Analysing Machine Learning based Approaches for Detecting Late Blight Disease in Potato Crop

Abstract

Agriculture is a significant contributor in the world economy. With the drastic change in the geographical conditions, the occurrence of extreme events like floods, droughts, heat waves, etc. are increasing, thereby harming crop yield. Additionally, crop yield is adversely impacted by crop diseases causing significant losses towards food production. Protecting against losses incurred by crop diseases can aid in improving food security as well as strengthening the economy. Traditional methods of crop disease detection are time and labor-intensive, whereas the use of machine learning (ML) based methods fastens up the process, thereby helping implement corrective actions at an early stage. Multiple ML algorithms find application in the field of crop disease detection. Nevertheless, there exists a need to investigate the accuracies of different ML algorithm with regard to disease detection for a specific crop and disease combination. The performance of three ML algorithms, namely Random Forest, Linear Discriminant Analysis, and k-Nearest Neighbors with respect to late blight disease in potatoes was investigated in this work.