Resilience Approaches and Comparative Evaluation of Classical Time Series Models in Modeling and Forecasting Wheat Production

Abstract

Prediction is a complex task that is based on different statistical methods and scientific knowledge. Using statistical methods, the current study aims to forecast wheat output in the Pakistani province of Sindh. To forecast the wheat harvest, secondary data on acreage, output, and yields have been gathered over the past thirty years (1984–1985 to 2013–2014). The objectives of this work are to utilize moving averages (MA) and exponential smoothing (ES) consisting of two time series methods that indicate the time series variable (wheat) with less error and compare both methods. We apply the 3,5 and 7-year moving average techniques to check which of them is performing the best without losing the information. In the second technique, we use the exponential smoothing methods to evaluate and forecast the time series variable. The models are contrasted with one another and chosen using the lowest Key Performance indicators. The analysis is conducted on R software (3.4.1) and Minitab. Based on the results, it is found that exponential smoothing performed better than the moving average. Moreover, using these methods, it is observed that the study variable follows an upward-increasing trend.