Digital intelligent marketing method and system applying AI large model

Abstract

The invention discloses a digital intelligent marketing method and system applying an AI large model, and relates to the technical field of digital intelligent marketing, and the method comprises the following steps: multi-source data collection, user activeness analysis, potential analysis, marketing comprehensive evaluation, user rating through a marketing strength algorithm in combination with the data of an activeness analysis algorithm and a potential analysis algorithm, and marketing evaluation. According to the method, users are divided into low-marketing users, medium-marketing users and high-marketing users, different marketing intensities are adopted for the users of different grades to optimize resource allocation, marketing content generation and pushing according to different marketing intensities, and the method has the advantages that the activity degree and purchase potential of the users are quantitatively evaluated through an activity analysis algorithm and a potential analysis algorithm; in combination with a marketing strength algorithm, the users are graded according to a set threshold, high-grade users are ensured to obtain more marketing resources, reasonable resource allocation is realized, the product sales volume and the customer conversion rate are improved, and the marketing quality is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of digital intelligence marketing, and specifically provides a digital intelligence marketing method and system applying an AI large model. Background Art

[0002] Digital intelligence marketing is a new marketing method based on digital and intelligent technologies, combining advanced technical means such as artificial intelligence and big data, aiming to achieve the intelligence, precision, and efficiency of marketing activities. Its core is to drive marketing decisions through data, use in-depth analysis of user data to formulate more effective marketing strategies, improve brand awareness, drive website traffic, generate potential customers, and increase customer conversion rates. It includes not only traditional forms such as online advertising, email, and search engines, but also emerging forms such as content marketing, social media marketing, video marketing, mobile marketing, and influencer marketing.

[0003] Currently, merchants place digital intelligence marketing advertisements with a fixed cost on various online channel platforms, and the marketing intensity for each user on the platform is the same, lacking analysis of the actual data of individual users. As a result, users with greater potential fail to receive sufficient attention, or too many resources are invested in users with low purchase potential, and marketing resources cannot be reasonably allocated to each user, thereby reducing the marketing effect. Summary of the Invention

[0004] The purpose of the present invention is to provide a digital intelligence marketing method and system applying an AI large model, which solves the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: including the following steps: Multi-source data collection, collecting user data from different network platforms and storing it; User activity analysis, extracting information on access duration and access frequency in user data through data mining technology, and quantitatively evaluating the user activity through an activity analysis algorithm to obtain user activity characteristics; Potential analysis, extracting consumption situations and purchase intentions in user data through data mining technology, and quantitatively evaluating the user purchasing power through a potential analysis algorithm to obtain user purchasing power characteristics; Marketing comprehensive evaluation, rating users through a marketing intensity algorithm, combining user activity characteristics and user purchasing power characteristics to obtain a single user grade score, setting different thresholds for the single user grade score, classifying users into low-marketing users, medium-marketing users, and high-marketing users according to different thresholds, and adopting different marketing intensities for users of different levels to optimize resource allocation; Marketing content generation analyzes user data through natural language processing (NLP) to extract users' interest characteristics, then uses the GPT model to generate personalized marketing content based on the interest characteristics, and pushes it according to different marketing intensities.

[0006] Optionally, the activity analysis algorithm extracts single-user access duration information, total-user average access duration information, single-user access frequency information, and total-user average access frequency information from user data. It conducts a comparative analysis of the single-user access duration information and the total-user average access duration information to obtain the access duration characteristics, and conducts a comparative analysis of the single-user access frequency information and the total-user average access frequency information to obtain the access frequency characteristics. The user activity characteristics are obtained by weighted fusion of the access duration characteristics and the access frequency characteristics, and the user activity characteristics represent the activity level of users on the Internet.

[0007] Optionally, the potential analysis algorithm extracts single-user consumption amount information and total-user average consumption amount information, single-user consumption frequency information and total-user average consumption frequency information from user data. It conducts a comparative analysis of the single-user consumption amount information and the total-user average consumption amount information to obtain the consumption amount characteristics, and conducts a comparative analysis of the single-user consumption frequency information and the total-user average consumption frequency information to obtain the consumption frequency characteristics. Then, by extracting page view information, shopping cart quantity information, and search volume information from user data, it conducts a weighted fusion of the page view information, the shopping cart quantity information, and the search volume information to obtain the purchase intention characteristics. The user purchasing power characteristics are obtained by weighted fusion of the consumption amount characteristics, the consumption frequency characteristics, and the purchase intention characteristics, and the user purchasing power characteristics represent the consumption level of users.

