Agricultural product intelligent sales management system based on AI digital human

Abstract

The invention relates to the technical field of agricultural product sales management, in particular to an agricultural product intelligent sales management system based on AI digital humans, and aims to solve the problems that in the prior art, sales volume, price and environmental factors cannot be comprehensively analyzed to construct a sales trend prediction model, a comprehensive view cannot be provided to improve prediction accuracy, and the sales trend cannot be predicted. And enterprises cannot quickly respond to market dynamics. According to the method, the sales trend prediction model is constructed by using historical sales and meteorological data, the method has significant advantages, the sales volume, price and environmental factors are comprehensively analyzed, a comprehensive perspective is provided to improve prediction accuracy, the sales trend prediction feature vector is generated, the module can predict future sales volume change in real time, and the prediction accuracy is improved. And enterprises are supported to quickly respond to market dynamics, and inventory, pricing and marketing strategies are optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural product sales management, and more specifically, to an intelligent sales management system for agricultural products based on an AI digital human. Background Art

[0002] With the rapid development of the agricultural industry and the continuous progress of Internet technology, the sales methods of agricultural products are also constantly innovating. Traditional sales methods of agricultural products often have problems such as information asymmetry, single sales channels, and backward marketing means, making it difficult to meet the diverse needs of modern consumers. At the same time, with the increasing maturity of AI technology, the application of AI digital humans in various fields is becoming more and more extensive, providing new ideas and solutions for the intelligent sales management of agricultural products.

[0003] The patent application with the publication number CN115660767A discloses an agricultural product sales management system, which includes a front-end function module and a back-end function module. The front-end function module includes a registration module, a personal information management module, an order management module, a shopping cart management module, and an online message module. The beneficial effects are as follows: By adding a dual-code payment verification module at the user end and a notice penalty management module and an overdue delivery claim module at the management end, on the one hand, the product sales management system can effectively improve the security of the payment environment at the user end and avoid the risk of information leakage during the payment process. On the other hand, during the use of the product sales management system, it can purify and warn the system operation environment through notice penalty prompts, and at the same time, combined with overdue delivery claims, make merchants compensate users to a certain extent, effectively improving the satisfaction of users when using the system, thus facilitating the efficient and orderly management and operation of the system. However, the above reference patent, by adding a dual-code payment verification at the user end, a notice penalty management and an overdue delivery claim module at the management end, improves the system payment security, purifies the system operation environment and improves user satisfaction, but cannot extract multi-dimensional data features and use machine learning models to accurately classify users, reducing the customer's understanding and response ability, unable to achieve personalized services and marketing, unable to comprehensively analyze sales volume, price and environmental factors to build a sales trend prediction model, unable to provide a comprehensive perspective to improve prediction accuracy, unable to support enterprises to quickly respond to market dynamics, unable to provide accurate marketing strategies and optimize the sales of agricultural products, and unable to adjust strategies in a timely manner according to sales trends to enhance market response ability.

[0004] Therefore, we propose an intelligent sales management system for agricultural products based on an AI digital human to address the above problems. Summary of the Invention

[0005] The object of the present invention is to provide an intelligent sales management system for agricultural products based on an AI digital human, which solves the problems that the prior art cannot extract multi-dimensional data features and use machine learning models to accurately classify users, reduces the customer's understanding and response ability, cannot achieve personalized services and marketing, cannot comprehensively analyze sales volume, price and environmental factors to build a sales trend prediction model, cannot provide a comprehensive perspective to improve prediction accuracy, cannot support enterprises to quickly respond to market dynamics, cannot provide accurate marketing strategies and optimize the sales of agricultural products, and cannot adjust strategies in a timely manner according to sales trends to enhance market response ability.

