Advertisement generation marketing method based on user behavior driving

Abstract

The invention provides an advertisement generation marketing method based on user behavior driving, and relates to the technical field of advertisement generation marketing methods.According to the method, shopping behavior characteristics of platform users are calculated by obtaining historical shopping behavior data, browsing behavior data, social behavior data and daily behavior data of each user in a platform; judging a shopping intention value according to the browsing behavior data, calculating a social association value, judging whether the social association value belongs to a social relationship, calculating social features of the social relationship, calculating a shopping idle value and an environment influence coefficient of the user, determining a shopping situation coefficient of the user, and obtaining historical advertisements watched by each user in the platform; and taking each advertisement content as a generation factor, updating shopping behavior characteristics, daily behavior characteristics and shopping situation coefficients at an advertisement opening moment, establishing an advertisement generation model based on the generative adversarial network according to the behavior factor characteristic set, and outputting the advertisement content according to the behavior factor characteristic set.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement generation and marketing methods, and in particular to an advertisement generation and marketing method driven by user behavior. Background Art

[0002] With the rapid development of Internet advertising and e-commerce, the accuracy of advertising generation has gradually become an important issue in marketing. At present, traditional advertising generation methods often rely on simple rules or outdated data, usually rely on a single or limited data source, fail to fully consider the multi-dimensional information of user behavior, and often ignore the diversity and real-time nature of user behavior, such as user shopping behavior, social interaction, environmental factors, and advertising viewing history. In addition, existing technologies mostly focus on advertising push based on static or outdated data, lack of real-time updates and personalized customization, resulting in insufficient relevance and timeliness of advertising content, affecting the effectiveness of advertising. Therefore, how to accurately predict users' needs and generate personalized advertising content based on their various behavioral characteristics has become an urgent problem to be solved in the field of advertising generation.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0004] The purpose of the present invention is to provide an advertisement generation and marketing method driven by user behavior to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A user behavior driven advertising generation marketing method, the specific steps include: Step 1: Obtain the historical shopping behavior data and browsing behavior data of each user on the platform, and calculate the user's interest in each product type on the platform, as well as shopping price preferences, discount tolerance and shopping cycle based on the shopping behavior data and browsing behavior data to form shopping behavior characteristics; Step 2: Obtain the historical social behavior data of each user on the platform, calculate the social association value between the user and other users through the historical social behavior data, determine whether it is a social relationship, and calculate the commodity influence coefficient and commodity influence timeliness of each social user based on the historical social behavior data of users in social relationship to form social characteristics; Step 3: Obtain the historical daily behavior data of each user on the platform, calculate the user's shopping idle value, and obtain the meteorological data of the user's area when shopping based on the historical shopping time, and calculate the environmental impact coefficient. Calculate the user's shopping situation coefficient based on the shopping idle value and the environmental impact coefficient; Step 4: Obtain the historical advertisements watched by each user on the platform, and obtain the content of each advertisement, including product type, discount price and advertisement duration, as generation factors, and update the shopping behavior characteristics, shopping intention value, daily behavior characteristics and shopping situation coefficient at the time of opening the advertisement, and form a behavioral factor feature set with the shopping intention value determined based on the shopping behavior data and browsing behavior data of the previous day; Step 5: Establish an advertisement generation model based on a generative adversarial network, take the updated behavioral factor feature set as the input of the model, and the advertisement content as the output of the model to train the model. By taking the real-time behavioral factor feature set as the input of the model, the advertisement content is output.

[0006] Furthermore, the shopping behavior characteristics include the type of goods purchased, the price of the goods, the time of purchase, the preferential price paid, and the use period of the goods; The browsing behavior data includes the type of goods browsed and the number of times the goods were browsed; The steps for calculating the interest level of each commodity type are as follows: classify the commodities purchased each time into categories, and calculate the interest level of each commodity type according to the commodity type: in, is the interest level of the product type at the current moment, is the interest level of the product type at the time of last purchase, The number of purchases of the product type up to the current moment. The incremental value of interest in purchasing the product type. The time between the current moment and the last time the product type was purchased. The usage cycle of the product type.

