Study On Forecast Model And Prevention Indexes Of Tobacco Wildfire And Angular Spot Disease

Tobacco wildfire disease and angular spot disease are two important leaf diseases in tobacco planting in Jilin province and even in the whole country.At present,most of the researches on these two diseases are related to the control methods and varieties resistance.However,there are few reports on the prevalence and prediction of these two diseases.This article mainly studies the disease growth models of tobacco wildfire and angular spot disease in Jilin Province,and establishes a tobacco wildfire forecast model and prevention indicators.The results are as follows:With SPSS18.0 software in six different mathematical function model of tobacco wildfire disease and angular spot disease index sequence value at any time and change of data analysis,through the determination coefficient R²,F value and standard error of the three indicators of these six kinds of different mathematical models are compared,and one by one the results show that the logistic function model can well simulate the tobacco wildfire disease and angular spot disease of illness development dynamic,three function also can well simulate the condition of tobacco wildfire disease development.Taking the tobacco wildfire disease index in early August as the prediction object,temperature,humidity,rainfall and sunshine hours as the predictors,a multiple regression prediction model was established with SPSS software.Regression analysis was performed on the predicted object?Y?and the predicted factor?X?.The average humidity in mid-May?X3?,the average humidity in mid-to-late May?X4?,and June were selected by comparing the determination coefficients R²and F values.The rainfall in the middle?X6?and the sunshine hours?X7?in mid-May were the optimal predictors.An optimal multiple regression prediction model was established,Y=-18.494+0.076 X3+0.348 X4-0.134 X6+0.138X7.After fitting test,the average fitting rate of the model is 75.38%.The core program of the BP neural network model was established using MATLAB 2018b software.The data of the prediction object and the prediction factor are trained.When the number of hidden layer nodes is 8,the regression coefficient R is 1,and the training error value of the BP neural network is the smallest.The normalized test sample was used to test the neural network with 8 hidden layer nodes.The test output was normalized to obtain a fitted disease index.The average fitting rate was 87.34%.The fitting degree of BP neural network model is higher than that of multiple regression prediction model,and the error is smaller.SPSS software was used to perform regression analysis on the tobacco wildfire disease index and output value loss rate data.By determining the coefficient R²,standard error and F value,the six models of linearity,quadratic term,cube,index,power,and logic were compared and compared.Test,the power function has the highest fitting degree,and the optimal regression equation is Y=4.8266X^0.6599.According to the formula:allowable loss rate?%?=3×hectare cost/?ha output value×prevention effect?,calculate the allowable loss rate of 5.895%.Substituting the value 5.895 as the Y value into the optimal regression equation Y=4.8266X^0.6599,and calculating the allowable loss of the disease index X is 2.45,which is the allowable level of economic harm?EIL?.The EIL value of 2.45 was substituted into the linear regression equation of the disease index?Y?after the control and the disease index?X?at the beginning of the control:Y=-0.17+3.192X.?X=ET?and ET=0.82,indicating that the indicators for the prevention indicators of tobacco wildfire are between 0.82 and 2.45.The proposed control index has practical guiding significance for the pharmaceutical control of tobacco wildfire.