Research And Realization Of Short - Term Forecasting Model Of Potato Late Blight

Late blight is a destructive pest of potato farming, with the late blight pathogen warming and climate variability, the harm is rising, accurate forecasting is an important basis for the prevention and control of the spread of the disaster. Therefore, on how to improve prediction accuracy problems, the problem has been the focus of potato late blight forecasting short-term studies.First, this paper analyzes common pest modeling methods, mainly regression analysis and time series, and Malong potato late blight disease in an area to establish a short-term forecasting data model based on a single method. Through the analysis of a single prediction model, although considered a single model can forecast the development trend of late blight, but there are some limitations, such as:time series forecasting model based on changes in the larger data for late blight description ineffective, in this paper, although the use of the seasonal difference method to smooth sequence, but in the95%confidence level forecast wide confidence intervals, prediction is not ideal; forecasting model based on linear regression analysis, for complex systems such as late blight, often due to the level of awareness of the limitations of the researcher, the model is very difficult to satisfy unrelated residual resistance, stability and normal distribution and other basic assumptions, the paper while taking a correlation analysis and cluster analysis to select variables from the regression, the regression equation can achieve a high fitting rate, but it still can not meet the basic assumptions of the residuals.On this basis, we design a hybrid forecasting model multivariate polynomial regression and time series-based forecasting to overcome the limitations of a single model. The specific modeling approach is:first, the use of multivariate polynomial regression method to extract information on late blight affecting the same period, to study the relationship between late blight disease with the same period of meteorological factors, grasp the essence of late blight disease causes of change, multivariate polynomial regression equations derived from a high fitting rate; second step, using time series method to extract information on the impact of late blight in different periods, which is part of multivariate polynomial regression residuals build ARIMA models. In this paper, a region of potato late blight disease Malong build hybrid forecasting model data, the results show that the measured values of the fitting rate reached92%, to overcome the limitations of a single model, and effectively improve the forecasting accuracy.Finally, a hybrid forecasting model as the core algorithm prepared potato late blight short-term forecasting system for short-term forecasting Malong other areas, late blight forecasting results and the actual situation is basically consistent, can provide effective protection advice to potato growers, but also mixed forecast model proved multivariate polynomial regression and time series based on potato late blight is an effective short-term forecasting models.