[0006] The object of the present invention is achieved by the following technical solutions: An intelligent sales management system for agricultural products based on an AI digital human, which is applied to a sales management platform, includes: A data collection and preprocessing module, which is used to collect the operation and management data and user data of target agricultural products, and perform preprocessing operations on the collected operation and management data and user data; A user portrait construction module, which is used to extract user portrait feature vectors from the preprocessed user data, construct a machine learning model for identifying user portrait types, identify the user portrait types through the model, and describe the user portrait types; A sales data analysis and prediction module, which is used to collect the historical operation and management data of target agricultural products, build a sales trend prediction model, and predict the future sales trend of target agricultural products through the model; A sales strategy optimization module, which formulates and optimizes agricultural product sales strategies based on user portrait types and sales trend prediction results; An AI digital human recommendation module, which provides personalized agricultural product recommendations for users based on user portrait types and agricultural product sales strategies.

[0007] As a preferred embodiment of the present invention, the specific process of the user portrait construction module extracting user portrait feature vectors from the preprocessed operation and management data is as follows: Obtain the preprocessed historical user data. The user data includes user age, user gender, purchase amount, purchase times, and the duration since the last purchase. Generate a collection period, divide the collection period into m consecutive sub-periods equally, and mark the midpoint moment of each sub-period to obtain m midpoint moments; Taking the m midpoint moments as the base points, expand forward and backward by equal time lengths respectively, mark s - 1 expansion moments, and after summarizing the midpoint moments and s - 1 expansion moments, obtain s collection moments; Obtain the purchase amounts at the s collection moments, obtain s collection purchase amount values, and after accumulating and averaging the s collection purchase amount values, obtain the purchase amounts for m sub-periods; The expression for the sub - cycle purchase amount is: ; In the formula, is the sub - cycle purchase amount of the m - th sub - cycle, is the n - th acquisition purchase amount value of the m - th sub - cycle; Remove the maximum and minimum values of the sub - cycle purchase amounts, accumulate the remaining m - 2 sub - cycle purchase amounts and then calculate the average to obtain the average purchase amount; The expression for the average purchase amount is: ; In the formula, is the average purchase amount, is the sub - cycle purchase amount of the p - th sub - cycle; Using the method to calculate the average purchase amount similarly, the average purchase frequency and the average duration since the most recent purchase can be obtained.

[0008] As a preferred embodiment of the present invention, the specific process for the user portrait construction module to construct a machine learning model for identifying user portrait types is as follows: Obtain the age of each user, represent the age of each user with corresponding numbers, obtain the gender of each user, number the gender of each user, number males as 01, number females as 02, and combine the age, gender, average purchase amount , average purchase frequency and the average duration since the most recent purchase extracted from each user to form a user portrait feature vector YHT; Take the extracted user portrait feature vector YHT as the input of the machine learning model, and take the label vector v as the output of the machine learning model. The label vector v represents the user portrait type, and the output of the label vector v is 0, 1, 2, or 3. 0 represents that the user portrait type is a churned user, 1 represents that the user portrait type is a potential user, 2 represents that the user portrait type is a price - sensitive user, and 3 represents that the user portrait type is a high - value user. Using the label vector v as the prediction target and minimizing the sum of prediction errors of all training data as the training target, train the machine learning model until the sum of prediction errors reaches convergence and then stop training to obtain a machine learning model that predicts the label vector v.

[0009] As a preferred embodiment of the present invention, the specific process for the user portrait construction module to identify and describe the user portrait type is as follows: Obtain the preprocessed real-time user data, process it, and construct the real-time user portrait feature vector YHT. Identify the user portrait type through a trained machine learning model; If the label vector output by the machine learning model is 0, it indicates that the user portrait type is a churned user, and the user portrait type description is that they haven't made a purchase for a long time, the last purchase time was a long time ago, and the purchase frequency is very low; If the label vector output by the machine learning model is 1, it indicates that the user portrait type is a potential user, and the user portrait type description is that the last purchase time was relatively long or they have never purchased, but they show interest in the target agricultural products; If the label vector output by the machine learning model is 2, it indicates that the user portrait type is a price-sensitive user, and the user portrait type description is that the purchase frequency is medium, the amount of each purchase is low, and they are very sensitive to price fluctuations; If the label vector output by the machine learning model is 3, it indicates that the user portrait type is a high-value user, and the user portrait type description is that they make high-frequency and high-amount purchases, and the last purchase time was relatively recent.

