Design And Implementation Of Agricultural Product Price Early Warning System Based On Deep Learning

China is a big agricultural country.The price of agricultural products not only reflects the economic development level of the country,but also affects the national happiness index to a certain extent.Agricultural products are mainly circulated in farmers' markets,and it is the direct trading place between agricultural practitioners and consumers.Our country is broad,the technical development level of each province has a big gap,which is also reflected in the agricultural informatization.The low level of agricultural informatization will make agricultural workers unable to obtain the latest agricultural information,resulting in blind planting,which will lead to the contradiction between supply and demand in the market and social unrest.Therefore,the wholesale market for agricultural products in reasonable management and timely release of information,and to establish relatively reasonable price prediction and early warning mechanism,can help agricultural practitioners for agricultural information in time,reasonable arrangement of agricultural cultivation plan,at the same time also can give market regulators to provide a set of perfect supervision and warning system.In recent years,the rapid development of emerging technologies,such as big data,artificial intelligence,Internet of Things and 5G,has also provided strong technical support for agricultural informatization.In view of the above mentioned problems,this paper plans to develop an agricultural product price early warning system,which integrates market information management,information display and price prediction and early warning.I conducted the following research:(1)Design of agricultural product price early warning system.The early warning system in this paper adopts the modularized design method to conduct business decoupling at the logical level.According to the different positioning of users,the system is divided into two systems: data display system and information management system.Data display is divided into price data query,prediction and early warning display and market information display;The background management system is composed of market data,personnel information and warning information.(2)Price prediction algorithm design.Aiming at the non-linear relationship between the features of agricultural products and the low multi-day accuracy of traditional price prediction algorithms,this paper proposes a hybrid neural network price prediction model based on feature selection.In the overall design of the model,it starts from the two aspects of feature selection and prediction model network structure.In terms of feature selection,considering the non-linear relationship between features and the redundant influence between features,LightGBM is selected to select features of different categories in the same market,price data of the same category in adjacent markets,environment,market economy and other factors.In terms of network structure,considering the output prediction for the future price in as many days,so choose Seq2 Seq structure,two parts of the network is divided into the Encode and Decode,Encode parts considering the multidimensional characteristics and scheduling characteristics of time series itself,select 1 DCNN and Lstm characteristics of the mixed neural network capture,Decode part using multiple output characteristic of the Lstm,forecast the demand of many days.In addition,in order to validate the presented prediction model on the data set of agricultural products,the effect of constructs such as LSTM,GRU helped,and the SVM and the commonly used prediction model for comparative experiments with better effect,compare the different root mean square error of model prediction RMSE,through analysis of experimental results,this article proposed prediction model in the comparative experiments with other model has a better performance,and better generalization ability.(3)Implementation of agricultural product price early warning system.Through Spring Boot,Vue and other related technologies,the agricultural product price early warning system is built.The Amap API is used to display the market geographic location information,and the Echarts chart library is used to visually display the relevant data of agricultural product price.And the agricultural product price prediction algorithm proposed in this paper is applied to the price warning module,to analyze the predicted price,to carry out the corresponding early warning through the price fluctuation,and to use telephone and email to carry out the early warning processing.This paper analyzes the problems existing in the existing algorithms and solves the problems of nonlinearity in feature selection and multi-day prediction accuracy.Moreover,based on industrial development,it improves the accuracy of agricultural product price prediction and ensures the fast running speed.Based on the algorithm proposed in this paper,the agricultural product price early warning system was built,and the early warning system was tested to ensure the integrity of each functional module of the system.