Intelligent advertisement putting system and method based on user behavior analysis

Abstract

The invention relates to the technical field of artificial intelligence, in particular to an intelligent advertisement putting system and method based on user behavior analysis, and the system comprises a data acquisition module, a behavior analysis module, an advertisement putting module and a monitoring feedback module. The behavior analysis module is used for constructing a user behavior analysis model and a cold start model, the advertisement putting module is used for generating and optimizing an advertisement putting strategy, and the monitoring feedback module is used for optimizing the strategy and the model according to the advertisement investment return rate. The advertisement putting accuracy, efficiency and income are improved, the method adapts to a cold start scene, and the market adaptability is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent advertising delivery system and method based on user behavior analysis. Background Art

[0002] In the field of advertising, the cold start problem has always been a major problem that has plagued advertising practitioners. For new users, the platform has no records of their past browsing, purchasing and other behaviors, and traditional advertising strategies based on user behavior analysis cannot work. In new advertising scenarios, since the advertisement has no historical data, it is impossible to know its appeal and conversion rate among different user groups, resulting in a waste of advertising resources. New platforms also face the cold start dilemma. New advertising platforms may lack a mature user behavior data accumulation and analysis system, and cannot effectively use user behavior information to optimize advertising. The cold start problem in advertising seriously restricts the effectiveness and efficiency of advertising, and an effective solution is urgently needed.

[0003] Chinese Patent Publication No. CN113465672A discloses an intelligent advertising delivery system, including an advertising space detection device. However, this solution still lacks a user behavior data accumulation and analysis system, resulting in insufficient accuracy and low efficiency in advertising delivery. Summary of the Invention

[0004] To this end, the present invention provides an intelligent advertising delivery system and method based on user behavior analysis to overcome the problems in the prior art of lacking a user behavior data accumulation and analysis system, resulting in insufficient accuracy and low efficiency in advertising delivery.

[0005] To achieve the above objectives, the present invention provides, on the one hand, an intelligent advertising delivery system based on user behavior analysis, comprising:

[0006] A data acquisition module is used to collect target advertising data;

[0007] a behavior analysis module for acquiring a target comprehensive feature vector based on target advertising data, for constructing a user behavior analysis model using a user behavior analysis model construction method, for constructing a cold start model using a cold start model construction method, and for optimizing the user behavior analysis model based on the cold start model;

[0008] An advertising delivery module, which is used to generate an advertising delivery strategy based on a user behavior analysis model using an advertising delivery strategy generation method, and is also used to adjust the advertising delivery strategy in real time;

[0009] The monitoring feedback module is used to calculate the advertising return on investment, obtain the advertising return on investment, and optimize the advertising delivery strategy based on the advertising return on investment.

[0010] Furthermore, the data acquisition module collects target advertising data, which includes e-commerce data and social media data. The e-commerce data is collected by embedding code in e-commerce websites and mobile applications, and the social media data is collected through an application programming interface. The e-commerce data includes the number of views, the number of clicks, the number of purchases, the price of purchased goods, the purchase time interval and the time the user browses the platform. The social media data includes user comments, user forwarding, user likes, user sharing, the number of fans in the user's social circle and historical advertising records of the e-commerce platform.

[0011] Furthermore, the behavior analysis module obtains the target comprehensive feature vector based on the target advertisement data, and the process includes:

[0012] Step J01: Calculate browsing frequency d1 according to the formula d1 = a1 / t, calculate click frequency d2 according to the formula d2 = a2 / t, and calculate purchase frequency d3 according to the formula d3 = a3 / t, where a1 is the number of browsing times, a2 is the number of clicks, a3 is the number of purchases, and t is the preset duration. The browsing frequency d1, click frequency d2, and purchase frequency d3 are used as behavioral frequency features.

[0013] Step J02: Obtain product association features based on product association rules;

[0014] Step J03: Compare the price h of the purchased product with the average price h0 of similar products. Based on the comparison results, determine the price trend of the purchased product h, and output the price sensitivity based on the determination results, where:

[0015] When h<h0, the price h of the purchased product is judged to be high, and the price sensitivity output is low, and the low sensitivity is used as the price sensitivity feature;

[0016] When h≥h0, the price h of the purchased product is judged to be low, the price sensitivity output is high, and the high sensitivity is used as the price sensitivity feature;

[0017] Step J04: Use behavior frequency features, product association features, and price sensitivity features as e-commerce data features;

[0018] Step J05: Perform sentiment analysis on user comments through natural language processing to obtain sentiment tendency features, perform interest analysis on user reposts, likes, and shares through natural language processing to obtain interest topic features, perform social analysis on user shares and the number of followers in the user's social circle through natural language processing to obtain social influence features, and use the sentiment tendency features, interest topic features, and social influence features as social media data features;

[0019] Step J06: normalize the e-commerce data features and social media data features to obtain a target comprehensive feature vector.

[0020] Furthermore, the behavior analysis module constructs a user behavior analysis model through a user behavior analysis model construction method, and the user behavior analysis model construction method includes:

[0021] Step U01: divide the historical comprehensive database into a 70% user behavior training set, a 20% user behavior verification set, and a 10% user behavior test set;

[0022] Step U02: Set the number of weak learners in the gradient boosting decision tree model parameters to 100, set the learning rate to 0.1, input the user behavior training set into the gradient boosting decision tree model after parameter setting, and train it to obtain the trained gradient boosting decision tree model;

[0023] In step U03, the user behavior verification set is input into the trained gradient boosting decision tree model, and the parameters of the trained gradient boosting decision tree model are optimized to obtain the parameter-optimized gradient boosting decision tree model. The user behavior test set is then input into the parameter-optimized gradient boosting decision tree model for testing, and the test accuracy R is output.

[0024] Furthermore, the user behavior analysis model construction method also includes: step U04, comparing the test accuracy R with the preset test accuracy R0, judging the compliance of the gradient boosting decision tree model after parameter optimization based on the comparison result, and outputting the gradient boosting decision tree model after parameter optimization based on the judgment result.

[0025] Furthermore, the behavior analysis module constructs a cold start model using a cold start model construction method, and the cold start model construction method includes:

[0026] Step Q01: compare the preset advertisement with the historical advertisement data search library, determine the existence of the preset advertisement in the historical advertisement data search library based on the comparison result, and output the definition of the preset advertisement based on the determination result, wherein:

[0027] When the existence of the preset delivery advertisement in the historical advertisement data search library is not found, the definition of the preset delivery advertisement is output as a new advertisement;

[0028] When the existence of the preset delivery advertisement in the historical advertisement data search library is "exist", the definition of the preset delivery advertisement is output as an old advertisement;

[0029] Step Q02: Compare the preset advertising e-commerce platform with the historical advertising e-commerce platform database, determine the existence of the preset advertising e-commerce platform in the historical advertising e-commerce platform database based on the comparison result, and output the definition of the preset advertising e-commerce platform based on the determination result, wherein:

[0030] When the preset advertising delivery e-commerce platform does not exist in the historical advertising e-commerce platform database, the definition of the preset advertising delivery e-commerce platform is output as a new platform;

[0031] When the preset advertising delivery e-commerce platform exists in the historical advertising e-commerce platform database, the definition of the preset advertising delivery e-commerce platform is output as an old platform;

[0032] Step Q03, comparing the user's platform browsing time t1 within the preset time with the preset browsing time t0, judging the attributes of the user's platform browsing time t1 within the preset time based on the comparison result, and outputting the user type based on the judgment result;

[0033] Step Q04: divide 70% of the target comprehensive feature vector into a cold start training set, and divide 30% of the target comprehensive feature vector into a cold start validation set;

[0034] Step Q05: Select the prototype network algorithm as the cold start model, initialize the weights and biases of the cold start model, input the cold start training set into the cold start model, and calculate the output of the cold start model;

[0035] Step Q06: Calculate the loss function value based on the output and label of the cold start model, calculate the gradient through the back propagation algorithm, and update the weights and bias of the cold start model. Repeat the forward propagation, loss function calculation and back propagation process to obtain the trained cold start model.