[0010] As a preferred implementation manner of the present invention, the specific process for the sales data analysis and prediction module to construct the sales trend prediction model is as follows: Obtain the historical operation and management data of the target agricultural product. The operation and management data includes sales data and meteorological data. The sales data includes sales volume, sales price, and cost price. The meteorological data includes ambient temperature, precipitation, and sunshine duration. Generate a collection period and equally divide the collection period into multiple collection time segments; Obtain the sales volume change rate of the target agricultural product within multiple collection time segments. The sales volume change rate represents the ratio of the sales volume change amount to the duration of the corresponding time period. Construct a set A of the sales volume change rate in this way, and record the mean value of the difference between the largest subset and the smallest subset in set A as the sales volume change rate difference XLC; Using the method of calculating the sales volume change rate difference XLC, in the same way, the sales price change rate difference XJC, the cost price change rate difference CJC, the ambient temperature change rate difference HWC, the precipitation change rate difference JLC, and the sunshine duration change rate difference RSC can be obtained.

[0011] As a preferred implementation manner of the present invention, combine the sales volume change rate difference XLC, the sales price change rate difference XJC, the cost price change rate difference CJC, the ambient temperature change rate difference HWC, the precipitation change rate difference JLC, and the sunshine duration change rate difference RSC to construct the sales trend prediction feature vector XYX; The constructed sales trend prediction feature vector XYX is used as the input of the machine learning model, and the sales volume change in a future period corresponding to each group of sales trend prediction feature vectors XYX is used as the output of the machine learning model. With the sales volume change in a future period as the prediction target and minimizing the sum of prediction errors of the training data as the training target, the machine learning model is trained until the sum of prediction errors converges and then the training stops, obtaining the sales trend prediction model.

[0012] As a preferred embodiment of the present invention, the specific process of the sales data analysis and prediction module for predicting the future sales trend of the target agricultural product through the model is as follows: The expression formula of the sales trend prediction model is as follows: ; Where represents the sales volume change in a future period, and are both regression coefficients, is the random error term; Obtain the real-time operation and management data of the target agricultural product, convert it into the corresponding sales trend prediction feature vector XYX and input it into the sales trend prediction model, and obtain the real-time sales volume change in a future period through the sales trend prediction model .

[0013] As a preferred embodiment of the present invention, the specific process of the sales strategy optimization module for formulating and optimizing the agricultural product sales strategy is as follows: Obtain the user portrait type and the sales trend prediction result. The user portrait types are high-value users, price-sensitive users, potential users, and lost users, and the sales trend prediction result is the sales volume change in a future period; Agricultural product sales strategies based on user portrait types: High-value users: Provide exclusive offers, membership point reward programs, priority experience of new products, and holiday greetings; Price-sensitive users: Carry out promotional activities, launch affordable product combinations, and emphasize product cost performance; Potential users: Provide trial products or free experiences, and produce attractive product promotion videos and graphics; Lost users: Send recovery emails or text messages, provide exclusive offers or compensation, and invite to participate in offline activities; Optimization of agricultural product sales strategies based on sales trend prediction results: When the sales trend prediction result is an increase in sales volume, continue to maintain the existing strategy, expand the production scale, increase inventory, and expand sales channels; When the sales trend prediction result indicates a decline in sales, corresponding countermeasures should be taken immediately for handling.

[0014] As a preferred embodiment of the present invention, the specific process of the AI digital human recommendation module providing personalized agricultural product recommendations to users is as follows: Obtain user portrait types and agricultural product sales strategies, and based on different user portraits and agricultural product sales strategies, provide personalized agricultural product recommendations for users of different portrait types: Products recommended for high-value users are: new products, high-quality products, scarce products, and organic products; Products recommended for price-sensitive users are: promotional products, cost-effective product combinations, special offer products, and discounted products; Products recommended for potential users are: trial products, experience products, and products with high sales volume and good reviews; Products recommended for lost users are: preferential products, compensatory products, and exclusive event products for old customers.

