ABSTRACT
Agriculture is often regarded as the principal means of ensuring food provision worldwide. Besides being the supplier of food, this sector has contributions to supply raw materials in different industries in our country. In light of declining crop production and food shortages around the world, one of the most important criteria in agriculture today is predicting future yields and selecting the right crop for the right land at the right time. Therefore, in this study we have proposed a method that will help us by predicting crop production based on previous year data analysis on some predictive parameters using machine learning. Although several studies have been conducted in this area, most of them are not in the context of Bangladesh. In this study, we applied Support Vector Machine (SVM), Random Forest (RF) and Lasso algorithms for crop yield prediction using our collected dataset of major rice crops (Aush, Aman, and Boro). The dataset contains the weather factors (temperature, rainfall, humidity) data from 2015 to 2022 as predictors. A comparative analysis of machine learning algorithms is performed based on evaluation metrics (MAE, MSE, RMSE) values. RF shows the best crop yield prediction results with least error values for two types of crops: Aman (MAE = 0.02511, MSE= 0.00138, RMSE = 0.0371) and Boro (MAE = 0.02576, MSE = 0.00112, RMSE = 0.03352). However,Lasso notes the least error values for Aus which account for MAE of 0.030260, MSE of .00124 and RMSE of 0.03524. From our experiments, we can estimate that RF demonstrates the optimum result while taking into account several predicting properties and performance metrics.
Key words: Rice crop, machine Learning, SVM, RF, Lasso
Recent Comments