[0007] Furthermore, the calculation formula of the shopping price preference is: in, For shopping price preference, For the The price of the purchase, is the total number of historical shopping times; The calculation formula of the shopping cycle is: in, For the shopping cycle, For each product type, For the The time of the first purchase of the product type; The calculation formula of the preferential tolerance is: in, To provide preferential tolerance, For the The preferential price paid for the second preferential shopping The number of times you purchase discount items.

[0008] Furthermore, the shopping intention value is determined based on the shopping behavior characteristics and browsing behavior data within the current day: in, is the shopping intention value, The number of views of the same product type in one day. The time when the type of product was last purchased, The usage cycle of the product type.

[0009] Furthermore, the social behavior data includes the number of shares, likes and comments on the product; The calculation steps of the social association value are: By calculating the delay time of purchasing the same product, we can determine whether it has an impact based on the delay time: in, is the delay time, The time when the user purchased the product. The time when other users purchased the same product; when When the social influence effect is judged; in, The delay time threshold for purchasing goods; The social association value is calculated by counting the number of users with social influence who purchase the same product and their social behavior data: in, is the social connection value, The number of items purchased by the user. The number of the same products purchased by the user and the user with social influence, The number of shares, likes and comments made by users on the product, The number of shares, likes and comments on the same product by users and users with social influence. The weights of purchasing the same product, sharing, liking and commenting are respectively: ; when When judging, it belongs to the social scope.

[0010] Furthermore, the calculation formula for the impact timeliness is: in, To affect timeliness, Buying for users and buying for social users The delay time when the same product is To purchase the same quantity of goods, and All are positive integers.

[0011] Furthermore, the calculation formula of the commodity impact coefficient is: in, is the commodity impact coefficient, is the total number of items purchased by the user. The number of times the same product was purchased by the social user.

[0012] Furthermore, the historical daily behavior includes the user's real-time coordinate position and number of steps; The meteorological data include light intensity, temperature, and humidity; The calculation formula of the environmental impact coefficient is: in, is the environmental impact coefficient, is the light intensity, is the temperature value, is the humidity value, are the weights of light intensity, temperature and humidity respectively. ; The calculation formula of the shopping idle value is: in, is the shopping idle value, For the current fixed time period The number of steps moved inside, The current time and fixed time period respectively The previous position, are the weights of moving steps and position changes respectively, ; The calculation formula of shopping situation coefficient is: in, is the shopping situation coefficient.

[0013] Furthermore, the advertisement generation model is based on a generative adversarial neural network, including a generator and a judger: The generator maps the input to the output of the advertising content through a neural network, and the calculation formula is: in, For the content of the advertisement, is the real-time behavioral factor feature set, are the parameters of the generator, is the mapping function of the generator; The goal of the discriminator is to determine whether the ad content is generated by the generator, and based on this judgment it gives a probability value: in, Indicates the output advertising content. is the weight of the discriminator, is the bias of the discriminator, is the sigmoid function.

[0014] Furthermore, the mapping function of the generator adopts a recurrent neural network: in, is the hidden state vector at the current moment, is the input value at the current moment, is the weight matrix of the hidden layer, is the weight matrix of the input layer, is the bias term of the hidden layer, is the hyperbolic tangent function, is the current output, is the output weight matrix, is the output bias term; Among them, the mathematical expressions of the sigmoid function and the hyperbolic tangent function are: in, is the sigmoid function, is the input of the sigmoid function, is the hyperbolic tangent function, is the input of the hyperbolic tangent function.