[0008] Optionally, the user activity characteristics include single-user activity information and user average activity information, and the user purchasing power characteristics include single-user purchasing power information and user average purchasing power information. The marketing intensity algorithm combines the single-user activity information and the single-user purchasing power information by weighting to obtain the single-user characteristics, then combines the user average activity information and the user average purchasing power information to obtain the user average characteristics, and divides the single-user characteristics by the user average characteristics to obtain the single-user rank score. A threshold is set for the single-user rank score for grading: High-marketing users: single-user rank score ≥ 1.5; Medium-marketing users: 1.5 > single-user rank score ≥ 1; Low-marketing users: single-user rank score < 1; The higher the user level, the more active the user is and the greater the purchase potential. More marketing resources should be allocated, and a higher marketing intensity should be invested. Conversely, the lower the user level, the lower the marketing intensity. Appropriate marketing resources should be invested according to different user levels to match the actual value of the users.

[0009] Optionally, when 1.5 > single user level score ≥ 1 and the user is determined as a medium-marketing user, adjust the weight of the purchase intention feature in the user purchasing power feature to encourage medium-marketing users to participate in more marketing activities, increase their purchase intention, and stimulate consumption to improve the conversion rate and sales volume.

[0010] Optionally, the network platform includes an e-commerce platform, a social platform, and an entertainment platform.

[0011] A digital marketing application using an AI large model includes a data collection module, a data analysis module, a marketing push module, and a data storage module; The data collection module is integrated with the API of the network platform and uses web crawler technology to capture public user data; The data analysis module includes an activity analysis unit, a purchasing power analysis unit, and a comprehensive evaluation unit. The activity analysis unit is used to analyze the user's login and access situation on the network platform. The purchasing power analysis unit is used to analyze the user's consumption level and purchase intention. The comprehensive evaluation unit comprehensively evaluates the user by combining the data of the activity analysis unit and the purchasing power analysis unit, and classifies the users into different levels according to the evaluation results; The marketing push module generates marketing content that meets the user's interests through NLP and GPT models and pushes it according to different user levels. The data storage module is responsible for managing and storing the user data collected from various network platforms.

[0012] Optionally, the data collection module includes a preprocessing unit, which is used to perform deduplication, formatting, and outlier detection on the collected user data.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention analyzes according to the access duration and access times of users on the network platform through an activity analysis algorithm, and quantitatively evaluates the results to judge the activity degree of users. The higher the activity degree, the better the marketing effect. Then, the purchase potential of users is analyzed through a potential analysis algorithm. The potential analysis algorithm improves the accuracy of purchase potential evaluation by introducing information such as analyzing users' consumption behaviors, browsing behaviors, and data of products added to the shopping cart. Through multi-dimensional comprehensive analysis, it can more accurately predict which users are more likely to make purchases in the future, so as to optimize the allocation of marketing resources, and also quantitatively evaluate the analysis results for marketers to intuitively understand the activity and purchase potential of users.

[0014] 2. The present invention comprehensively evaluates users through a marketing intensity algorithm and combines the data of the activity analysis algorithm and the potential analysis algorithm, and sets a threshold for the results. Users are divided into low-marketing users, medium-marketing users, and high-marketing user levels through the threshold. The higher the user level, the more active the user and the greater the purchase potential, and more marketing resources should be obtained and the marketing intensity invested should be higher. On the contrary, the lower the user level, the lower the marketing intensity invested. Appropriate marketing resources are invested according to different user levels to match the actual value of users, thereby realizing the reasonable allocation of marketing resources, increasing the sales volume of merchants' products and the customer conversion rate, and improving the marketing effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flowchart of the method of the present invention; Figure 2 is a block diagram of the system module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] Embodiment 1: Please refer to Figure 1 , this embodiment provides a digital marketing method applying an AI large model, including the following steps: Multi-source data collection, collecting user data from different network platforms and storing it; Specifically, the network platforms include e-commerce platforms, social platforms, and entertainment platforms; Data collection is limited to publicly available data, and strict compliance with relevant laws and regulations is ensured. All data collection operations require explicit consent from users. The user data collected will only be used for legitimate purposes, such as academic research or internal analysis, and will not involve commercial profit-making or unfair competition; User activity analysis: Quantitatively evaluate the user activity based on user data and through an activity analysis algorithm; Potential analysis: Use a potential analysis algorithm to quantitatively evaluate the purchasing power according to the consumption situation in user data and combined with the purchase intention factor; Comprehensive marketing evaluation: Rate users through a marketing intensity algorithm and combined with the data of the activity analysis algorithm and the potential analysis algorithm. Classify users into low-marketing users, medium-marketing users, and high-marketing users, and adopt different marketing intensities for users at different levels to optimize resource allocation; Marketing content generation: Analyze user data through natural language processing (NLP), extract the interest characteristics of users from it, then use the GPT model to generate personalized marketing content according to the interest characteristics, and push it according to different marketing intensities.