[0015] Compared with the prior art, the advantages of the present invention are as follows: (1) In the present invention, the user portrait construction module extracts multi-dimensional data features and uses machine learning models to accurately classify users, enhancing customer understanding and response capabilities. Based on age, gender, purchase frequency, and amount information, it realizes personalized services and marketing, improves user experience and loyalty. This module reduces manual analysis costs, improves decision-making efficiency, and has real-time response capabilities, enabling enterprises to quickly respond to market changes; (2) In the present invention, the sales data analysis and prediction module constructs a sales trend prediction model using historical sales and meteorological data, which has significant advantages. It comprehensively analyzes sales volume, price, and environmental factors, providing a comprehensive perspective to improve prediction accuracy. By generating sales trend prediction feature vectors, this module can instantly predict future sales changes, support enterprises to quickly respond to market dynamics, and optimize inventory, pricing, and marketing strategies; (3) In the present invention, the sales strategy optimization module provides accurate marketing strategies and optimizes agricultural product sales based on user portraits and sales trend predictions. It improves user satisfaction through personalized strategies and adjusts strategies in a timely manner according to sales trends to enhance market response capabilities. This module automates data analysis to improve decision-making efficiency and continuously optimizes to ensure the accuracy of strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is the system block diagram of Embodiment 1 in the present invention; Figure 2 It is the system block diagram of Embodiment 2 in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] The following will describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention; it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment 1: As Figure 1 shown, an AI digital human-based intelligent sales management system for agricultural products proposed by the present invention is applied to a sales management platform and includes: A data collection and preprocessing module for collecting operation management data and user data of target agricultural products and performing preprocessing operations on the collected operation management data and user data. The preprocessing operations include but are not limited to data cleaning, data conversion, and data integration; The data collection and preprocessing module improves data quality and consistency through cleaning, converting, and integrating data, provides a comprehensive data view, supports accurate analysis to predict trends, identify risks, and discover business opportunities. It simplifies data analysis, improves decision-making efficiency, and realizes personalized services through user data processing, enhancing customer satisfaction.