[0015] Compared with the prior art, the beneficial effect of the present invention is that the present invention obtains the historical shopping behavior data, browsing behavior data, social behavior data, and daily behavior data of each user in the platform, calculates the shopping behavior characteristics of the platform users, and judges the shopping intention value according to the browsing behavior data, and judges whether it belongs to a social relationship by calculating the social association value, and then calculates the social characteristics of the social relationship, and determines the user's shopping situation coefficient by calculating the user's shopping idle value and environmental impact coefficient, obtains the historical advertisements watched by each user in the platform, takes each advertisement content as a generation factor, and updates the shopping behavior characteristics, daily behavior characteristics, and shopping situation coefficient at the time of opening the advertisement to form a behavior factor feature set, establishes an advertisement generation model based on a generative adversarial network, and outputs the advertisement content according to the behavior factor feature set; The present invention uses a generative adversarial network to establish an advertisement generation model based on the user's shopping behavior, social behavior, daily behavior and environmental factors, so that the personalization and accuracy of the advertisement content are significantly improved. By acquiring and analyzing these multi-dimensional user behavior data in real time, the system can generate relevant advertisements based on dynamic factors such as user interests, social relationships and shopping situations, thereby improving the conversion rate of advertisements and user participation. In addition, the practice of generating advertisements based on a real-time updated behavioral factor feature set makes the advertisement content more in line with the user's current needs and situations, thereby effectively solving the problems of timeliness and lack of personalization of traditional advertising recommendation systems, and further improving the efficiency of advertisement generation and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0017] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0018] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0019] Example

[0020] See also Figure 1 , the present invention provides a technical solution: A user behavior driven advertising generation marketing method, the specific steps include: Step 1: Obtain the historical shopping behavior data and browsing behavior data of each user on the platform, and calculate the user's interest in each product type on the platform, as well as shopping price preferences, discount tolerance and shopping cycle based on the shopping behavior data and browsing behavior data to form shopping behavior characteristics.

[0021] Browsing behavior can reflect a user's interest in a specific product or product type. For example, if a user browses a certain product category or brand but does not purchase it immediately, it indicates that the user has potential interest in the product. Based on this data, relevant ads can be pushed accurately to increase the likelihood of users clicking on ads.

[0022] In this embodiment, the shopping behavior characteristics include the type of goods purchased, the price of the goods, the time of purchase, the preferential price paid, and the use period of the goods; The browsing behavior data includes the type of goods browsed and the number of times the goods were browsed; The steps for calculating the interest level of each commodity type are as follows: classify the commodities purchased each time into categories, and calculate the interest level of each commodity type according to the commodity type: in, is the interest level of the product type at the current moment, is the interest level of the product type at the time of last purchase, The number of purchases of the product type up to the current moment. The incremental value of interest in purchasing the product type. The time between the current moment and the last time the product type was purchased. The usage cycle of the product type.

[0023] Interest refers to the user's preference for a certain product type or category. By understanding the user's interest, the advertising platform can accurately recommend products related to the user's interests and improve the relevance of the advertisement. This makes the advertisement closer to the user's needs, reduces the interruption of irrelevant advertisements, and improves user satisfaction. At the same time, more targeted advertising content can attract users to click, and increase the click-through rate and conversion rate of the advertisement.

[0024] In this embodiment, the calculation formula of the shopping price preference is: in, For shopping price preference, For the The price of the purchase, The total number of historical shopping times.

[0025] Shopping price preference refers to the user's budget range or price sensitivity when purchasing goods. By analyzing the price range of goods purchased by users in the past, it is possible to infer the user's price acceptance and consumption preferences in specific product categories. Based on the user's price preference, the advertising platform can recommend products or promotions that meet the user's budget and avoid displaying products that are too high or too low in price.

[0026] The calculation formula of the shopping cycle is: in, For the shopping cycle, For each product type, For the The time when the product type was purchased.