[0018] Among them, natural language processing (NLP) can analyze according to the user's browsing records, search records, purchase behaviors, etc. of goods, select the goods with the most keywords for subsequent processing. For example, analyze according to the corresponding tags of each good, and the tags include home appliances, clothing, beauty, etc. Select the one with the highest frequency as the user's interest characteristic, and then generate a user profile according to the user data and use it as the information input for the GPT model.

[0019] The specific content of the personalized marketing content is that even under the same marketing intensity, the marketing content is different. For example, the promoted product is a smart watch; User 1's user profile is health, sports, fitness; Marketing title: Help you with every challenge - Smart watch, a good partner for a healthy life Marketing content: The smart watch is designed for sports enthusiasts, with built-in professional sports modes to accurately monitor every piece of sports data such as your running, cycling, swimming, etc. It can not only track your heart rate and calorie consumption in real time, but also help you formulate a scientific training plan, making fitness more goal-oriented and efficient! Whether you are challenging the limit or having daily workouts, the smart watch can help you better manage your health.

[0020] User 2's user profile is learning, entertainment, cost-effectiveness; Marketing title: A must-have for students - Smart watch, combining learning and entertainment without compromise; The marketing content is: The smartwatch can not only help you manage your time, but also help you maintain a balance between study and entertainment! It can give you reminders during breaks, manage your study plans, and handle your schedule with one click. At the same time, it also supports various entertainment functions, allowing you to check social media, listen to music, answer calls at any time, and easily meet the multiple needs of study and life. With extremely high cost performance, you can easily enjoy the convenience brought by technology.

[0021] Thus, the push of personalized marketing content for each user is realized.

[0022] More specifically, in this embodiment: User data is collected from different network platforms. After the user data is collected, first, the activity analysis algorithm is used to analyze the activity level of the user on the network platform, and the results are quantitatively evaluated. The higher the activity level of the user on the network platform, the greater the chance the user will see when advertising is placed, and the better the marketing effect. Then, the potential analysis algorithm is used to analyze the purchase potential of the user, and the analysis results are also quantitatively evaluated, which is convenient for marketers to intuitively understand the activity and purchase potential of the user. And by introducing the purchase intention factor into the potential analysis algorithm, the accuracy of the potential analysis algorithm is improved. Finally, through the marketing intensity algorithm, combined with the data of the activity analysis algorithm and the potential analysis algorithm, a comprehensive evaluation of the user is carried out, and a threshold is set for the results. The user is divided into levels through the threshold. The higher the user level, the more active and greater the purchase potential the user is, and more marketing resources should be obtained, and the higher the marketing intensity should be invested. On the contrary, the lower the user level, the lower the marketing intensity should be invested. Thus, appropriate marketing resources are accurately allocated to different users, avoiding resource waste, and improving the digital marketing effect through in-depth analysis of user data.

[0023] Furthermore, the process of the activity analysis algorithm is as follows: ; where US i is the activity score of user i; T st,i is the access duration of user i within a unit time; T ave is the average access duration within a unit time, representing the average access duration of all users within a unit time; By comparing the access duration T st,i of user i within a unit time with the average access duration T ave within a unit time, the position of the access duration of user i among all users can be evaluated; W 1 is the access duration influence coefficient, and its value range is from 0 to 1; F i is the number of accesses of user i within a unit time; Fave is the average number of visits per unit time; Similarly, the number of visits F of user i per unit time i is compared with the average number of visits F per unit time ave to evaluate the position of user i's number of visits among all users; W 2 is the influence coefficient of the number of visits, and its value range is from 0 to 1; The average visit duration T per unit time ave and the average number of visits F per unit time ave are both calculated based on the data of all platforms; Specifically, the activity score US of user i i represents the activity level of user i on the platform. The larger the activity score US of user i i the more frequently the user uses the platform, and the easier it is to be seen during advertising placement on the platform, and the better the advertising placement. The smaller the activity score US of the user, the less frequently the user uses the platform, and the worse the advertising placement. According to the activity level of the user on the network platform, the marketing efforts can be adjusted, and the activity analysis algorithm introduces data from different dimensions to comprehensively analyze the user activity, making the analysis results more accurate.

