Early Cocoa Blackpod Pathogen Prediction with Machine Learning Ensemble Algorithm based on Climatic Parameters

Abstract

Machine learning has been useful for prediction in the various sectors of the economy. The research work proposed an ensemble SA-CCT machine learning algorithm that gives early and accurate prediction of blackpod disease to farmers and agricultural extension officers in South-West, Nigeria. Since data mining put into consideration the types of pattern in a given dataset, the study considered the pattern in climatic dataset retrieved from Nigeria Meteorological agency (NIMET). The proposed model uses climatic parameters (Rainfall and Temperature) to predict the outbreak of blackpod disease. The ensemble SA-CCT model was formulated by hybridizing a linear algorithm Seasonal Auto Regressive Integrated Moving Average (SARIMA) and a nonlinear algorithm Compact Classification Tree (CCT), the implementation was done with python programming. The proposed SA-CCT model gives the following results after evaluation. Precision: 0.9429, Recall 0.9167, Mean Square Error: 0.2357, Accuracy: 0.9444