[0027] A shopping cycle refers to the frequency of a user's shopping or periodic purchasing behavior within a certain period of time. For example, some users may purchase a certain type of product every month, while others may purchase it only during a certain season or at a specific time. Based on the user's shopping cycle, ads can be pushed at the time when the user is expected to purchase a certain product, thereby improving the timeliness of the ads. By understanding the shopping cycle, the platform can proactively push related products or remind users to purchase when the user is about to enter the next purchase cycle, thereby promoting repeat purchases.

[0028] The calculation formula of the preferential tolerance is: in, To provide preferential tolerance, For the The preferential price paid for the second preferential shopping The number of times you purchase discount items.

[0029] Discount tolerance refers to the user's acceptance of discounts and promotions. Some users may make purchases under certain discount conditions, while other users may not be very sensitive to discounts. Based on the user's discount tolerance, ads can display promotions that are most likely to attract users' interest, such as full-amount discounts, discounts, limited-time offers, etc. By accurately grasping the user's discount tolerance, the platform can increase user stickiness and enhance user loyalty through appropriate promotions.

[0030] In this embodiment, the shopping intention value is determined based on the shopping behavior data and browsing behavior data in the current day: in, is the shopping intention value, The number of views of the same product type in one day. The time when the product type was last purchased. The usage cycle of the product type.

[0031] The shopping intention value is a value calculated based on the user's current shopping behavior and browsing data, representing the strength of the user's shopping intention at a certain moment. The higher the shopping intention value, the more likely the user is to make a purchase. For users with higher shopping intention values, the advertising platform can prioritize showing them related products, thereby improving the effectiveness of advertising and achieving accurate advertising delivery. The shopping intention value is a dynamically changing indicator, and advertisements can adjust their delivery strategies in real time based on changes in this value to avoid excessive push to users with low intention values.

[0032] These shopping behavior characteristics provide rich user portraits for ad generation, making ad delivery more accurate, personalized and timely. This can not only increase the click-through rate and conversion rate of ads, but also effectively improve the user's shopping experience and the overall benefits of the platform. By better understanding user needs and behaviors, advertising platforms can more intelligently match ad content to achieve the goal of optimizing ad delivery results.

[0033] Step 2: Obtain the historical social behavior data of each user on the platform, calculate the social association value between the user and other users through the historical social behavior data, and determine whether it is a social relationship. According to the historical social behavior data of users in social relationships, calculate the product influence coefficient and product influence timeliness of each social user to form social characteristics.

[0034] Social relationships are key to information dissemination. Interactions and relationships between users may greatly affect the efficiency of advertising information dissemination. For example, if a user's friends or followers are interested in a product, they may be more likely to buy it than other users. Therefore, judging whether there is a social relationship can identify more valuable target groups and help brands target potential customers more accurately. By identifying social relationships between users, it is possible to avoid pushing ads to objects that have no relationship with the user, thereby increasing the relevance of the ads and user engagement, and improving advertising effectiveness. Users in social relationships are more likely to believe in and buy products, so the conversion rate of advertising is higher.

[0035] In this embodiment, the social behavior data includes the number of shares, likes, and comments on the product; The calculation steps of the social association value are: By calculating the delay time of purchasing the same product, we can determine whether it has an impact based on the delay time: in, is the delay time, The time when the user purchased the product. The time when other users purchased the same product; when When the social influence effect is judged; in, The delay time threshold for purchasing goods; The social association value is calculated by counting the number of users with social influence who purchase the same product and their social behavior data: in, is the social connection value, The number of items purchased by the user. The number of the same products purchased by the user and the user with social influence, The number of shares, likes and comments made by users on the product, The number of shares, likes and comments on the same product by users and users with social influence. The weights of purchasing the same product, sharing, liking and commenting are respectively: ; when When judging, it belongs to the social scope.

[0036] In this embodiment, the calculation formula for the timeliness of the impact is: in, To affect timeliness, Buying for users and buying for social users The delay time when the same product is To purchase the same quantity of goods, and All are positive integers.

