness B.

8. The method for managing location-based advertising based on big data according to claim 1, characterized in that: According to step S500, a multivariate linear regression model is used to describe the relationship between competitive advertising information, weather information and advertising effect. The model is expressed as: Among them, B is the advertising effect, A is the competitive advertising information, and A' is the weather information. is the error term, , and are the coefficients of the multiple linear regression model, is the correlation between competitive advertising information and advertising effectiveness, is the correlation between weather information and advertising effectiveness, and Need to be obtained in the training of the multivariate linear regression model; Divide the set of advertising points into training set and test set, select features for training from the training set, use the training set data to train the linear regression model, and find the best coefficients through the square loss function , minimize the error between the predicted value and the actual value; use the test set data to evaluate the trained model, use the mean square error to calculate the evaluation result, and tune the multivariate linear regression model according to the evaluation result; The established regression model is used to analyze the impact of competitive advertising information and weather information on advertising effectiveness, and to determine the time period with the least impact; the determined time period with the least impact is combined with the optimal advertising delivery point to generate a targeted advertising delivery management plan.

9. A positioning advertising delivery management system based on big data, using a positioning advertising delivery management method based on big data according to any one of claims 1 to 8, characterized in that: include: Geographic information acquisition module: including: geographic range acquisition unit and advertisement placement point information acquisition unit; wherein the geographic range acquisition unit is responsible for acquiring geographic range information of the expected advertisement placement from the advertiser, and the advertisement placement point information acquisition unit is responsible for acquiring the names and latitude and longitude coordinate information of all advertisement placement points within the geographic range; Data acquisition and integration module: including: a traffic information acquisition unit, a crowd portrait information acquisition unit, a crowd behavior data acquisition unit, a competitive advertising information acquisition unit, a weather information acquisition unit and an information integration unit; wherein the traffic information acquisition unit is responsible for acquiring the traffic information of each advertising delivery point from relevant data sources, the crowd portrait information acquisition unit is responsible for acquiring the crowd portrait information of each advertising delivery point, the crowd behavior data acquisition unit is responsible for acquiring the crowd behavior data of each advertising delivery point, the competitive advertising information acquisition unit is responsible for acquiring relevant information of competitive advertising, the weather information acquisition unit is responsible for acquiring the weather information of the area where each advertising delivery point is located, and the information integration unit is responsible for integrating the various types of information acquired into the set of advertising delivery points; Related product information processing module: including: related product information acquisition unit, new product information processing unit, old product information processing unit and principal component analysis unit; wherein the related product information acquisition unit is responsible for acquiring related product information provided by advertisers for advertisement placement, the new product information processing unit performs corresponding processing and analysis based on the new product information, the old product information processing unit performs corresponding processing and analysis based on the old product information, and the principal component analysis unit is responsible for performing principal component analysis on the related product information, reducing the dimension and performing feature extraction and analysis; Customer clustering module: including: a first K-means clustering unit and a customer portrait model building unit; wherein the first K-means clustering unit uses K-means clustering to group customers into groups with similar characteristics and determine their proportions, and the customer portrait model building unit is responsible for building a customer portrait model according to the processing results of the first K-means clustering unit; Advertisement delivery point module: including: an advertisement delivery point feature extraction unit, a second K-means clustering unit and an optimal advertisement delivery point acquisition unit; wherein the advertisement delivery point feature extraction unit extracts and analyzes features of data in the advertisement delivery point set, the second K-means clustering unit uses K-means clustering to group advertisement delivery points and establish crowd portrait models for different advertisement delivery points; the optimal advertisement delivery point acquisition unit compares the customer portrait model with the crowd portrait model by calculating a comprehensive similarity index to obtain the optimal advertisement delivery point; Positioning advertising delivery management module: including a Pearson correlation coefficient calculation unit, a regression analysis modeling unit and a positioning advertising delivery management plan generation unit; wherein, the Pearson correlation coefficient calculation unit is responsible for calculating the correlation between competing advertising information and weather information and advertising effects, the regression analysis modeling unit is responsible for establishing a mathematical model between competing advertising information and weather information and advertising effects, and determining the delivery time period with the least impact, and the positioning advertising delivery management plan generation unit is responsible for combining the time period with the least impact of competing advertising and weather on advertising delivery with the optimal advertising delivery point to generate a positioning advertising delivery management plan.

Images