[0024] Furthermore, the potential analysis algorithm process is as follows: ; where UQ i is the purchasing power score of user i; M i is the consumption amount of user i per unit time, representing the total consumption within a unit time; M ave is the average consumption amount per unit time; W 3 is the influence coefficient of the consumption amount, and its value range is from 0 to 1; BS i is the purchase intention score of user i, and its value range is from 0 to 1; W 4 is the influence coefficient of the purchase intention, and its value range is from 0 to 1; PL i is the consumption frequency of user i per unit time; PL iave is the average consumption frequency of user i per unit time; W 5 is the influence coefficient of the consumption frequency, and its value range is from 0 to 1; The data of consumption amount, purchase intention score, and consumption frequency are all related to purchasing power. Therefore, they are weighted and added together to reflect the purchasing power level of users as a whole; Specifically, the purchasing power score UQ of user i i represents the consumption level of the user. The larger the purchasing power score UQ of user i i , the higher the consumption level of the user and the better the benefits brought to the enterprise. Conversely, it indicates a lower consumption level and worse benefits brought to the enterprise; The purchase intention score BS of user i i is obtained from the normalized user operation behavior data, and the process is as follows: ; where P i is the page view score of user i, and its value range is from 0 to 1, representing the number of products viewed on the platform; M 1 is the page view times impact coefficient; A i is the score of the number of products added to the shopping cart by user i, and its value range is from 0 to 1; M 2 is the page view times impact coefficient; G i is the search volume score of user i, and its value range is from 0 to 1; M 3 is the search volume impact coefficient, M 1 +M 2 +M 3 =1; Among them, the user operation behavior data includes the page view score P of user i i , the score A of the number of products added to the shopping cart by user i I and the search volume score G of user i i . The maximum-minimum normalization is used to process the user operation behavior data, and the process is as follows: ; where GX is the normalized data; X is the original data, that is, the specific user operation behavior data; X min is the minimum value of this item of data, that is, the minimum value of this item of data among all users; X max is the maximum value of this item of data, that is, the maximum value of this item of data among all users; By converting the data of different dimensions and dimensions in the user operation behavior data into a unified range, it is possible to avoid the influence of a certain type of data on the entire scoring system due to too large or too small values, which is convenient for analysis and comparison.

[0025] Page view score P of user i i 、Score A for the number of items added to the shopping cart by user i i and search volume score G of user i i All reflect the degree of interest of user i in the product. The higher the score, the greater the probability of purchasing the product. When users purchase products, they usually search and browse relevant products for comparison, and add suitable ones to the shopping cart for final comparison or directly purchase them all at once. Therefore, adding them with weights can represent the purchase intention score BS of user i i .

[0026] Specifically, the purchase intention score BS of user i i represents the strength of the user's purchase intention. The greater the purchase intention score BS of user i i , the stronger the user's interest in the product and the higher the probability of purchasing the product. On the contrary, the weaker the interest in the product and the lower the probability of purchasing the product. The potential analysis algorithm improves the accuracy of purchase potential assessment by introducing information such as the user's consumption behavior, browsing behavior, and data of items added to the shopping cart. Through this multi-dimensional comprehensive analysis, it can more accurately predict which users are more likely to make purchases in the future, thus optimizing the allocation of marketing resources

[0027] Furthermore, the marketing intensity algorithm process is as follows: ; E i Is the level score of user i; US i Is the activity score of user i; β is the activity score influence coefficient, and its value range is from 0 to 1; UQ i Is the purchasing power score of user i; γ is the activity score influence coefficient, and its value range is from 0 to 1; US ave Is the average activity score; UQ ave Is the average activity score; Activity score US of user i i And the purchasing power score UQ of user i i And as two independent indicators, adding them with weights can more reasonably reflect their contribution to the level score E of user i i ;

[0028] Specifically, set the threshold one Y1 of the level score E of user i i To be 1, and the threshold two Y2 to be 1.5. When the level score E of user i iWhen it is less than the first threshold Y1, it is determined that user i is a low-marketing user. When the first threshold Y1 ≤ the level score E of user i i When it is less than the second threshold Y2, it is determined as a medium-marketing user. When the level score E of user i i ≥ the second threshold Y2, it is determined as a high-marketing user. The higher the user level, the more active the user is and the greater the purchase potential. More marketing resources should be obtained, and the marketing intensity should be higher. On the contrary, the lower the user level, the lower the marketing intensity. According to the different user levels, corresponding marketing resources are invested to match the actual value of the user, thus realizing the reasonable allocation of marketing resources, improving the sales volume of the merchant's products and the customer conversion rate, and increasing the enterprise's benefits.