[0037] The timeliness of influence reflects the time of influence of other social users on the user after they purchase the product. Since the influence of social behavior on the product is time-sensitive, especially when a user just purchases, the product is in a fresh period, which will increase the user's desire to buy and induce the user to have a strong interest. After a long time, this interest gradually fades. Therefore, evaluating the timeliness of influence can help advertisers determine when to push ads to which users, thereby improving the timeliness and effectiveness of advertising.

[0038] In this embodiment, the calculation formula of the commodity impact coefficient is: in, is the commodity impact coefficient, is the total number of items purchased by the user, The number of times the same product was purchased by the social user.

[0039] Each social user has different influence in their social network. Some users' recommendations, shares or comments may have a greater impact and can lead to more attention or purchase behavior. Therefore, by calculating the product influence coefficient, we can evaluate the potential impact of each social user on the behavior of other users. Pushing ads to those social users with higher influence can have a wider impact in their social circles, thereby improving the overall effect of the ads.

[0040] Step 3: Obtain the historical daily behavior data of each user on the platform, calculate the user's shopping idle value, and obtain the meteorological data of the user's area when shopping based on the historical shopping time, and calculate the environmental impact coefficient. Calculate the user's shopping situation coefficient based on the shopping idle value and the environmental impact coefficient.

[0041] The meteorological data include light intensity, temperature, and humidity; The calculation formula of the environmental impact coefficient is: in, is the environmental impact coefficient, is the light intensity, is the temperature value, is the humidity value, are the weights of light intensity, temperature and humidity respectively. ; The environmental impact coefficient measures the degree of influence of the external environment on the user's shopping behavior by analyzing the meteorological data of the user's area, such as temperature and humidity. This coefficient indicates the possibility or tendency of users to shop under specific environmental conditions. Environmental factors such as weather, temperature, humidity, etc. have a significant impact on user shopping behavior. For example, cold weather may make people unwilling to go out shopping, and hot weather may prompt users to choose indoor shopping. By calculating the environmental impact coefficient, we can better understand the impact of the external environment on users' shopping behavior.

[0042] The calculation formula of the shopping idle value is: in, is the shopping idle value, For the current fixed time period The number of steps moved inside, The current time and fixed time period respectively The previous position, are the weights of moving steps and position changes respectively, .

[0043] The shopping idle value refers to the probability that a user has free time to shop within a certain period of time. By calculating this value, the advertising platform can accurately grasp the user's shopping time window, thereby avoiding pushing ads when the user is busy or inconvenient to shop, ensuring that the timing of ad push is more appropriate. By accurately placing ads during the user's free time or high-probability shopping period, the click-through rate and conversion rate of ads can be improved.

[0044] The calculation formula of shopping situation coefficient is: in, is the shopping situation coefficient.

[0045] The shopping idleness value and the environmental impact coefficient are combined to form a more comprehensive shopping situation coefficient. The shopping situation coefficient is a comprehensive indicator that takes into account the user's shopping idleness value and the environmental impact coefficient. It represents the possibility and suitability of users to shop under specific time and environmental conditions, and provides accurate timing and situation judgment for the production and delivery of advertisements. For example, if a user is in an idle time period and the weather is suitable for shopping, the shopping situation coefficient may be high, indicating that this moment is more suitable for shopping.

[0046] Step 4: Obtain the historical advertisements watched by each user on the platform, and obtain the content of each advertisement, including product type, discount price and advertisement duration, as generation factors, and update the shopping behavior characteristics, shopping intention value, daily behavior characteristics and shopping situation coefficient at the moment of opening the advertisement, and form a behavioral factor feature set with the shopping intention value judged based on the shopping behavior data and browsing behavior data of the previous day.

[0047] Product type refers to the category or type of product displayed in the advertisement. For example, electronic products, clothing, household goods, beauty products, etc. By understanding the types of products in the ads that users have viewed in the past, the platform can determine the user's areas of interest or potential needs. By identifying the types of products in the ads that users have viewed in the past, the advertising platform can push relevant ads based on the user's interests and preferences. For example, if a user frequently views electronic product ads, the platform can push more electronic product-related ads to increase the relevance and attractiveness of the ads.