[0036] In step Q07, the cold start validation set is input into the trained cold start model for testing, and the trained cold start model with an accuracy rate of 90% is output as the cold start model.

[0037] Furthermore, the behavior analysis module obtains the time t1 of the user browsing the platform within the preset time t2, calculates the user like frequency according to the formula Fd=f / t2, where f is the number of likes of the user, calculates the user comment frequency FP according to the formula FP=f2 / t2, where f2 is the number of comments of the user, and calculates the matching degree P between the user group and the platform according to the formula P=ω1×t1+ω2×Fd+ω3×FP, where ω1 is the weight parameter of the user browsing the platform time t1 within the preset time t2, ω2 is the weight parameter of the user like frequency, and ω3 is the weight parameter of the user comment frequency. The matching degree P of the user group and the platform is compared with the preset matching degree P0, and the matching degree of the user group and the platform is judged according to the comparison result, and the cold start model is adjusted according to the judgment result.

[0038] Furthermore, the advertisement delivery module generates an advertisement delivery strategy through an advertisement delivery strategy generation method, and the advertisement delivery strategy generation method includes:

[0039] Step K01: The probability G of a user purchasing a product obtained through a user behavior analysis model is compared with various preset purchase probabilities, including a first preset purchase probability G1 and a second preset purchase probability G2. The likelihood of the user purchasing the product is determined based on the comparison results, and the user's purchase attributes are output based on the determination results, where:

[0040] When G≤G1, the advertisement delivery module determines that the probability G of the user purchasing the product is low, and outputs the user's purchase attribute as a user with low purchase probability;

[0041] When G1<G≤G2, the advertisement delivery module determines that the probability G of the user purchasing the product is medium, and outputs the user's purchase attribute as a medium purchase possibility user;

[0042] When G>G2, the advertisement delivery module determines that the probability G of the user purchasing the product is high, and outputs the user's purchase attribute as a high purchase possibility user;

[0043] Step K02: When the user's purchase attribute is output as a low purchase possibility user, the advertising delivery strategy is output as follows: the number of delivery times within the preset period is N1, the advertising content is popular products and newly launched products, and product information and purchase guidance are provided;

[0044] When the user's purchase attribute is output as a medium purchase possibility user, the output advertising delivery strategy is: the number of delivery times within the preset period is N2, and the advertising content is the value and practicality of the product, providing product information and purchase guidance;

[0045] When the user's purchase attribute output is a high purchase possibility user, the output advertising delivery strategy is: the number of delivery times within the preset period is N3, and the advertising content is the advantages and characteristics of the product, providing product information and purchase guidance;

[0046] The advertising delivery module compares the purchase time interval TG with the preset purchase time interval TG0, determines the key attributes of the purchase time interval TG according to the comparison result, and adjusts the advertising delivery strategy according to the determination result;

[0047] When the advertising delivery module adjusts the advertising delivery strategy, for users with low purchase probability, the adjusted delivery times are set to N1', N1'=N1. After the preset delivery time, the probability G of the user purchasing the product is updated, and the purchase update coefficient is set to β, 1.1≤β≤1.5. The updated probability of the user purchasing the product is G', G'=β×G.

[0048] For users with medium purchase probability, the adjusted number of delivery times is set to N2`, N2` = 1.2 × N2;

[0049] For users with high purchase probability, the adjusted number of delivery times is set to N3`, N3` = 2.5 × N3;

[0050] The advertising module obtains the product's sales growth rate M1, inventory M2, and sales growth rate M3 of competing products based on a market research agency report, calculates the product promotion index C according to the formula C=0.4×M1+0.2×M2+0.4×M3, compares the product promotion index C with a preset promotion index C0, determines the degree of influence of the comparison result on the product promotion index C, and updates the purchase time interval TG based on the determination result, wherein:

[0051] When C≤C0, the advertising delivery module determines that the influence of the product promotion index C is low, and does not update the purchase time interval TG;

[0052] When C>C0, the advertising delivery module determines that the influence of the product promotion index C is high, and updates the purchase time interval TG. For users with low purchase possibility, N1`=N2`=N3` is set; for users with medium purchase possibility, N1`=N2`=N3` is set; for users with high purchase possibility, the advertising delivery strategy is not adjusted.

[0053] Furthermore, the monitoring feedback module obtains the product investment cost I0 and product investment income I according to the market research agency report, and calculates the product investment cost I0 and product investment income I according to the formula Calculate the advertising return on investment (ROI), compare the advertising return on investment (ROI) with the preset return rate (ROI0), determine the return attribute of the advertising return on investment (ROI) based on the comparison results, and optimize the advertising delivery strategy based on the determination results, including:

[0054] When ROI≥ROI0, the monitoring feedback module determines that the return attribute of the advertising return on investment ROI is high return, and does not optimize the advertising delivery strategy;

[0055] When ROI<ROI0, the monitoring feedback module determines that the return attribute of the advertising return on investment ROI is low, and optimizes the advertising delivery strategy:

[0056] For users with low purchase probability, the first optimization coefficient is set to θ1, θ1 = 1.7-3.2e -1.1×[(ROIO-ROI)-2 The optimized number of launches is N1``, N1``=θ1×N1`;

[0057] For users with medium purchase probability, the second optimization coefficient is set to θ2, θ2 = 1.2-3.3e -1.2×[(ROIO-ROI)+1] , the optimized delivery times are N2``, N2``=θ2×N2`

[0058] For users with high purchase probability, the third optimization coefficient is set to θ3, θ3 = 1.5-3.2e -1.2×[(ROIO-ROI)+2] , the optimized number of delivery times is N3``, N3``=θ3×N3`.

[0059] On the other hand, the present invention also provides a method for intelligent advertising delivery based on user behavior analysis, comprising:

[0060] Step S01, for collecting target advertising data;

[0061] Step S02 is used to obtain a target comprehensive feature vector based on the target advertising data, to construct a user behavior analysis model using a user behavior analysis model construction method, to construct a cold start model using a cold start model construction method, and to optimize the user behavior analysis model based on the cold start model;

[0062] Step S03 is used to generate an advertisement delivery strategy according to the user behavior analysis model using an advertisement delivery strategy generation method, and also to adjust the advertisement delivery strategy in real time;

[0063] Step S04 is to calculate the advertising investment return rate, obtain the advertising investment return rate, and optimize the advertising delivery strategy based on the advertising investment return rate.