[0029] The first threshold Y1 and the second threshold Y2 can be set according to the percentile method, that is, the level scores of all users are sorted from low to high, and a certain percentage of users are selected as high-marketing users, medium-marketing users, and low-marketing users. For example, if you want to select the top 20% of all users as high-marketing users, you can take the score of the last user in the top 20% as the second threshold Y2. When the level score E of user i i is greater than the second threshold Y2, it is a high-marketing user, and so on for medium-marketing users and low-marketing users.

[0030] For high-marketing users, high-frequency push is carried out, with 6 to 8 pushes per day, the discount intensity is increased, the goods are sold at a 20% discount, and at the same time, long-term full-reduction qualifications are given, and coupons are returned according to the consumption amount; For medium-marketing users, medium-frequency push is carried out, with 3 to 5 pushes per day, the goods are sold at a 10% discount, and limited-time full-reduction qualifications are given; For low-marketing users, low-frequency push is carried out, with 1 to 2 pushes per day, and a discount for the first purchase.

[0031] Furthermore, when the first threshold Y1 < the level score E of user i i < the second threshold Y2 and it is determined as a medium-marketing user, the purchase intention influence coefficient W 4 is increased, and the process is as follows: ; where NW 4 is the newly added purchase intention influence coefficient; k is the adjustment coefficient; Specifically, when the user is determined as a medium-marketing user, the newly added purchase intention influence coefficient NW 4 is used to replace the purchase intention influence coefficient W 4 , so as to improve the purchase intention score BS i of user i, and finally improve the purchasing power score UQ i and the level score E of user i i, since medium-marketing users usually have a certain purchasing potential and activity level, their purchase intention scores generally fall between those of high-potential users and low-potential users. They may be hesitant between buying and not buying. At this time, more attention needs to be paid to this group of users. By investing more marketing resources in these users, encouraging them to participate in more marketing activities, increasing their purchase intention, and stimulating consumption, the conversion rate and sales volume can be improved.

[0032] An increase in the purchase intention score does not necessarily result in an increase in the level score sufficient to enter the score range of high-marketing users. However, there may be medium-marketing users who were originally close enough to the high-marketing user score to achieve an increase. And in the future, when the activity level, purchasing power, etc. of medium-marketing users further improve, the original medium-marketing users are more likely to be converted into high-marketing users. Therefore, even if the increase in the purchase intention score cannot temporarily cross over to the score range of high-marketing users, through long-term optimization, more medium-marketing users can still be promoted to high-marketing users.

[0033] Embodiment 2: Based on the above embodiment, please refer to Figure 2 , the present invention provides a system, a digital marketing system applying an AI large model, including a data collection module, a data analysis module, a marketing push module, and a data storage module; The data collection module is integrated with the API of the network platform and uses web crawler technology to capture public user data. The data collection module includes a preprocessing unit, which is used to perform deduplication, formatting, and outlier detection processing on the collected user data; The data analysis module includes an activity analysis unit, a purchasing power analysis unit, and a comprehensive evaluation unit. The activity analysis unit is used to analyze the user's login and access situation on the network platform. The purchasing power analysis unit is used to analyze the user's consumption level and purchase intention. The comprehensive evaluation unit comprehensively evaluates the user by combining the data of the activity analysis unit and the purchasing power analysis unit, and classifies the users into different levels according to the evaluation results; The marketing push module generates marketing content that meets the user's interests through NLP and GPT models and pushes it according to different user levels. The data storage module is responsible for managing and storing the user data collected from various network platforms.

