Machine Learning Approaches for Banana Quality Classification: A Performance Comparison of J48 and Random Forest Algorithms

This study evaluates the performance of using machine learning models; J48 and Random Forest to classify bananas quality. The existing methods of visual inspection are qualitative and take much time more than the ML technique which offers quantitative approaches to the same issue. The study adopted data from Kaggle’s Banana Quality Dataset it has 1200 instances and different features. Data preparation was done by converting the data into ARFF format, normalizing data values as well as selecting the most relevant features out of all features which are then used to split the dataset into training and testing set. In the current study, both J48 and Random Forest models developed in WEKA software were used. It was found that results of Random Forest were high in comparison to J48 with regard to accuracy, precision, recall and economic aspects. Based on this, this study also assumes that Random Forest is indeed better to use in the prediction of the quality of bananas as it improves on the measurement precision and that it saves on time significantly.