[0064] Compared with the existing technology, the beneficial effect of the present invention is that this embodiment builds a refined management and analysis system covering intelligent advertising delivery through the efficient collaborative operation of multiple modules, so as to facilitate multi-dimensional analysis of advertising delivery duration and effectively improve the accuracy and efficiency of advertising delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a schematic diagram of the structure of the intelligent advertising delivery system based on user behavior analysis in this embodiment;

[0066] Figure 2 Schematic diagram of the process of intelligent advertising delivery method based on user behavior analysis in this embodiment. DETAILED DESCRIPTION

[0067] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0068] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0069] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0070] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0071] See also Figure 1 As shown in FIG, it is a structural diagram of the intelligent advertising delivery system based on user behavior analysis in this embodiment, and the system includes:

[0072] A data acquisition module is used to collect target advertising data;

[0073] a behavior analysis module for acquiring a target comprehensive feature vector based on target advertising data, for constructing a user behavior analysis model using a user behavior analysis model construction method, for constructing a cold start model using a cold start model construction method, and for optimizing the user behavior analysis model based on the cold start model; the behavior analysis module is connected to the data acquisition module;

[0074] An advertising delivery module, which is used to generate an advertising delivery strategy based on a user behavior analysis model using an advertising delivery strategy generation method, and is also used to adjust the advertising delivery strategy in real time. The advertising delivery module is connected to the behavior analysis module;

[0075] The monitoring and feedback module is used to calculate the advertising investment return rate, obtain the advertising investment return rate, and optimize the advertising delivery strategy based on the advertising investment return rate. The monitoring and feedback module is connected to the advertising delivery module.

[0076] Specifically, the system is applied to the advertising delivery system terminal, and the entire process of intelligent advertising delivery is finely managed through the collaborative operation of multiple modules, thereby improving the accuracy, efficiency and benefits of advertising delivery. The system collects e-commerce data and social media data through the data acquisition module code embedding and application programming interface to improve the comprehensiveness and real-time nature of data collection. The system constructs a user behavior analysis model and a cold start model through the behavior analysis module, and analyzes the matching degree between user behavior and platform in combination with the target comprehensive feature vector to improve the accuracy of user behavior analysis and the adaptability of the cold start model. The system generates an advertising delivery strategy through the probability classification of user purchases of products through the advertising delivery module, and dynamically adjusts the strategy based on the purchase time interval and product promotion index to improve the pertinence and flexibility of advertising delivery. The system calculates the advertising return on investment through the monitoring feedback module to optimize the advertising delivery strategy and improve the economic benefits and model performance of advertising delivery.

[0077] Specifically, the data acquisition module collects target advertising data, which includes e-commerce data and social media data. The e-commerce data includes the number of views, clicks, purchases, prices of purchased goods, purchase time intervals, and user browsing platform time. This embodiment does not limit the method of collecting e-commerce data. Those skilled in the art can freely set it according to actual conditions, and only need to meet the collection requirements of users' purchase characteristics on e-commerce platforms. For example, in this embodiment, the data acquisition module is set to collect e-commerce data by embedding code on e-commerce websites and mobile applications. The social media data includes user comments, user forwarding, user likes, user sharing, and the number of users' social circle fans. This embodiment does not limit the method of collecting social media data. Those skilled in the art can freely set it according to actual conditions, and only need to meet the collection requirements of users' behavioral characteristics on social media platforms. For example, in this embodiment, the data acquisition module is set to collect social media data through an application programming interface.

[0078] Specifically, the e-commerce data refers to the data generated when users conduct shopping-related activities on e-commerce platforms, reflecting the users' consumption behavior. The social media data refers to the interactive behavior data of users on social media platforms, reflecting the users' social behavior. The e-commerce website refers to the website used for online commodity transactions. The mobile application refers to the e-commerce application installed on the mobile device. The code embedding refers to the insertion of specific statistical code in the code of the e-commerce website and mobile application. The specific statistical code refers to the code used to track user behavior data, such as JavaScript tags. The application programming interface refers to the standardized interface provided by the social media platform, and the social media data can be obtained after authorization by the social media platform. The number of views refers to the total number of times users visit the e-commerce platform. The number of clicks refers to the total number of times users click on the interactive entrance on the e-commerce platform. This embodiment does not limit the interactive entrance, such as product pictures and purchase buttons. The number of purchases refers to the number of users The total number of times an order for goods is placed and paid for on an e-commerce platform; the purchase time interval refers to the time difference between two adjacent purchases by a user; the price of a purchased good refers to the actual amount paid by a user for a single good; the user's platform browsing time refers to the total length of time a user visits the e-commerce platform; the user's comments refer to the text content posted by a user on social media regarding advertisements and goods, including the number of user comments; the user's forwarding refers to the user sharing advertisement content and product links to his or her own social circle; the user's likes refer to the number of clicks by the user expressing approval of the advertisement and goods, including the number of user likes; the user's sharing refers to the user's behavior of publicly disseminating goods on the e-commerce platform through social media; the number of fans in the user's social circle refers to the total number of fans the user has on social media. This embodiment does not limit the type of social media, and those skilled in the art can freely set it according to actual conditions, as long as it meets the user's needs for online social networking, such as Weibo.

[0079] Specifically, the data acquisition module collects e-commerce data and social media data through data acquisition module code embedding and application programming interface, thereby improving the comprehensiveness and real-time performance of data collection.

[0080] Specifically, the behavior analysis module obtains the target comprehensive feature vector based on the target advertisement data, and the process includes:

[0081] Step J01: Calculate browsing frequency d1 according to the formula d1 = a1 / t, calculate click frequency d2 according to the formula d2 = a2 / t, and calculate purchase frequency d3 according to the formula d3 = a3 / t, where a1 is the number of browsing times, a2 is the number of clicks, a3 is the number of purchases, and t is the preset duration. The browsing frequency d1, click frequency d2, and purchase frequency d3 are used as behavioral frequency features.

[0082] Step J02: Obtain product association features based on product association rules;

[0083] Step J03: Compare the price h of the purchased product with the average price h0 of similar products. Based on the comparison results, determine the price trend of the purchased product h, and output the price sensitivity based on the determination results, where:

[0084] When h<h0, the price h of the purchased product is judged to be high, and the price sensitivity output is low, and the low sensitivity is used as the price sensitivity feature;

[0085] When h≥h0, the price h of the purchased product is judged to be low, the price sensitivity output is high, and the high sensitivity is used as the price sensitivity feature;

[0086] Step J04: Use behavior frequency features, product association features, and price sensitivity features as e-commerce data features;

[0087] Step J05: Perform sentiment analysis on user comments through natural language processing to obtain sentiment tendency features, perform interest analysis on user reposts, likes, and shares through natural language processing to obtain interest topic features, perform social analysis on user shares and the number of followers in the user's social circle through natural language processing to obtain social influence features, and use the sentiment tendency features, interest topic features, and social influence features as social media data features;

[0088] Step J06: normalize the e-commerce data features and social media data features to obtain a target comprehensive feature vector.