[0034] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A digital marketing method using an AI big model, characterized in that: The following steps are involved: Step S1: Multi-source data collection, collecting user data from different network platforms and storing them; Step S2: User activity analysis: extracting access duration and access frequency information from user data through data mining technology, and quantitatively evaluating user activity through activity analysis algorithm to obtain user activity characteristics; Step S3: Potential analysis, extracting consumption and purchase intention from user data through data mining technology, and quantitatively evaluating user purchasing power through potential analysis algorithm to obtain user purchasing power characteristics; Step S4: Comprehensive marketing evaluation: users are rated by the marketing intensity algorithm in combination with user activity characteristics and user purchasing power characteristics to obtain a single user grade. Different thresholds are set for the single user grade. Users are divided into low marketing users, medium marketing users and high marketing users according to different thresholds. Different marketing intensities are adopted for users of different grades to optimize resource allocation. Step S5: Marketing content generation: Analyze user data through natural language processing (NLP) to extract user interest features, and then use the GPT model to generate personalized marketing content based on the interest features, and push it according to different marketing intensities.

2. According to claim 1, a digital marketing method using an AI big model is characterized by: The activity analysis algorithm extracts single user access time information, total user average access time information, single user access times information and total user average access times information from user data, compares and analyzes single user access time information and total user average access time information to obtain access time features, compares and analyzes single user access times information and total user average access times information to obtain access times features, the user activity features are obtained by weighted fusion of the access time features and the access times features, and the user activity features represent the user's activity level on the Internet.

3. According to claim 2, a digital marketing method using an AI big model is characterized by: The potential analysis algorithm extracts the single user consumption amount information and the total user average consumption amount information, the single user consumption frequency information and the total user average consumption frequency information from the user data, compares and analyzes the single user consumption amount information and the total user average consumption amount information to obtain the consumption amount characteristics, compares and analyzes the single user consumption frequency information and the total user average consumption frequency information to obtain the consumption frequency characteristics, and then extracts the page view information, the shopping cart quantity information and the search volume information from the user data, and weightedly fuses the page view information, the shopping cart quantity information and the search volume information to obtain the purchase intention characteristics. The user purchasing power characteristics are obtained by weightedly fusing the consumption amount characteristics, the consumption frequency characteristics and the purchase intention characteristics, and the user purchasing power characteristics represent the user's consumption level.

4. According to claim 3, a digital marketing method using an AI big model is characterized by: The user activity feature includes single user activity information and user average activity information, the user purchasing power feature includes single user purchasing power information and user average purchasing power information, the marketing strength algorithm weights and combines single user activity information and single user purchasing power information to obtain single user features, then combines user average activity information and user average purchasing power information to obtain user average features, divides single user features by user average features to obtain single user rating points, sets thresholds for single user rating points for grading: High marketing users: single user rating score ≥ 1.5; Medium marketing users: 1.5>single user rating score ≥1; Low marketing users: single user rating score < 1; The higher the user level, the more active the user is and the greater the purchasing potential. Therefore, the user should receive more marketing resources and the higher the marketing investment should be. Conversely, the lower the user level, the lower the marketing investment should be. Appropriate marketing resources should be invested according to the user level to match the user's actual value.

5. According to claim 4, a digital marketing method using an AI big model is characterized by: When 1.5>single user rating score≥1, the user is determined to be a mid-marketing user. The weight of the purchase intention feature in the user purchasing power feature is adjusted to encourage the mid-marketing user to participate in more marketing activities, increase purchase intention, stimulate consumption, and improve conversion rate and sales.

6. According to claim 1, a digital marketing method using an AI big model is characterized by: The network platforms include e-commerce platforms, social platforms and entertainment platforms.

7. A digital marketing system for executing the digital marketing method using an AI big model as claimed in claim 1, characterized in that: It includes data collection module, data analysis module, marketing push module and data storage module; The data collection module is integrated with the API of the network platform and uses web crawler technology to capture public user data; The data analysis module includes an activity analysis unit, a purchasing power analysis unit and a comprehensive evaluation unit. The activity analysis unit is used to analyze the user's login access situation on the network platform, the purchasing power analysis unit is used to analyze the user's consumption level and purchasing intention, and the comprehensive evaluation unit comprehensively evaluates the user by combining the data of the activity analysis unit and the purchasing power analysis unit, and divides the user into different levels according to the evaluation results; The marketing push module generates marketing content that meets the user's interests through NLP and GPT models, and pushes it according to different user levels. The data storage module is responsible for managing and storing user data collected from various network platforms.

8. According to claim 7, a digital marketing system using an AI big model is characterized by: The data collection module includes a preprocessing unit, which is used to perform deduplication, formatting, and outlier detection processing on the collected user data.

Images