Time Series forecasting of Agriculture Yield of Cotton with Regression Model Implementation

The farmers benefit from yield prediction as it lowers down their expenses and assists them in obtaining the best prices for their crops. Time series of a signal contains vital information that can be used for predicting the future points in the series. However, such prediction requires careful analysis of the correlations in the time series. Getting the right number of past observations is a very important part of time-series forecasting (lags). We have used auto correlation in the time series of yield of Cotton of all over India from the period 1964-2020. This study employs a comparative analysis between three approaches to examine the accuracy of prediction based on suitable value of time lag. These include a statistical method based on the auto correlation function, the widely used machine learning method of regression analysis, and the best prediction is chosen based on the ideal time-lag value. An experimental data set with a single predictor was used to test the techniques i.e actual yield and its different lags moving a 1-step ahead, 2-step ahead and so on till 5 as a target values (predicted yield). Root Mean Square Error (RMSE) and R-squared were the performance metrics. The investigation demonstrated that the proposed linear, multi linear regression and non-linear regression at degree 4 analysis is the better method for determining the best time-lag value.