[0089] Specifically, the product association rule refers to a data analysis method that discovers the association characteristics of different products during the sales process by mining sales data and user behavior data. In this embodiment, the product association rule is set to group similar products and customers to discover the association between products in the group, such as clustering products with similar functions, price ranges, and target customer groups into one category. The sales data refers to the core data of users completing transactions on the e-commerce platform, such as product category, unit price, and product sales volume. The user behavior data refers to the behavioral records generated by users during the interaction with the system and platform, such as the page content browsed and the browsing time. The product association characteristics refer to the characteristics that reflect the association relationship between products obtained based on the analysis of product association rules. This embodiment does not specifically limit the form of product association characteristics. Those skilled in the art can freely choose according to actual needs, as long as they meet the requirements for obtaining product association characteristics, such as product group association. The average price of similar products refers to the price value obtained by statistically averaging the prices of a category of products with similar attributes, functions, and uses. The price sensitivity refers to the measurement of consumers' response to changes in product prices. Natural language processing refers to the technical means of allowing computers to classify, analyze, and extract information from data, allowing computers to understand, process, and generate human language. The embodiment does not limit the specific method of natural language processing. People in this field can freely choose according to actual needs, and only need to meet the needs of obtaining emotional tendency characteristics and interest topic characteristics. For example, a keyword extraction method can be set to obtain emotional tendency characteristics, and a semantic understanding method can be set to obtain interest topic characteristics. The emotional analysis refers to an application of natural language processing. By analyzing the vocabulary and sentence structure expressed in the user's comments, the emotional tendency contained in the text is obtained. This embodiment does not limit the emotional tendency contained in the text, such as positive, negative and neutral. The interest analysis refers to the use of natural language processing to analyze the user's forwarding and user's likes The content shared by the user is analyzed to obtain the topic areas of interest to the user. This embodiment does not limit the topic areas of interest to the user, such as fashion, technology, and food. The social analysis refers to the use of natural language processing to obtain the influence of the user in the social network. This embodiment does not limit the influence of the user in the social network, such as the user's sharing and the number of fans in the user's social circle. The data normalization processing refers to the data preprocessing technology that converts the data to a unified scale to eliminate the differences in numerical size and dimension of different features, making the data more suitable for data analysis. This embodiment does not limit the data normalization processing, such as linear normalization.

[0090] Specifically, the behavior analysis module improves the accuracy of advertising delivery by extracting e-commerce data features and social media data features and normalizing them.

[0091] Specifically, the behavior analysis module constructs a user behavior analysis model through a user behavior analysis model construction method, and the user behavior analysis model construction method includes:

[0092] Step U01: divide the historical comprehensive database into a 70% user behavior training set, a 20% user behavior verification set, and a 10% user behavior test set;

[0093] Step U02: Set the number of weak learners in the gradient boosting decision tree model parameters to 100, set the learning rate to 0.1, input the user behavior training set into the gradient boosting decision tree model after parameter setting, and train it to obtain the trained gradient boosting decision tree model;

[0094] Step U03: Input the user behavior validation set into the trained gradient boosting decision tree model, optimize the parameters of the trained gradient boosting decision tree model, and obtain a gradient boosting decision tree model with optimized parameters. Then, input the user behavior test set into the gradient boosting decision tree model with optimized parameters for testing, and output the test accuracy R.

[0095] In step U04, the test accuracy R is compared with the preset test accuracy R0, and the compliance of the gradient boosting decision tree model after parameter optimization is judged based on the comparison result, and the gradient boosting decision tree model after parameter optimization is output based on the judgment result, where:

[0096] When R≥R0, the behavior analysis module determines that the gradient boosting decision tree model after parameter optimization meets the standards, and outputs the gradient boosting decision tree model after parameter optimization as the user behavior analysis model;

[0097] When R<R0, the behavior analysis module determines that the gradient boosting decision tree model after parameter optimization does not meet the standards, obtains the second e-commerce data features and the social media data features, and trains the gradient boosting decision tree model according to the second e-commerce data features and the social media data features until the gradient boosting decision tree model after parameter optimization meets the standards, and outputs the gradient boosting decision tree model after parameter optimization as the user behavior analysis model.

[0098] Specifically, the historical comprehensive database includes the e-commerce data features and social media data features in the historical records as the input data of the user behavior analysis model, and the probability of the user purchasing the product as the output data of the user behavior analysis model. The user behavior training set refers to the data set used to train the gradient boosting decision tree model. The user behavior verification set refers to the data set used to verify the performance of the model and evaluate the performance of the model on new data during the model training process. The user behavior test set refers to the data set for testing the accuracy of the gradient boosting decision tree model after the gradient boosting decision tree model training and parameter optimization are completed. The number of weak learners refers to the number of gradient boosting decision trees. The number of simple decision trees included in the tree model; the learning rate refers to the parameter that controls the degree of influence of the number of newly added weak learners on the improved decision tree model in each iteration during the training process of the gradient boosting decision tree model; the parameter optimization refers to the process of adjusting the parameters of the model, such as the weights of the weak learners and the splitting conditions of the decision tree, to minimize the loss function of the model on the validation set and the training set; the loss function refers to a numerical value used to measure the degree of difference between the model prediction results and the actual results; the preset test accuracy R0 refers to a pre-set value used to measure the compliance of the gradient boosting decision tree model after parameter optimization. In this embodiment, the test accuracy R0 is set to 90%.

[0099] Specifically, the behavior analysis module adopts a gradient boosting decision tree model to improve the accuracy and reliability of the user behavior analysis model through scientific data set division and parameter optimization.

[0100] Specifically, the behavior analysis module constructs a cold start model using a cold start model construction method, and the cold start model construction method includes:

[0101] Step Q01: compare the preset advertisement with the historical advertisement data search library, determine the existence of the preset advertisement in the historical advertisement data search library based on the comparison result, and output the definition of the preset advertisement based on the determination result, wherein:

[0102] When the existence of the preset delivery advertisement in the historical advertisement data search library is not found, the definition of the preset delivery advertisement is output as a new advertisement;

[0103] When the existence of the preset delivery advertisement in the historical advertisement data search library is "exist", the definition of the preset delivery advertisement is output as an old advertisement;

[0104] Step Q02: Compare the preset advertising e-commerce platform with the historical advertising e-commerce platform database, determine the existence of the preset advertising e-commerce platform in the historical advertising e-commerce platform database based on the comparison result, and output the definition of the preset advertising e-commerce platform based on the determination result, wherein:

[0105] When the preset advertising delivery e-commerce platform does not exist in the historical advertising e-commerce platform database, the definition of the preset advertising delivery e-commerce platform is output as a new platform;

[0106] When the preset advertising delivery e-commerce platform exists in the historical advertising e-commerce platform database, the definition of the preset advertising delivery e-commerce platform is output as an old platform;

[0107] Step Q03, compare the user's browsing time t1 within the preset time with the preset browsing time t0, judge the attribute of the user's browsing time t1 within the preset time based on the comparison result, and output the user type based on the judgment result, where:

[0108] When t1≥t0, the behavior analysis module determines that the attribute of the user browsing the platform time t1 within the preset time is a long time, and outputs the user type as an old user;

[0109] When t1<t0, the behavior analysis module determines that the attribute of the user browsing the platform time t1 within the preset time is a short time, and outputs the user type as a new user;

[0110] Step Q04: divide 70% of the target comprehensive feature vector into a cold start training set, and divide 30% of the target comprehensive feature vector into a cold start validation set;

[0111] Step Q05: Select the prototype network algorithm as the cold start model, initialize the weights and biases of the cold start model, input the cold start training set into the cold start model, and calculate the output of the cold start model;

[0112] Step Q06: Calculate the loss function value based on the output and label of the cold start model, calculate the gradient through the back propagation algorithm, and update the weights and bias of the cold start model. Repeat the forward propagation, loss function calculation and back propagation process to obtain the trained cold start model.

[0113] In step Q07, the cold start validation set is input into the trained cold start model for testing, and the trained cold start model with an accuracy rate of 90% is output as the cold start model.

