ENETIC ALGORITHM BASED FEATURE SELECTION AND RANDOM FOREST MODEL FOR RICE YIELD PREDICTION

Abstract

As a major contributor of Rice crop production, India ranks number two after China in whole over the world as
a leading country to produce 116.42 million metric tons of Rice in 2018-2019. The production of rice is most
suitable for a country like India with favourable climatic and weather condition and majority of population with
farmer as occupation. So in order to meet the population demand and growth of countrys economy, the major
saviour is the rice yield prediction which can predict the amount of Rice by giving the pre-productive
information to farmers and agronomist to boost up the production level. In this paper, we come up with an
approach to predict the rice yield by making use of collected weather, climatic and agricultural raw materials
related data to measure the quantifiable amount of Rice collected from fields. In order to predict the amount of
Rice yield, agricultural materials like amount of seeds, pesticides, fertilizers, weather condition like amount of
rain and temperature in addition with PH and moisture content of soil are considered which can be correlated
with other environmental factors and the yield amount can be predicted by using these determinant factors
performing regression analysis by implementing different machine learning models like Linear Regression,
SVR, K-Neighbors Regression, Random Forest and Decision tree to compare the value of each regression
model with a range analysis of MAPE value collected from each model and genetic algorithm is implemented
for feature selection purpose in order to reduce measured MSE and MAPE values up to certain extent