[0019] A user portrait construction module for extracting user portrait feature vectors from the preprocessed user data, constructing a machine learning model for identifying user portrait types, identifying the user portrait types through the model, and describing the user portrait types; The specific process of the user portrait construction module extracting user portrait feature vectors from the preprocessed operation management data is as follows: Obtain the preprocessed historical user data. The user data includes user age, user gender, purchase amount, purchase frequency, and the duration since the most recent purchase. Generate a collection period, and set the collection period duration to one year. Divide the collection period into m consecutive sub-periods equally, and mark the midpoint moment of each sub-period to obtain m midpoint moments; Taking the m midpoint moments as the base points, expand forward and backward by the same duration respectively, mark s - 1 expansion moments, and after summarizing the midpoint moments and s - 1 expansion moments, obtain s collection moments; Obtain the purchase amounts at the s collection moments to obtain s collection purchase amount values, and accumulate and average the s collection purchase amount values to obtain the purchase amounts for m sub-periods; The expression for the purchase amount of a sub-period is: ; In the formula, is the purchase amount of the mth sub-period, is the nth collection purchase amount value of the mth sub-period; Remove the maximum and minimum values of the sub-period purchase amounts, accumulate the remaining m - 2 sub-period purchase amounts and then calculate the average to obtain the average purchase amount; The expression for the average purchase amount is: ; In the formula, is the average purchase amount, is the sub-period purchase amount of the p-th sub-period; Using the method to calculate the average purchase amount similarly, the average purchase frequency and the average duration since the most recent purchase can be obtained; The specific process of the user profile construction module for constructing a machine learning model to identify user profile types is as follows: Obtain the age of each user, represent the age of each user with corresponding numbers, obtain the gender of each user, number the gender of each user, number male as 01, number female as 02, and combine the age, gender, average purchase amount , average purchase frequency and average duration since the most recent purchase of each user and extract them to form a user profile feature vector YHT; Use the extracted user profile feature vector YHT as the input of the machine learning model, and use the label vector v as the output of the machine learning model. The label vector v represents the user profile type, and the output of the label vector v is 0, 1, 2, or 3. 0 indicates that the user profile type is a churned user, 1 indicates that the user profile type is a potential user, 2 indicates that the user profile type is a price-sensitive user, and 3 indicates that the user profile type is a high-value user. Use the label vector v as the prediction target and minimize the sum of the prediction errors of all training data as the training target to train the machine learning model until the sum of the prediction errors converges and then stop training to obtain the machine learning model that predicts the label vector v; The specific process of the user profile construction module for identifying and describing the user profile type is as follows: Obtain the preprocessed real-time user data, process it to construct a real-time user profile feature vector YHT, and identify the user profile type through the trained machine learning model; If the label vector output by the machine learning model is 0, it indicates that the user profile type is a churned user, and the user profile type description is that the user has not made a purchase for a long time, the most recent purchase time was a long time ago, and the purchase frequency is very low; If the label vector output by the machine learning model is 1, it indicates that the user profile type is a potential user, and the user profile type description is that the most recent purchase time was relatively long ago or the user has never made a purchase, but shows interest in the target agricultural products; If the label vector output by the machine learning model is 2, it indicates that the user portrait type is price-sensitive users, and the description of the user portrait type is medium purchase frequency, low purchase amount per time, and very sensitive to price fluctuations; If the label vector output by the machine learning model is 3, it indicates that the user portrait type is high-value users, and the description of the user portrait type is high-frequency and high-amount purchases, and the time of the last purchase is relatively recent; The user portrait construction module extracts multi-dimensional data features and uses a machine learning model to accurately classify users, improving customer understanding and response capabilities. Based on age, gender, purchase frequency, and amount information, it realizes personalized services and marketing, improves user experience and loyalty. This module reduces manual analysis costs, improves decision-making efficiency, and has real-time response capabilities, enabling enterprises to quickly respond to market changes; The sales data analysis and prediction module is used to collect historical operation and management data of target agricultural products, construct a sales trend prediction model, and predict the future sales trend of target agricultural products through the model; The specific process of the sales data analysis and prediction module constructing the sales trend prediction model is as follows: Obtain the historical operation and management data of target agricultural products. The operation and management data include sales data and meteorological data. The sales data include sales volume, sales price, and cost price. The meteorological data include environmental temperature, precipitation, and sunshine duration. Generate a collection period and divide the collection period into multiple collection time periods; Obtain the sales volume change rate of target agricultural products in multiple collection time periods. The sales volume change rate represents the ratio of the sales volume change amount to the duration of the corresponding time period. Based on this, construct a set A of sales volume change rates, and record the mean value of the difference between the largest subset and the smallest subset in set A as the sales volume change rate difference XLC; Using the method of calculating the sales volume change rate difference XLC, the sales price change rate difference XJC, the cost price change rate difference CJC, the environmental temperature change rate difference HWC, the precipitation change rate difference JLC, and the sunshine duration change rate difference RSC can be obtained in the same way; Combine the sales volume change rate difference XLC, the sales price change rate difference XJC, the cost price change rate difference CJC, the environmental temperature change rate difference HWC, the precipitation change rate difference JLC, and the sunshine duration change rate difference RSC to construct a sales trend prediction feature vector XYX; The constructed sales trend prediction feature vector XYX is used as the input of the machine learning model, and the sales volume change in a future period corresponding to each group of sales trend prediction feature vectors XYX is used as the output of the machine learning model. With the sales volume change in a future period as the prediction target and minimizing the sum of prediction errors of the training data as the training target, the machine learning model is trained until the sum of prediction errors reaches convergence and then the training stops, obtaining a sales trend prediction model; The specific process of the sales data analysis and prediction module predicting the future sales trend of the target agricultural product through the model is as follows: The expression formula of the sales trend prediction model is as follows: ; Where represents the sales volume change in a future period, and are both regression coefficients, is the random error term; Obtain the real-time operation and management data of the target agricultural product, convert it into the corresponding sales trend prediction feature vector XYX and input it into the sales trend prediction model, and obtain the real-time sales volume change in a future period through the sales trend prediction model ; The sales data analysis and prediction module constructs a sales trend prediction model using historical sales and meteorological data, which has significant advantages. It comprehensively analyzes sales volume, price and environmental factors (such as temperature, precipitation), provides a comprehensive perspective to improve prediction accuracy. By generating sales trend prediction feature vectors, this module can instantly predict future sales volume changes, support enterprises to quickly respond to market dynamics, and optimize inventory, pricing and marketing strategies.