[0114] Specifically, the preset advertisement refers to an advertisement that needs to be delivered, and the historical advertisement data search library refers to a database that stores advertisements that have been delivered by the historical advertisement delivery e-commerce platform. The historical advertisement data search library stores data in the form of data, such as the preset delivery advertisement name-historical advertisement content. The existence of the preset delivery advertisement in the historical advertisement data search library means that there is a preset delivery advertisement in the historical advertisement data search library. This embodiment does not limit the method for judging the existence of the preset delivery advertisement in the historical advertisement data search library. Relevant technicians in this field can freely choose according to actual needs, and only need to meet the need to judge the existence of the preset delivery advertisement in the historical advertisement data search library, such as through The historical advertising data search library is directly searched through the preset advertising to determine whether there are exactly the same records. The preset advertising delivery e-commerce platform refers to the e-commerce platform that is predetermined by the preset advertising delivery. The historical advertising e-commerce platform database refers to a data set that stores historical advertising e-commerce platform data. The historical advertising e-commerce platform database stores data in the form of data, such as the preset advertising delivery e-commerce platform name-historical platform name. The existence of the preset advertising delivery e-commerce platform in the historical advertising e-commerce platform database refers to the existence of the preset advertising delivery e-commerce platform in the historical advertising e-commerce platform database. This embodiment does not determine the existence of the preset advertising delivery e-commerce platform in the historical advertising e-commerce platform database. The preset time is limited in a specific way. Relevant technical personnel in this field can freely choose according to actual needs. It only needs to meet the need to judge the existence of the preset advertising e-commerce platform in the historical advertising e-commerce platform database, such as directly searching the historical advertising e-commerce platform database through the preset advertising e-commerce platform to judge whether there are exactly the same records. The preset time refers to a preset length of time. This embodiment does not limit the preset time. For example, the preset time is set to 24 hours. The preset browsing time t0 refers to a preset time length for comparing with the user browsing platform time within the preset time to judge the attribute of the user browsing platform time within the preset time. This embodiment does not limit the preset browsing time t0. The line is limited. In this embodiment, t0 is set to 2 hours. The prototype network algorithm refers to achieving rapid classification by extracting category prototypes. The category prototype refers to a representative feature vector. In this embodiment, the category prototype is a target comprehensive feature vector. The cold start training set refers to a data set used to train the cold start model. The cold start validation set refers to a data set used to verify the performance of the cold start model during the cold start model training process. The initialization refers to the process of setting initial values for the weights and biases of the model when constructing the cold start model. This embodiment does not limit the initialization method. The relevant technical personnel of this embodiment can freely choose according to actual needs. It is only necessary to meet the requirements of setting initial values for the weights and biases of the model, such as random initialization.The loss function value is a quantitative indicator that measures the error between the model's prediction results and the true labels. In this embodiment, the loss function is set to mean square error. The labels are the output values corresponding to the input values in the cold start model. The accuracy test is the process of evaluating the trained cold start model using the cold start validation set to determine the proportion of cold start model prediction results that are consistent with the true results.

[0115] Specifically, the behavior analysis module analyzes the matching degree between user behavior and platform in combination with the target comprehensive feature vector, thereby improving the accuracy of user behavior analysis and the adaptability of the cold start model.

[0116] Specifically, the behavior analysis module obtains the time t1 of the user browsing the platform within the preset time t2, calculates the user like frequency according to the formula Fd=f / t2, where f is the number of the user's likes, calculates the user comment frequency FP according to the formula FP=f2 / t2, where f2 is the number of the user's comments, and calculates the matching degree P of the user group and the platform according to the formula P=ω1×t1+ω2×Fd+ω3×FP, where ω1 is the weight parameter of the user browsing the platform time t1 within the preset time t2, ω2 is the weight parameter of the user like frequency, and ω3 is the weight parameter of the user comment frequency. The matching degree P of the user group and the platform is compared with the preset matching degree P0, and the matching degree of the user group and the platform is judged according to the comparison result, and the cold start model is adjusted according to the judgment result, where:

[0117] When P≥P0, the behavior analysis module determines that the matching degree between the user group and the platform is highly matched, and does not adjust the cold start model;

[0118] When P<P0, the behavior analysis module determines that the degree of match between the user group and the platform is low, adjusts the cold start model, and adds the e-commerce data feature vector and social media data feature vector of old users whose platform features are similar to those of new users to the target comprehensive feature vector.

[0119] Specifically, the number of likes of the user refers to the total number of like operations performed by the user on the platform within the preset time t2, the platform refers to a preset advertising e-commerce platform, the number of comments of the user refers to the total number of comment operations performed by the user on the platform within the preset time t2, the weight parameter refers to a preset parameter value used to measure the relative importance of influencing factors in evaluating the matching degree between the user group and the platform, the measured influence refers to the user browsing time t1 on the platform, the user like frequency and the user comment frequency within the preset time t2, the preset matching degree P0 refers to a preset standard value used to measure the matching degree between the user group and the platform. This embodiment does not limit the preset matching degree P0, and this embodiment sets P0= 0.7, the new user characteristics refer to the various behaviors, attributes, and interests exhibited by the new user on the platform. The old users with similar new user characteristics refer to old users who are similar to the new user in terms of behavior, attributes, and interests. The e-commerce data feature vectors and social media data feature vectors of the old users with similar new user characteristics refer to the behavioral data of the old users on the e-commerce platform and social media, which are extracted and converted into numerical feature vectors to supplement the input information of the cold start model. This embodiment extracts features from the e-commerce data and social media data of the old users and normalizes the extracted features to obtain the e-commerce data feature vectors and social media data feature vectors of the old users with similar new user characteristics.

[0120] Specifically, the behavior analysis module calculates the matching degree P between the user group and the platform through multi-dimensional data, dynamically adjusts the cold start model, and enhances the matching between new users and the platform with the help of similar old user characteristics, thereby improving the accuracy of the cold start model.

[0121] Specifically, the advertisement delivery module generates an advertisement delivery strategy through an advertisement delivery strategy generation method, and the advertisement delivery strategy generation method includes:

[0122] Step K01: The probability G of a user purchasing a product obtained through a user behavior analysis model is compared with various preset purchase probabilities, including a first preset purchase probability G1 and a second preset purchase probability G2. The likelihood of the user purchasing the product is determined based on the comparison results, and the user's purchase attributes are output based on the determination results, where:

[0123] When G≤G1, the advertisement delivery module determines that the probability G of the user purchasing the product is low, and outputs the user's purchase attribute as a user with low purchase probability;

[0124] When G1<G≤G2, the advertisement delivery module determines that the probability G of the user purchasing the product is medium, and outputs the user's purchase attribute as a medium purchase possibility user;

[0125] When G>G2, the advertisement delivery module determines that the probability G of the user purchasing the product is high, and outputs the user's purchase attribute as a high purchase possibility user;

[0126] Step K02: When the user's purchase attribute is output as a low purchase possibility user, the advertising delivery strategy is output as follows: the number of delivery times within the preset period is N1, the advertising content is popular products and newly launched products, and product information and purchase guidance are provided;

[0127] When the user's purchase attribute is output as a medium purchase possibility user, the output advertising delivery strategy is: the number of delivery times within the preset period is N2, and the advertising content is the value and practicality of the product, providing product information and purchase guidance;

[0128] When the user's purchase attribute is output as a user with high purchase possibility, the advertising delivery strategy is output as: the number of delivery times within the preset period is N3, and the advertising content is the advantages and characteristics of the product, providing product information and purchase guidance.