[0048] A preferential price refers to the discount or promotional price of the product displayed in the advertisement. The platform collects users' responses to advertisements with different preferential prices, including whether they click, buy, etc. By analyzing users' interactions with advertisements with different preferential prices, the advertising platform can identify users' preferences for specific discounts. This provides a basis for generating advertisements with similar preferential prices in the future, thereby improving the conversion rate of advertisements.

[0049] Ad duration refers to the length of time that an ad is played, that is, the duration of time that the ad content is continuously displayed. Ad duration is crucial to maintaining user attention. Ads that are too long or too short may affect the user's viewing experience and reaction. By understanding the user's acceptance of different ad durations, the platform can adjust the length of the generated ads to ensure that the ad duration matches the user's viewing habits. For example, if the user prefers short ads, the system will prioritize generating short ads.

[0050] Step 5: Establish an advertisement generation model based on a generative adversarial network, take the updated behavioral factor feature set as the input of the model, and the advertisement content as the output of the model to train the model. By taking the real-time behavioral factor feature set as the input of the model and outputting the advertisement content, the platform can accurately identify the user's interest preferences by analyzing the user's historical purchase behavior. For example, if a user frequently purchases high-end electronic products, relevant high-end electronic product advertisements can be pushed to improve the relevance of the advertisements. In this embodiment, the historical daily behavior includes the user's real-time coordinate position and number of steps.

[0051] Generative adversarial networks consist of two parts: a generator and a discriminator. The goal of the generator is to generate fake data that is similar to real data, while the discriminator evaluates the difference between the generated data and the real data, thereby forcing the generator to produce more realistic content. Generative adversarial networks can make advertising content more creative and diverse by continuously optimizing the output of the generator, avoiding the monotony of advertising content. For ad generation, diverse creativity can attract different user groups and increase user engagement and click-through rate. Through adversarial training, the generator can generate text or images that are highly close to real advertising content. This helps ensure that the generated advertising content meets user expectations and behaviors, increasing the effectiveness of advertising. With the adversarial training of ad generation, the generator can gradually adapt to real-time updated user behaviors and market demands, making advertising content timely and flexible.

[0052] In this embodiment, the advertisement generation model is based on a generative adversarial neural network, including a generator and a judger: The generator maps the input to the output of the advertising content through a neural network, and the calculation formula is: in, For the content of the advertisement, is the real-time behavioral factor feature set, are the parameters of the generator, is the mapping function of the generator; The goal of the discriminator is to determine whether the ad content is generated by the generator, and based on this judgment it gives a probability value: in, Indicates the output advertising content. is the weight of the discriminator, is the bias of the discriminator, is the sigmoid function.

[0053] The mapping function of the generative adversarial network, that is, the architecture of the generator, uses a recurrent neural network to better handle timing and contextual dependencies, allowing advertising content to be more effectively adjusted to dynamically changing user behaviors.

[0054] Recurrent neural networks are good at processing time series and sequence data, and can learn the order relationship and context information between words in advertising copy. In advertising generation, advertising copy often contains multiple elements, and these elements have a certain order and dependency relationship. Through its inherent memory mechanism, recurrent neural networks can capture the context and grammatical structure of words in advertising copy and generate more natural and coherent advertising content.

[0055] In the user behavior-driven advertising generation model, the user's real-time behavioral factor feature set usually changes over time. Recurrent neural networks can dynamically generate personalized advertising content based on these changing input features, ensuring that the advertisements accurately match the user's needs at different time points.

[0056] Using generative adversarial networks and recurrent neural networks as ad generation models can combine the advantages of generated diversity and time series modeling to effectively generate personalized and highly relevant ad content in real time based on user behavior. This approach can not only increase the attractiveness of ads, but also ensure the dynamic adaptability and timeliness of ad content, thereby significantly improving the effectiveness of advertising marketing.

