Forecasting Rainfed Agricultural Production in Arid and Semi-Arid Lands Using Learning Machine Methods: A Case Study

With the rising demand for food products and direct impact of climate change on production in many parts world, recent years have seen growing interest subject security role rainfed farming this area. Machine learning methods can be used to predict crop yield based a combination remote sensing data collected by ground weather stations. This paper argues that forecasting drylands reliable management purpose under uncertain conditions using machine determines which indicators are most important predicting chickpea. In study, chickpea farms 11 top producing counties Kermanshah province, Iran, was predicted three methods, namely support vector regression (SVR), random forest (RF), K-nearest neighbors (KNN). To improve prediction accuracy, each county, were overlaid satellite images with suitable slope altitude farming. An integrated database created combining data, statistics. The evaluated leave-one-out cross-validation (LOOCV) technique compared terms multiple measures. Given sensitivity time predictions made two scenarios: (1) averages all months, (2) months. results showed RF provides more accurate than other methods. method 7–8% different from statistics reported Statistical Center Ministry Agriculture Iran. It found pre-harvest yield, March–April period (the months) offers best result correlation coefficient relationship between predictor indices.