[0129] Specifically, the first preset purchase probability G1 refers to a preset lower threshold value for judging the probability of a user purchasing a product, and the second preset purchase probability G2 refers to a preset upper threshold value for judging the probability of a user purchasing a product. In this embodiment, G1=0.3 and G2=0.7 are set. The preset period refers to a preset length of time. In this embodiment, no limitation is imposed on the preset period, such as setting the preset period to one week. The number of placements refers to the number of advertisement placements within the preset period. In this embodiment, no limitation is imposed on the number of placements to N1, such as setting N1=2 times. The popular product refers to a product with high sales volume and popularity on the platform. In this embodiment, popular products are obtained through online statistics. The online statistics refer to a method of statistically analyzing the click-through rate and purchase rate of a product. The newly launched product refers to a product that is promoted online by the platform within the preset period of a new product. The preset period of a new product refers to the length of the most recent period of time from the current moment. In this embodiment, the preset period of a new product is set to The period is one week. The product information refers to the key features, value and differentiated advantages of the product delivered to users through advertising. This embodiment is layered according to the user's purchase probability and matches the information content. The purchase guidance refers to the operation of prompting users to complete the purchase behavior through copywriting, activities, etc. This embodiment sets the purchase guidance to welfare temptation. This embodiment does not limit the number of delivery to N2, such as setting N2=4 times. The value of the product refers to the benefits and advantages that the product can bring to users and the degree to which it meets user needs. This embodiment obtains the value of the product through user research. This embodiment does not limit the method of user research, such as questionnaire survey. The practicality refers to the operability and applicability of the product in actual life and work. This embodiment sets practical content, such as long battery life, through user research. This embodiment does not limit the number of delivery to N3, such as setting N3=6 times. The advantages and characteristics of the product refer to the unique strengths and competitive advantages of the product, such as unique design and lower price.

[0130] Specifically, the advertisement delivery module generates an advertisement delivery strategy by classifying the probability of users purchasing products, and dynamically adjusts the strategy in combination with the purchase time interval and the product promotion index to improve the pertinence and flexibility of advertisement delivery.

[0131] Specifically, the advertising delivery module compares the purchase time interval TG with the preset purchase time interval TG0, determines the key attributes of the purchase time interval TG based on the comparison result, and adjusts the advertising delivery strategy based on the determination result, wherein:

[0132] When TG≤TG0, the advertising delivery module determines that the key attribute of the purchase time interval TG is a short time interval and does not adjust the advertising delivery strategy;

[0133] When TG>TG0, the advertisement delivery module determines that the key attribute of the purchase time interval TG is a long time interval, and adjusts the advertisement delivery strategy:

[0134] When the advertising delivery module adjusts the advertising delivery strategy, for users with low purchase probability, the adjusted delivery times are set to N1', N1'=N1. After the preset delivery time, the probability G of the user purchasing the product is updated, and the purchase update coefficient is set to β, 1.1≤β≤1.5. The updated probability of the user purchasing the product is G', G'=β×G.

[0135] For users with medium purchase probability, the adjusted number of delivery times is set to N2`, N2` = 1.2 × N2;

[0136] For users with high purchase probability, the adjusted number of delivery times is set to N3`, N3` = 2.5 × N3;

[0137] The advertising module obtains the product's sales growth rate M1, inventory M2, and sales growth rate M3 of competing products based on a market research agency report, calculates the product promotion index C according to the formula C=0.4×M1+0.2×M2+0.4×M3, compares the product promotion index C with a preset promotion index C0, determines the degree of influence of the comparison result on the product promotion index C, and updates the purchase time interval TG based on the determination result, wherein:

[0138] When C≤C0, the advertising delivery module determines that the influence of the product promotion index C is low, and does not update the purchase time interval TG;

[0139] When C>C0, the advertising delivery module determines that the influence of the product promotion index C is high, and updates the purchase time interval TG. For users with low purchase possibility, N1`=N2`=N3` is set; for users with medium purchase possibility, N1`=N2`=N3` is set; for users with high purchase possibility, the advertising delivery strategy is not adjusted.

[0140] Specifically, the preset purchase time interval TG0 refers to a pre-set time length threshold for judging the key attributes of the purchase time interval. This embodiment does not limit the preset purchase time interval TG0. This embodiment sets TG0 = 25 days. The market research institution report refers to a report written by a professional market research institution after conducting extensive market surveys, collecting data, and analyzing and studying the market. This embodiment does not limit professional market research institutions, and relevant personnel in this field can freely choose according to actual needs. The sales growth rate of the product refers to the growth ratio of the sales quantity and sales amount of the product in a certain period compared with the sales quantity and sales amount of the previous period. This embodiment does not limit it to a certain period, such as The certain period is set to one month. The inventory refers to the number of products that have not been sold. The sales growth rate of the competing product refers to the growth ratio of the sales quantity and sales amount of the competing product in a certain period compared with the sales quantity and sales amount in the previous period. The competing product refers to the products of other companies that are in direct competition with the product. The preset promotion index C0 refers to a preset standard value for measuring the impact of the product promotion index. This embodiment does not limit the preset promotion index C0. This embodiment sets C0=0.5. The preset delivery duration refers to the preset length of time for measuring the duration of advertising delivery. The purchase update coefficient β refers to the coefficient for updating the probability G of users purchasing the product.

[0141] Specifically, the advertising delivery module adjusts the probability of users purchasing products based on the purchase time interval and the promotion index, so as to optimize the delivery frequency for different user layers, thereby improving the accuracy of advertising delivery and market adaptability. For users with medium purchase possibility, the adjusted number of delivery is set to N2`=1.2×N2, so as to avoid excessive disturbance while maintaining user attention, strengthen product value awareness by moderately increasing the contact frequency, and gradually promote users to high intention conversion. For users with medium purchase possibility, the adjusted number of delivery is set to N2`=2.5×N2 to increase the adjusted number of delivery, thereby seizing the user's potential demand window or leveraging market heat to accelerate decision-making, and improving the purchase probability of medium-intent users.

[0142] Specifically, the monitoring feedback module obtains product investment cost I0 and product investment income I according to the market research agency report, and uses the formula Calculate the advertising return on investment (ROI), compare the advertising return on investment (ROI) with the preset return rate (ROI0), determine the return attribute of the advertising return on investment (ROI) based on the comparison results, and optimize the advertising delivery strategy based on the determination results, including:

[0143] When ROI≥ROI0, the monitoring feedback module determines that the return attribute of the advertising return on investment ROI is high return, and does not optimize the advertising delivery strategy;

[0144] When ROI<ROI0, the monitoring feedback module determines that the return attribute of the advertising return on investment ROI is low, and optimizes the advertising delivery strategy:

[0145] For users with low purchase probability, the first optimization coefficient is set to θ1, θ1 = 1.7-3.2e -1.1×[(ROIO-ROI)-2 The optimized number of launches is N1``, N1``=θ1×N1`;

[0146] For users with medium purchase probability, the second optimization coefficient is set to θ2, θ2 = 1.2-3.3e -1.2×[(ROIO-ROI)+1] , the optimized delivery times are N2``, N2``=θ2×N2`;

[0147] For users with high purchase probability, the third optimization coefficient is set to θ3, θ3 = 1.5-3.2e -1.2×[(ROIO-ROI)+2] , the optimized number of delivery times is N3``, N3``=θ3×N3`.