[0057] In this embodiment, the mapping function of the generator adopts a recurrent neural network: in, is the hidden state vector at the current moment, is the input value at the current moment, is the weight matrix of the hidden layer, is the weight matrix of the input layer, is the bias term of the hidden layer, is the hyperbolic tangent function, is the current output, is the output weight matrix, is the output bias term; Among them, the mathematical expressions of the sigmoid function and the hyperbolic tangent function are: in, is the sigmoid function, is the input of the sigmoid function, is the hyperbolic tangent function, is the input of the hyperbolic tangent function.

[0058] Through shopping behavior characteristics, daily behavior characteristics and shopping context coefficients, the model can dynamically generate advertising content based on users' real-time needs and interests, thereby achieving highly personalized advertising recommendations. These input features help the advertising generation model accurately identify users' preferences and behavior patterns, so that the product type, preferential prices and advertising duration of the advertising content can match the actual needs of users, improve the relevance and attractiveness of the advertisement, and thus increase user engagement and conversion rate.

[0059] In addition, this ad generation method based on real-time features can enhance the timeliness and accuracy of ad content, avoiding the monotony and obsolescence of static ad content. By continuously optimizing ad generation strategies, ads can not only be pushed to the right users at the right time, but also adjusted according to the user's shopping context and interest fluctuations, thereby improving ad effectiveness and user experience.

[0060] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0061] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0062] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A user behavior driven advertising generation marketing method, characterized in that: The specific steps include: Step 1: Obtain the historical shopping behavior data and browsing behavior data of each user on the platform, and calculate the user's interest in each product type on the platform, as well as shopping price preferences, discount tolerance and shopping cycle based on the shopping behavior data and browsing behavior data to form shopping behavior characteristics; Step 2: Obtain the historical social behavior data of each user on the platform, calculate the social association value between the user and other users through the historical social behavior data, determine whether it is a social relationship, and calculate the commodity influence coefficient and commodity influence timeliness of each social user based on the historical social behavior data of users in social relationship to form social characteristics; Step 3: Obtain the historical daily behavior data of each user on the platform, calculate the user's shopping idle value, and obtain the meteorological data of the user's area when shopping based on the historical shopping time, and calculate the environmental impact coefficient. Calculate the user's shopping situation coefficient based on the shopping idle value and the environmental impact coefficient; Step 4: Obtain the historical advertisements watched by each user on the platform, and obtain the content of each advertisement, including product type, discount price and advertisement duration, as generation factors, and update the shopping behavior characteristics, shopping intention value, daily behavior characteristics and shopping situation coefficient at the time of opening the advertisement, and form a behavioral factor feature set with the shopping intention value determined based on the shopping behavior data and browsing behavior data of the previous day; Step 5: Establish an advertisement generation model based on a generative adversarial network, take the updated behavioral factor feature set as the input of the model, and the advertisement content as the output of the model to train the model. By taking the real-time behavioral factor feature set as the input of the model, the advertisement content is output.

2. The method for generating advertisements based on user behavior drive according to claim 1, characterized in that: Shopping behavior characteristics include the type of goods purchased, the price of the goods, the time of purchase, the preferential price paid, and the life cycle of the goods; The browsing behavior data includes the type of goods browsed and the number of times the goods were browsed; The steps for calculating the interest level of each commodity type are as follows: classify the commodities purchased each time into categories, and calculate the interest level of each commodity type according to the commodity type: in, is the interest level of the product type at the current moment, is the interest level of the product type at the time of last purchase, The number of purchases of the product type up to the current moment. The incremental value of interest in purchasing the product type. The time between the current moment and the last time the product type was purchased. The usage cycle of the product type.

