predictable models

The Sunn pest, Eurygaster integriceps Put. is one of the main pests of wheat and one of the most important plant protection problems in Iran. Multiple linear regression models have been used to predict the fluctuation of various pest populations by using environmental variables. Using intelligent systems to better estimate insect population fluctuations can lead to better results. Therefore, the current study was conducted to predict population fluctuation of the Sunn pest by using a neuro adaptive fuzzy inference system, a response surface method and multiple linear regression. This study was done during in 2015 and 2016 on two wheat farms each one with an area of one hectare in the Chadegan county. In these models, average temperature, average relative humidity, rainfall, wind speed and direction, sampling date, degree- day and altitude from sea level were used as response variables. The collected data randomly divided in two categories of training (70%) and testing (30%) and they used for train and test of ANFIS and response surface methodology. The accuracy of the prediction was evaluated by R2 and RMSE. The higher performance of the ANFIS model (R2= 0.93, RMSE= 0.0614) and RSM (R2 = 0.88, RMSE= 0.0836) resulted comparing to the multiple linear regression (R2 = 0.23, RMSE= 0.34). Also the results of sensitivity analysis indicated that the average of temperature, relative humidity, wind speed and date of sampling were the most important parameters for predicting density of adult Sunn pest.