[0148] Specifically, the product investment cost I0 refers to the total cost invested in the process of advertising the product, the product investment return I refers to the economic benefit obtained from the product advertising, the advertising return on investment (ROI) refers to an indicator that measures the benefits brought by advertising based on the product investment cost and the product investment return, the preset return rate (ROI) 0 refers to a preset value for judging the return attribute of the advertising return on investment (ROI), and this embodiment does not limit the advertising return on investment (ROI). This embodiment sets the advertising return on investment (ROI) = 160%. The first optimization coefficient refers to a coefficient for optimizing the number of advertising delivery times for users with low purchase likelihood when the advertising return on investment (ROI) is determined to have a low return. The second optimization coefficient refers to a coefficient for optimizing the number of advertising delivery times for users with medium purchase likelihood when the advertising return on investment (ROI) is determined to have a low return. The third optimization coefficient refers to a coefficient for optimizing the number of advertising delivery times for users with high purchase likelihood when the advertising return on investment (ROI) is determined to have a low return.

[0149] Specifically, the monitoring and feedback module calculates the advertising return on investment through the monitoring and feedback module to optimize the advertising delivery strategy, so as to optimize the advertising delivery strategy and thus improve the economic benefits and model performance of advertising delivery. For users with low purchase probability, the first optimization coefficient is set to θ1=0.8, so as to reduce invalid exposure to low-intention users, reduce advertising delivery costs, focus resources on more potential user groups, avoid excessive contact and cause user disgust, thereby improving the economy of the delivery strategy. For users with medium purchase probability, the second optimization coefficient is set to θ2=1.5, so as to strengthen users' memory of product value by moderately increasing the contact frequency when the advertising return on investment is low, thereby balancing costs and conversion efficiency. For users with high purchase probability, the third optimization coefficient is set to θ3=2.0, so as to shorten the decision-making cycle for high-intention users through high-frequency and precise contact, thereby quickly improving overall advertising revenue.

[0150] See also Figure 2 , which is a flow chart of the intelligent advertising delivery method based on user behavior analysis in this embodiment, the method includes:

[0151] Step S01, for collecting target advertising data;

[0152] Step S02 is used to obtain a target comprehensive feature vector based on the target advertising data, to construct a user behavior analysis model using a user behavior analysis model construction method, to construct a cold start model using a cold start model construction method, and to optimize the user behavior analysis model based on the cold start model;

[0153] Step S03 is used to generate an advertisement delivery strategy according to the user behavior analysis model using an advertisement delivery strategy generation method, and also to adjust the advertisement delivery strategy in real time;

[0154] Step S04 is used to calculate the advertising investment return rate to obtain the advertising investment return rate, and optimize the advertising delivery strategy according to the advertising investment return rate.

[0155] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. An intelligent advertising delivery system based on user behavior analysis, characterized in that: The system comprises: A data acquisition module is used to collect target advertising data; a behavior analysis module for acquiring a target comprehensive feature vector based on target advertising data, for constructing a user behavior analysis model using a user behavior analysis model construction method, for constructing a cold start model using a cold start model construction method, and for optimizing the user behavior analysis model based on the cold start model; An advertising delivery module, which is used to generate an advertising delivery strategy based on a user behavior analysis model using an advertising delivery strategy generation method, and is also used to adjust the advertising delivery strategy in real time; The monitoring feedback module is used to calculate the advertising return on investment, obtain the advertising return on investment, and optimize the advertising delivery strategy based on the advertising return on investment.

2. The intelligent advertising delivery system based on user behavior analysis according to claim 1 is characterized in that: The data acquisition module collects target advertising data, which includes e-commerce data and social media data. The e-commerce data is collected by embedding code in e-commerce websites and mobile applications, and the social media data is collected through an application programming interface. The e-commerce data includes the number of views, the number of clicks, the number of purchases, the price of purchased goods, the purchase time interval and the user's browsing time on the platform. The social media data includes user comments, user forwarding, user likes, user sharing, the number of users' social circle fans and historical advertising records on the e-commerce platform.

3. The intelligent advertising delivery system based on user behavior analysis according to claim 1 is characterized in that: The behavior analysis module obtains the target comprehensive feature vector based on the target advertisement data, and the process includes: Step J01: Calculate browsing frequency d1 according to the formula d1 = a1 / t, calculate click frequency d2 according to the formula d2 = a2 / t, and calculate purchase frequency d3 according to the formula d3 = a3 / t, where a1 is the number of browsing times, a2 is the number of clicks, a3 is the number of purchases, and t is the preset duration. The browsing frequency d1, click frequency d2, and purchase frequency d3 are used as behavioral frequency features. Step J02: Obtain product association features based on product association rules; Step J03: Compare the price h of the purchased product with the average price h0 of similar products. Based on the comparison results, determine the price trend of the purchased product h, and output the price sensitivity based on the determination results, where: When h<h0, the price h of the purchased product is judged to be high, and the price sensitivity output is low, and the low sensitivity is used as the price sensitivity feature; When h≥h0, the price h of the purchased product is judged to be low, the price sensitivity output is high, and the high sensitivity is used as the price sensitivity feature; Step J04: Use behavior frequency features, product association features, and price sensitivity features as e-commerce data features; Step J05: Perform sentiment analysis on user comments through natural language processing to obtain sentiment tendency features, perform interest analysis on user reposts, likes, and shares through natural language processing to obtain interest topic features, perform social analysis on user shares and the number of followers in the user's social circle through natural language processing to obtain social influence features, and use the sentiment tendency features, interest topic features, and social influence features as social media data features; Step J06: normalize the e-commerce data features and social media data features to obtain a target comprehensive feature vector.

4. The intelligent advertising delivery system based on user behavior analysis according to claim 3 is characterized in that: The behavior analysis module constructs a user behavior analysis model using a user behavior analysis model construction method, and the user behavior analysis model construction method includes: Step U01: divide the historical comprehensive database into a 70% user behavior training set, a 20% user behavior verification set, and a 10% user behavior test set; Step U02: Set the number of weak learners in the gradient boosting decision tree model parameters to 100, set the learning rate to 0.1, input the user behavior training set into the gradient boosting decision tree model after parameter setting, and train it to obtain the trained gradient boosting decision tree model; In step U03, the user behavior verification set is input into the trained gradient boosting decision tree model, and the parameters of the trained gradient boosting decision tree model are optimized to obtain the parameter-optimized gradient boosting decision tree model. The user behavior test set is then input into the parameter-optimized gradient boosting decision tree model for testing, and the test accuracy R is output.

5. The intelligent advertising delivery system based on user behavior analysis according to claim 4 is characterized in that: The user behavior analysis model construction method also includes: step U04, comparing the test accuracy R with the preset test accuracy R0, judging the compliance of the gradient boosting decision tree model after parameter optimization based on the comparison result, and outputting the gradient boosting decision tree model after parameter optimization based on the judgment result.

