The Application Research Of Fruit Demand Forecasting Based On Hybrid Model

With the rapid development of economy,people's living standards have been continuously improved.While the demand for high-quality fresh fruits has continued to increase.In addition,because of the perishable nature of fruits,unreasonable judgments about the future market demand will cause a large increase in the operating costs of related companies.Therefore,how to reasonably forecast the demand for fruit and establish a demand inventory model to achieve a reasonable relationship between the supply and demand of fruits in a certain area.And to reduce the inventory cost of vendors and suppliers is an urgent problem to be solved.This article analyzes and compares the current mainstream quantitative prediction methods,and in view of the low prediction accuracy of the forecasting model and the overly subjective choice of model parameters,which combined with case analysis done the following work:Firstly,under the premise that the main research content of this paper is to forecast the fruit demand.By comparing the advantages,disadvantages and applicability of various current prediction methods,the BP neural network prediction model,RBF neural network prediction model,and support vector regression prediction model are selected as the main components of this prediction model.And the inertial weight adjustment strategy of basic particle swarm optimization algorithm is improved.The optimized particle swarm algorithm is used to optimize the initial parameters of the three prediction methods to achieve the objective of objectively determining the initial parameters of the basic prediction model,thereby reducing the possibility of subjectively determining the initial parameters that adversely affect the performance of the predictive model.Meanwhile,the thesis analyzes the influencing factors affecting fruit demand and determine the seven subjective and objective input indicators such as product price and weather factors.After the grey correlation analysis was performed to sort the indicators,the 731 sets of raw data from January 1,2016 to December 31,2017 were selected as the input data for the prediction model to make predictions.Secondly,the data samples are trained and predicted which used to the optimized prediction method.Combining the prediction residuals with the original data of the sample,the Kalman filter algorithm is used to modify the initial prediction results twice.Then,the information entropy method was used to construct the two indicators of the relative error and the absolute deviation rate of the error of the second-fitting value,and the result fusion process of the three forecasting methods was transformed into the evaluation of the three forecasting methods.The weights are determined objectively according to the actual comprehensive scores of each plan,so as to obtain the predicted value of the final fruit demand.Thirdly,after obtaining the final prediction data,RMSE,MAE,and other evaluation indicators are selected to evaluate and validate the prediction model.The verification results show that the accuracy of the hybrid prediction model proposed in this paper reaches 98.56% when it predicts the 631 training samples.The accuracy of the prediction of the 100 groups of samples using the trained hybrid prediction model also reaches 95.28%,achieving the goal of accurate prediction.The accuracy is improved by about 5% compared with the conventional prediction model and the single prediction model optimized by the particle swarm optimization algorithm in this paper.Finally,after analyzing the existing inventory strategies of fruit sellers and superior suppliers and considering the basic factors such as safety stock and order cycle,a mathematical model of reordering strategy based on demand forecasting is given.In addition,on the basis of analyzing the inventory costs,transportation costs,and stock loss costs of the vendors and suppliers,the corresponding inventory optimization models were established and solved with examples.The result of the example shows that the inventory level of the two-level sellers is reduced by 12.27% compared with the subjective judgment,which effectively reduces the inventory costs of the suppliers and sellers and provides an important reference for the company to formulate inventory strategies.It also can avoid the losses caused by unreasonable inventory management strategies and maximize the profits and efficiency of the company.From the above,we can see that the proposed hybrid prediction model and inventory control model have high theoretical significance and practical value.