A Machine Learning Approach to Predict Crop Yield and Success Rate

In India agriculture contributes approximately 23% of GDP and employed workforce percentage is 59%. India is the second-largest producer of agriculture crops. the technological contribution may help the farmer to get more yield. The prediction of the yield of different crops may help the farmer regarding taking the decision about which crop to grow. The research focuses on the prediction of different crops yield using neural network regression modeling. The data of crop cycle for summer, Kharif, rabi, autumn and whole year is used. The dataset is resourced from an Indian government website. The experimental parameters considered for study are cultivation area, crop, state, district, season, year and production or yield for the period of 1998 to 2014. The dataset consists of 2 lakh 40 thousand records. The dataset is filtered using Python Pandas and Pandas Profiling tools to retrieve data for Maharashtra state. The model is developed using a Multilayer perceptron neural network. Initially the result obtained considering optimizer RMS prop with accuracy 45 %, later it will be enhanced to 90% by increasing layers, adjusting weight, bias and changing optimizer to Adam. This research describes the development of a different crop yield prediction model with ANN, with 3 Layer Neural Network. The ANN model develops a formula to ascertain the relationship using a large number of input and output examples, to establish model for yield predictions an Activation function: Rectified Linear activation unit (Relu) is used. The backward and forward propagation techniques are used.