6. The intelligent advertising delivery system based on user behavior analysis according to claim 5 is characterized in that: The behavior analysis module constructs a cold start model using a cold start model construction method, and the cold start model construction method includes: Step Q01: compare the preset advertisement with the historical advertisement data search library, determine the existence of the preset advertisement in the historical advertisement data search library based on the comparison result, and output the definition of the preset advertisement based on the determination result, wherein: When the existence of the preset delivery advertisement in the historical advertisement data search library is not found, the definition of the preset delivery advertisement is output as a new advertisement; When the existence of the preset delivery advertisement in the historical advertisement data search library is "exist", the definition of the preset delivery advertisement is output as an old advertisement; Step Q02: Compare the preset advertising e-commerce platform with the historical advertising e-commerce platform database, determine the existence of the preset advertising e-commerce platform in the historical advertising e-commerce platform database based on the comparison result, and output the definition of the preset advertising e-commerce platform based on the determination result, wherein: When the preset advertising delivery e-commerce platform does not exist in the historical advertising e-commerce platform database, the definition of the preset advertising delivery e-commerce platform is output as a new platform; When the preset advertising delivery e-commerce platform exists in the historical advertising e-commerce platform database, the definition of the preset advertising delivery e-commerce platform is output as an old platform; Step Q03, comparing the user's platform browsing time t1 within the preset time with the preset browsing time t0, judging the attributes of the user's platform browsing time t1 within the preset time based on the comparison result, and outputting the user type based on the judgment result; Step Q04: divide 70% of the target comprehensive feature vector into a cold start training set, and divide 30% of the target comprehensive feature vector into a cold start validation set; Step Q05: Select the prototype network algorithm as the cold start model, initialize the weights and biases of the cold start model, input the cold start training set into the cold start model, and calculate the output of the cold start model; Step Q06: Calculate the loss function value based on the output and label of the cold start model, calculate the gradient through the back propagation algorithm, and update the weights and bias of the cold start model. Repeat the forward propagation, loss function calculation and back propagation process to obtain the trained cold start model. In step Q07, the cold start validation set is input into the trained cold start model for testing, and the trained cold start model with an accuracy rate of 90% is output as the cold start model.

7. The intelligent advertising delivery system based on user behavior analysis according to claim 6, characterized in that: The behavior analysis module obtains the time t1 that the user browses the platform within the preset time t2, calculates the user like frequency according to the formula Fd=f / t2, where f is the number of likes of the user, calculates the user comment frequency FP according to the formula FP=f2 / t2, where f2 is the number of comments of the user, and calculates the matching degree P between the user group and the platform according to the formula P=ω1×t1+ω2×Fd+ω3×FP, where ω1 is the weight parameter of the user browsing the platform time t1 within the preset time t2, ω2 is the weight parameter of the user like frequency, and ω3 is the weight parameter of the user comment frequency. The matching degree P between the user group and the platform is compared with the preset matching degree P0, and the matching degree between the user group and the platform is judged according to the comparison result, and the cold start model is adjusted according to the judgment result.

8. The intelligent advertising delivery system based on user behavior analysis according to claim 1, characterized in that: The advertisement delivery module generates an advertisement delivery strategy through an advertisement delivery strategy generation method, and the advertisement delivery strategy generation method includes: Step K01: The probability G of a user purchasing a product obtained through a user behavior analysis model is compared with various preset purchase probabilities, including a first preset purchase probability G1 and a second preset purchase probability G2. The likelihood of the user purchasing the product is determined based on the comparison results, and the user's purchase attributes are output based on the determination results, where: When G≤G1, the advertisement delivery module determines that the probability G of the user purchasing the product is low, and outputs the user's purchase attribute as a user with low purchase probability; When G1<G≤G2, the advertisement delivery module determines that the probability G of the user purchasing the product is medium, and outputs the user's purchase attribute as a medium purchase possibility user; When G>G2, the advertisement delivery module determines that the probability G of the user purchasing the product is high, and outputs the user's purchase attribute as a high purchase possibility user; Step K02: When the user's purchase attribute is output as a low purchase possibility user, the advertising delivery strategy is output as follows: the number of delivery times within the preset period is N1, the advertising content is popular products and newly launched products, and product information and purchase guidance are provided; When the user's purchase attribute is output as a medium purchase possibility user, the output advertising delivery strategy is: the number of delivery times within the preset period is N2, and the advertising content is the value and practicality of the product, providing product information and purchase guidance; When the user's purchase attribute output is a high purchase possibility user, the output advertising delivery strategy is: the number of delivery times within the preset period is N3, and the advertising content is the advantages and characteristics of the product, providing product information and purchase guidance; The advertising delivery module compares the purchase time interval TG with the preset purchase time interval TG0, determines the key attributes of the purchase time interval TG according to the comparison result, and adjusts the advertising delivery strategy according to the determination result; When the advertising delivery module adjusts the advertising delivery strategy, for users with low purchase probability, the adjusted delivery times are set to N1', N1'=N1. After the preset delivery time, the probability G of the user purchasing the product is updated, and the purchase update coefficient is set to β, 1.1≤β≤1.

5. The updated probability of the user purchasing the product is G', G'=β×G. For users with medium purchase probability, the adjusted number of delivery times is set to N2`, N2` = 1.2 × N2; For users with high purchase probability, the adjusted number of delivery times is set to N3`, N3` = 2.5 × N3; The advertising delivery module obtains the product's sales growth rate M1, inventory M2, and sales growth rate M3 of competing products based on a market research agency report, calculates the product promotion index C according to the formula C=0.4×M1+0.2×M2+0.4×M3, compares the product promotion index C with a preset promotion index C0, determines the degree of influence of the comparison result on the product promotion index C, and updates the purchase time interval TG based on the determination result.

9. The intelligent advertising delivery system based on user behavior analysis according to claim 1, characterized in that: The monitoring feedback module obtains product investment cost I0 and product investment income I according to the market research agency report, and the formula Calculate the advertising return on investment (ROI), compare the advertising return on investment (ROI) with the preset return rate (ROI0), determine the return attribute of the advertising return on investment (ROI) based on the comparison results, and optimize the advertising delivery strategy based on the determination results, including: When ROI≥ROI0, the monitoring feedback module determines that the return attribute of the advertising return on investment ROI is high return, and does not optimize the advertising delivery strategy; When ROI<ROI0, the monitoring feedback module determines that the return attribute of the advertising return on investment ROI is low, and optimizes the advertising delivery strategy: For users with low purchase probability, the first optimization coefficient is set to θ1, θ1 = 1.7-3.2e -1.1×[(ROIO-ROI)-2 The optimized number of launches is N1``, N1``=θ1×N1`; For users with medium purchase probability, the second optimization coefficient is set to θ2, θ2 = 1.2-3.3e -1.2×[(ROIO-ROI)+1] , the optimized delivery times are N2``, N2``=θ2×N2` For users with high purchase probability, the third optimization coefficient is set to θ3, θ3 = 1.5-3.2e -1.2×[(ROIO-ROI)+2] , the optimized number of delivery times is N3``, N3``=θ3×N3`.

10. A method for applying to the intelligent advertising delivery system based on user behavior analysis according to any one of claims 1 to 9, characterized in that: include: Step S01, for collecting target advertising data; Step S02 is used to obtain a target comprehensive feature vector based on the target advertising data, to construct a user behavior analysis model using a user behavior analysis model construction method, to construct a cold start model using a cold start model construction method, and to optimize the user behavior analysis model based on the cold start model; Step S03 is used to generate an advertisement delivery strategy according to the user behavior analysis model using an advertisement delivery strategy generation method, and also to adjust the advertisement delivery strategy in real time; Step S04 is used to calculate the advertising investment return rate to obtain the advertising investment return rate, and optimize the advertising delivery strategy according to the advertising investment return rate.

Images