Prediction of Crop Yield with Feature Dimension Reduction Method Using Regression Models

The important factor in the development of the Indian economy is agriculture. Agricultural production supports more than 70% of rural households. Approximately 17% of the GDP of India is contributed by the agricultural sector, making it a significant economic sector. The main aim of the proposed method is to predict crop yield and analyze the best model to predict the yield of crops among various machine learning algorithms, This model can certainly help farmers to take suitable decisions to grow crops and improve the Indian economy. The dataset employed here includes various parameters such as the amount of fertilizer, the area planted, harvested area, etc. In this work lint crop dataset is considered for crop yield prediction. In the initial step, Principal Component Analysis (PCA) is used for the feature dimension reduction without losing important information/context of the data. It achieves this by reducing the number of dimensions in the data. After reducing the dimension of the dataset using PCA, various machine learning regression algorithms such as Linear rearcssion, Lasso regression, Decision tree, Adaboost, and XGBoost are employed to predict the crop yield. Further, the best model is found by comparing the model's performance metrics which include Mean absolute error (MAE), R2 Score, and Root mean square error (RMSE).