[0020] Embodiment 2: The technical solution of this embodiment of the present invention is different from that of Embodiment 1 in that: As Figure 2 shown, the sales strategy optimization module formulates and optimizes the agricultural product sales strategy based on the user portrait type and the sales trend prediction result; The specific process of the sales strategy optimization module formulating and optimizing the agricultural product sales strategy is as follows: Obtain the user portrait type and the sales trend prediction result. The user portrait type is high-value users, price-sensitive users, potential users and lost users, and the sales trend prediction result is the sales volume change in a future period; The agricultural product sales strategy based on the user portrait type: High-value users: Provide exclusive offers, membership point reward programs, priority experience of new products and festival greetings; Price-sensitive users: Conduct promotional activities (such as discounts, coupons, full reduction, etc.), launch affordable product combinations, and emphasize the cost performance of products; Potential users: Provide trial products or free experiences, and produce attractive product promotion videos and graphics; Lapsed users: Send recovery emails or text messages, provide exclusive offers or compensations, and invite them to participate in offline activities; Optimization of agricultural product sales strategies based on the results of sales trend prediction: When the sales trend prediction result is an increase in sales, continue to maintain the existing strategy, expand the production scale, increase inventory, and expand sales channels; When the sales trend prediction result is a decrease in sales, immediately take corresponding countermeasures for handling. The specific content of the countermeasures is: Adjust product prices, improve product quality, strengthen marketing promotion, develop new products, seek new sales channels, and optimize supply chain management; The sales strategy optimization module provides precise marketing strategies and optimizes agricultural product sales based on user portraits and sales trend prediction. It improves user satisfaction through personalized strategies (such as providing exclusive offers for high-value users and conducting promotions for price-sensitive users), and adjusts strategies in a timely manner according to sales trends to enhance market response capabilities. This module automates data analysis to improve decision-making efficiency and continuously optimizes to ensure the accuracy of strategies.

[0021] The AI digital human recommendation module provides personalized agricultural product recommendations for users based on user portrait types and agricultural product sales strategies; The specific process of the AI digital human recommendation module providing personalized agricultural product recommendations for users is as follows: Obtain user portrait types and agricultural product sales strategies, and provide personalized agricultural product recommendations for users of different portrait types based on different user portraits and agricultural product sales strategies: Products recommended for high-value users are: new products, high-quality products, scarce products, and organic products; Products recommended for price-sensitive users are: promotional products, cost-effective product combinations, special offer products, and discounted products; Products recommended for potential users are: trial products, experience products, and products with high sales volume and good reviews; Products recommended for lapsed users are: preferential products, compensatory products, and products exclusive to old customers' activities; The AI digital human recommendation module provides personalized recommendations based on user portraits and sales strategies, improves user experience and satisfaction, increases conversion rates and customer loyalty by precisely matching different user needs, adjusts recommendations in combination with real-time trends, uses big data to analyze user preferences to achieve precise recommendations, and reduces manual intervention through an automated system to improve marketing efficiency.

[0022] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its improved concept, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An intelligent sales management system for agricultural products based on AI digital human, applied to the sales management platform, characterized in that: include: A data collection and preprocessing module is used to collect the management data and user data of the target agricultural products and perform preprocessing operations on the collected management data and user data; The user portrait construction module is used to extract the user portrait feature vector from the preprocessed user data, build a machine learning model to identify the user portrait type, identify the user portrait type through the model, and describe the user portrait type; The sales data analysis and prediction module is used to collect the historical operation and management data of the target agricultural products, build a sales trend prediction model, and use the model to predict the future sales trend of the target agricultural products; Sales strategy optimization module, which formulates and optimizes agricultural product sales strategies based on user portrait types and sales trend forecast results; The AI ​​digital human recommendation module provides users with personalized agricultural product recommendations based on user portrait types and agricultural product sales strategies.

