Sales Forecast Model Of Fresh Agricultural Products Considering Weather Factors

Under the new retail boom,the fresh product market in China has broad prospects.However,the perishable nature of fresh agricultural products makes them require high timeliness,and it is easy to produce imbalances in supply and marketing,which causes fresh supermarkets to face risks such as high inventory costs and high damage rates.Accurate sales forecasting has become an effective way to solve the imbalance between supply and marketing.At present,most fresh supermarkets' sales forecasts are limited to traditional analysis methods,and some companies even rely entirely on the personal experience of managers to make decisions.At the same time,various hidden variable factors will also have a direct or indirect impact on sales,making it difficult to make accurate predictions.Therefore,mining the hidden variable factors that affect the sales volume of fresh produce and constructing an accurate sales forecast model have become the key issues that need to be resolved in the supply chain of fresh produce.This paper carries out the following research on the above issues:(1)The value of weather factors in the sales forecast model of fresh agricultural product.In view of the sales forecast of fresh agricultural products,the independent variables considered by the current forecasting models are limited to historical sales,weekends,holidays.However,the impact of weather factors(weather conditions,wind level,air quality index)in the sales forecast of fresh produce has not been rigorously analyzed.Based on daily sales of a fresh supermarket in Hangzhou,we fitted the baseline prediction model and weather prediction model based on three machine learning algorithms(RR,RF,SVR).The value of weather factors on the sales forecast of fresh produce was verified by comparing the prediction results of the two models.(2)Sales forecast model of LSTM neural network considering weather factors.Aiming at the problems that fresh produce sales are susceptible to multivariate characteristics and low prediction accuracy,this paper uses deep neural network optimization technology to build a multivariate multi-input sales prediction model based on LSTM recurrent neural networks.Through the multiple comparison experiments of LSTM model and control model(RR,RF,SVR)under different data sets and different time window data input,the validity and superiority of the LSTM model are verified from three dimensions: accuracy,generalization and robustness.This paper validates that weather factors can improve the accuracy of the fresh produce sales forecast model,and based on this,an LSTM recurrent neural network sales forecast model is constructed,and accurate sales forecast results are obtained.To a certain extent,the research in this paper expands the theory and methods of sales forecast of fresh agricultural products,enriches the research of sales forecast problems,and provides theoretical reference for further research in the future.At the same time,it provides decision support for the operation and management of fresh food retail enterprises.