Sales Forecasting and Consumption Recommendation System of E-commerce Agricultural Products Based on LSTM Model

Abstract

With the rapid development of global e-commerce, the sales volume of agricultural products, as an important consumer product category, in e-commerce platforms is increasing. However, affected by seasonal changes and market demand fluctuations, the sales forecast of agricultural products has always been a key challenge faced by e-commerce platforms. With the increase of personalized consumption demand, how to use recommendation system to improve users’ shopping satisfaction has become an important issue in the competition of e-commerce platforms. This study proposes a Long Short-Term Memory Network (LSTM) model based on e-commerce agricultural product sales forecasting method, combined with a recommendation system to provide users with personalized consumption suggestions. By analyzing the agricultural product sales data of a large e-commerce platform, the LSTM model can effectively capture complex patterns in time series data and provide high-precision sales forecast. In addition, the recommendation system designed in this paper realizes personalized commodity recommendation by combining users’ historical behaviors with sales forecast results, and improves users’ click-through rate and purchase rate. The experimental results show that the sales forecasting system based on LSTM model shows high forecasting accuracy in many agricultural products categories, and the user click-through rate after combining with the recommendation system increases by 1.3 percentage points, and the recommendation accuracy rate reaches 68.2%.