3. The method for generating advertisements based on user behavior as claimed in claim 2, characterized in that: The calculation formula of the shopping price preference is: in, For shopping price preference, For the The price of the purchase, is the total number of historical shopping times; The calculation formula of the shopping cycle is: in, For the shopping cycle, For each product type, For the The time of the first purchase of the product type; The calculation formula of the preferential tolerance is: in, To provide preferential tolerance, For the The preferential price paid for the second preferential shopping The number of times you purchase discount items.

4. The method for generating advertisements based on user behavior drive according to claim 3, characterized in that: Determine the shopping intention value based on the shopping behavior data and browsing behavior data within the current day: in, is the shopping intention value, The number of views of the same product type in one day. The time when the type of product was last purchased, The usage cycle of the product type.

5. The method for generating advertisements based on user behavior drive according to claim 1, characterized in that: The social behavior data includes the number of shares, likes and comments on the product; The calculation steps of the social association value are: By calculating the delay time of purchasing the same product, we can determine whether it has an impact based on the delay time: in, is the delay time, The time when the user purchased the product. The time when other users purchased the same product; when When the social influence effect is judged; in, The delay time threshold for purchasing goods; The social association value is calculated by counting the number of users with social influence who purchase the same product and their social behavior data: in, is the social connection value, The number of items purchased by the user. The number of the same products purchased by the user and the user with social influence, The number of shares, likes and comments made by users on the product, The number of shares, likes and comments on the same product by users and users with social influence. The weights of purchasing the same product, sharing, liking and commenting are respectively: ; when When judging, it belongs to the social scope.

6. The method for generating advertisements based on user behavior drive according to claim 5, characterized in that: The calculation formula for the timeliness of the impact is: in, To affect timeliness, Buying for users and buying for social users The delay time when the same product is To buy the same quantity of goods, and All are positive integers.

7. The method for generating advertisements based on user behavior drive according to claim 6, characterized in that: The calculation formula of the commodity impact coefficient is: in, is the commodity impact coefficient, is the total number of items purchased by the user, The number of times the same product was purchased by the social user.

8. The method for generating advertisements based on user behavior drive according to claim 1, characterized in that: The historical daily behavior includes the user's real-time coordinate location and number of steps; The meteorological data include light intensity, temperature, and humidity; The calculation formula of the environmental impact coefficient is: in, is the environmental impact coefficient, is the light intensity, is the temperature value, is the humidity value, are the weights of light intensity, temperature and humidity respectively. ; The calculation formula of the shopping idle value is: in, is the shopping idle value, For the current fixed time period The number of steps moved inside, The current time and fixed time period respectively The previous position, are the weights of moving steps and position changes respectively, ; The calculation formula of shopping situation coefficient is: in, is the shopping situation coefficient.

9. The method for generating advertisements based on user behavior drive according to claim 1, characterized in that: The advertisement generation model is based on a generative adversarial neural network, including a generator and a judger: The generator maps the input to the output of the advertising content through a neural network, and the calculation formula is: in, For the content of the advertisement, is the real-time behavioral factor feature set, are the parameters of the generator, is the mapping function of the generator; The goal of the discriminator is to determine whether the ad content is generated by the generator, and based on this judgment it gives a probability value: in, Indicates the output advertising content. is the weight of the discriminator, is the bias of the discriminator, is the sigmoid function.

10. The method for generating advertisements based on user behavior drive according to claim 9, characterized in that: The mapping function of the generator adopts a recurrent neural network: in, is the hidden state vector at the current moment, is the input value at the current moment, is the weight matrix of the hidden layer, is the weight matrix of the input layer, is the bias term of the hidden layer, is the hyperbolic tangent function, is the current output, is the output weight matrix, is the output bias term; Among them, the mathematical expressions of the sigmoid function and the hyperbolic tangent function are: in, is the sigmoid function, is the input of the sigmoid function, is the hyperbolic tangent function, is the input of the hyperbolic tangent function.

Images