2. According to claim 1, the intelligent sales management system for agricultural products based on AI digital human is characterized in that: The specific process of extracting the user portrait feature vector from the pre-processed business management data by the user portrait construction module is as follows: Obtain preprocessed historical user data, including user age, gender, purchase amount, number of purchases, and time since the last purchase, generate a collection cycle, divide the collection cycle into m consecutive sub-cycles, and mark the midpoint of each sub-cycle to obtain m midpoints; Taking m midpoint moments as the base point, expand forward and backward by equal lengths, mark s-1 expansion moments, and then aggregate the midpoint moments and s-1 expansion moments to obtain s collection moments; Obtain the purchase amount at s collection moments, obtain s collected purchase amount values, and accumulate and average the s collected purchase amount values ​​to obtain the purchase amount of m sub-periods; The expression for the sub-period purchase amount is: ; In the formula, is the sub-period purchase amount of the mth sub-period, The nth collected purchase amount value for the mth sub-period; Remove the maximum and minimum purchase amounts of the sub-periods, and add up the remaining m-2 sub-period purchase amounts and average them to obtain the mean purchase amount; The expression for the mean purchase amount is: ; In the formula, is the mean purchase amount, is the sub-period purchase amount for the p-th sub-period; Use the mean purchase amount calculated The mean number of purchases can be obtained by the same method. and the average time from the last purchase .

3. According to claim 2, the intelligent sales management system for agricultural products based on AI digital human is characterized in that: The specific process of the user portrait construction module constructing a machine learning model for identifying user portrait types is as follows: Get the age of each user, represent each user's age with a corresponding number, get the gender of each user, number each user's gender, number male as 01, number female as 02, and average the age, gender, and purchase amount of each user. , average number of purchases And the average time since the last purchase Extract and combine them into a user portrait feature vector YHT; The extracted user portrait feature vector YHT is used as the input of the machine learning model, and the label vector v is used as the output of the machine learning model. The label vector v represents the user portrait type. The output of the label vector v is 0, 1, 2 or 3. 0 indicates that the user portrait type is a lost user, 1 indicates that the user portrait type is a potential user, 2 indicates that the user portrait type is a price-sensitive user, and 3 indicates that the user portrait type is a high-value user. The label vector v is used as the prediction target, and the training target is to minimize the sum of prediction errors of all training data. The machine learning model is trained until the sum of prediction errors converges and the training is stopped to obtain a machine learning model that predicts the label vector v.

4. According to claim 3, the intelligent sales management system for agricultural products based on AI digital human is characterized in that: The specific process of the user portrait construction module identifying the user portrait type and describing the user portrait type is as follows: Obtain preprocessed real-time user data, process it to construct a real-time user portrait feature vector YHT, and use the trained machine learning model to identify the user portrait type; If the label vector output by the machine learning model is 0, it means that the user profile type is a lost user. The user profile type is described as not having made a purchase for a long time, the last purchase was a long time ago, and the purchase frequency is very low; If the label vector output by the machine learning model is 1, it indicates that the user profile type is a potential user, and the user profile type is described as the last purchase was a long time ago or has never been purchased, but shows interest in the target agricultural products; If the label vector output by the machine learning model is 2, it means that the user profile type is a price-sensitive user. The user profile type is described as having a medium purchase frequency, a low purchase amount each time, and is very sensitive to price fluctuations; If the label vector output by the machine learning model is 3, it means that the user profile type is a high-value user. The user profile type is described as high-frequency and high-amount purchases, and the most recent purchase was made recently.

