AI in Agriculture: Yield Prediction Using Machine Learning

Abstract

Weather prediction especially in terms of rain and climate is important in the germination and growth of food crops to feed the populace as well as in the proper utilization of inputs like fertilizers and water supplies. This work falls under the broad research area of how certain environmental factors in the context of rainfall, temperature; affects the use of fertilizer and macronutrient, thereby affecting crop yield. Because agricultural systems are diverse and dynamic, there is a clear need to incorporate these strong analysis methods to aid with making better predictions concerning yields, as well as farming practices in the future. In the current study, the density estimation techniques and analysis clearly indicated bimodal distribution of the selected input variables thus revealing the presence of two different crops within the data set. One crop type is less annual water demand type that prefers 400-500 mm rainfall and 25-30oC temperature whereas the other crop type demands heavier more than 1100 mm rainfall and 35-40oC temperature. The investigation also described the means yield relation, and it depicted that there is a simple relationship between yield and nutrient concentration though it also indicated high coefficient of variation emphasized effect of other factors such as type of soil, climate, variety of crops. When it comes to quantitative prediction of crops, algorithms like Decision Tree Regressor and Random Forest Regressor are utilized as and when possible. Random Forest Regressor thus seems to be a better option than Decision Tree Regressor due to reasons of accuracy and the potential to handle non-linear data. The results reiterate that agricultural productivity is not a unidimensional one, it is a multifaceted construct and there is a clear need to identify more predictors of yield. The research is useful in extending knowledge on the factors that determine the crop yield.