Reserch On Agricultural Product Price Prediction System Based On Micro-service Architecture

In recent years,people’s quality of life has been continuously improved,and the consumption structure of agricultural products has also under gone some changes,especially the proportion of vegetable consumption is incre asing.Changes in the prices of agricultural products have a great impact on ag ricultural practitioners.Therefore,with accurate and effective agricultural market information,it is possible to infer the price fluctuations of agricultural product s in advance,prevent them as soon as possible,and make corresponding decisi ons to avoid major losses.Therefore,the relevant research on forecasting the p rice of agricultural products can enable agricultural related personnel to grasp f orward-looking information,timely and effectively change the thinking of agricultural production and sales,and avoid losses caused by abnormal fluctuations in agricultural prices.However,my country has a large population and a large area,and the information on agricultural products is asymmetric.People canno t accurately grasp the price information of agricultural products and predict the price trend of agricultural products in the future.With the popularization of t he Internet,it has become more convenient for people to obtain information fr om the Internet.Therefore,relevant professionals have developed a system that can understand the price information of agricultural products at the first time and can predict the price of agricultural products for a period of time in the f uture.But most of them are monolithic architecture systems,and there are man y problems in this conventional software design architecture.In order to impro ve the performance of the agricultural product price prediction system,this pap er takes the agricultural product price prediction system as the carrier and buil ds the system combined with the micro-service architecture.The main research work of this paper is as follows:(1)Microservice division based on domain-driven design.Domain-driven design analyzes the use cases of each participant in the system,and uses the domain modeling method to divide the system into domains,analyzes and identifi es each domain element,finds the boundary of each domain,and identifies the bounded context.Bounded context divides the agricultural product price foreca sting system into four microservices: data acquisition,data management,statisti cal analysis and price forecasting.(2)Research on price forecasting model.This paper mainly takes vegetables as an example,selects the representative vegetables of the three major vegetable types,root,solanaceous and leaf,and the data of onion,ginger and garlic from 2014 to 2019 to conduct experiments,and analyzes the experimental results.Among them,the empirical mode decomposition-extreme learning machine(EMD-ELM)combined prediction model is used to analyze the experimental results of three types of vegetables.were 4.59%,2.72%,4.70%,4.79%,3.91%,and 3.83%,and the root mean square errors were 0.38,0.10,0.16,0.28,0.06,and 0.20,respectively.And the evaluation indicators are better than the single ELM and BP neural network prediction models.Then,the Prophet algorithm model was used to predict the prices of vegetables without seasonal characteristics such as onion,ginger and garlic.The results showed that the average absolute errors of the seven-day ones were 0.01,0.12,and 0.23,respectively,and the average percentage errors were 0.15%,0.04%,and 0.10.%,the root mean square error is 0.02,0.12,0.24,respectively.It can be concluded that the Prophe talgorithm model has a better performance in predicting the price of vegetables such as onion,ginger and garlic.(3)System construction based on the Spring Cloud microservice framework.Based on the Spring Cloud microservice framework,this paper designs these models into microservices in terms of collaborative communication of predictive models,uses Rabbit MQ message middleware to coordinate and combine the communication between these services,and then unifies all service call interface s It is connected to the API gateway,and the gateway performs unified access and output.In addition,the Nacos service registry is used,the Sentinel service circuit breaker component,and the Zipkin middleware for service distributed tracking and performance monitoring.The persistent database adopts My SQL,and uses Rediscache to store the results of price prediction,so as to avoid repeated calculation in the future.Finally,the built system was deployed into the Docker container,and the function and performance tests were carried out.The test results showed that all the functions of the system can be used normally,and the performance also meets the normal use requirements of relevant personnel.The EMD-ELM prediction model proposed in this paper is better than the single prediction model in the price prediction effect of three types of vegetables: root,stem,nightshade and leaf,and can better complete the prediction task.After the system,the security and scalability of the system have been improved to a certain extent compared with the traditional agricultural product price prediction system,and the development and deployment efficiency of the system has been significantly improved,and the software delivery has become more agile.Shortened software delivery cycle.