5. According to claim 1, the intelligent sales management system for agricultural products based on AI digital human is characterized in that: The specific process of the sales data analysis and prediction module collecting and processing the historical operation and management data of the target agricultural products is as follows: Obtain historical management data of target agricultural products, including sales data and meteorological data. Sales data includes sales volume, sales price and cost price. Meteorological data includes ambient temperature, precipitation and sunshine duration. Generate a collection cycle and divide the collection cycle into multiple collection periods. Obtain the sales volume change rate of the target agricultural product in multiple collection periods. The sales volume change rate represents the ratio between the sales volume change and the length of the corresponding time period. This is used to construct a set A of sales volume change rates, and the mean of the difference between the largest subset and the smallest subset in set A is recorded as the sales volume change rate difference XLC; By using the method of calculating the sales volume change rate difference XLC, we can also obtain the sales price change rate difference XJC, the cost price change rate difference CJC, the ambient temperature change rate difference HWC, the precipitation change rate difference JLC and the sunshine duration change rate difference RSC.

6. The agricultural product intelligent sales management system based on AI digital human according to claim 5 is characterized in that: The specific process of the sales data analysis and prediction module to construct the sales trend prediction model is as follows: The sales volume change rate difference XLC, sales price change rate difference XJC, cost price change rate difference CJC, ambient temperature change rate difference HWC, precipitation change rate difference JLC and sunshine duration change rate difference RSC are combined to construct the sales trend prediction feature vector XYX; The constructed sales trend prediction feature vector XYX is used as the input of the machine learning model, and the sales change in the future period corresponding to each group of sales trend prediction feature vectors XYX is used as the output of the machine learning model. The sales change in the future period is taken as the prediction target, and minimizing the sum of prediction errors of training data is taken as the training target. The machine learning model is trained until the sum of prediction errors converges and the training is stopped to obtain a sales trend prediction model.

7. The agricultural product intelligent sales management system based on AI digital human according to claim 6 is characterized in that: The specific process of the sales data analysis and prediction module predicting the future sales trend of the target agricultural product through the model is as follows: The sales trend forecasting model is expressed as follows: ； in Indicates the change in sales volume over a period of time in the future. as well as are regression coefficients, is the random error term; Obtain the real-time management data of the target agricultural products, convert it into the corresponding sales trend prediction feature vector XYX and input it into the sales trend prediction model. The sales trend prediction model can be used to obtain the real-time sales change in the future. .

8. The agricultural product intelligent sales management system based on AI digital human according to claim 1 is characterized in that: The specific process of the sales strategy optimization module to formulate and optimize the agricultural product sales strategy is as follows: Obtain user portrait types and sales trend forecast results. User portrait types include high-value users, price-sensitive users, potential users, and lost users. Sales trend forecast results are sales volume changes over a period of time in the future. Agricultural product sales strategies based on user portrait types: High-value users: Provide exclusive discounts, membership points reward programs, priority experience of new products, and holiday greetings; Price-sensitive users: Carry out promotional activities, launch economical product combinations, and emphasize product cost-effectiveness; Potential users: Provide trial products or free experiences, and create attractive product promotion videos and graphics; Lost users: send recovery emails or text messages, provide exclusive discounts or compensation, and invite them to participate in offline activities; Optimization of agricultural product sales strategies based on sales trend forecast results: When the sales trend forecast results in an increase in sales, continue to maintain the existing strategy and expand production scale, increase inventory, and expand sales channels; When the sales trend forecast results in a decline in sales, take appropriate countermeasures immediately to deal with it.

9. The agricultural product intelligent sales management system based on AI digital human according to claim 1 is characterized in that: The specific process of the AI ​​digital human recommendation module providing personalized agricultural product recommendations to users is as follows: Obtain user portrait types and agricultural product sales strategies, and make personalized agricultural product recommendations for users with different portrait types based on different user portraits and agricultural product sales strategies: Products recommended by high-value users are: new products, high-quality products, scarce products, and organic products; The recommended products for price-sensitive users are: promotional products, cost-effective product combinations, special-priced products, and discounted products; Products recommended to potential users are: trial products, experience products, and products with high sales and good reviews; Recommended products for lost users are: preferential products, compensatory products, and exclusive products for old customers.

Images