Applying Machine Learning Clustering and Classification to Predict Banana Ripeness States and Shelf Life

Abstract

Food waste accounts for over $15 billion annually. Not only is food waste a financial problem, it is also an ethical problem. The use of machine learning algorithms for clustering and classification provide an opportunity to help reduce food waste. K-Means clustering is proposed to determine banana ripeness states and the Decision Tree Classifier algorithm is proposed to classify banana shelf-life. An experiment is undertaken to provide data by imaging bananas and extracting color features using computer vision. The resultant data is then clustered to determine banana ripeness states. The states are used to determine the end-point of data collection. Seven different machine learning classification algorithms are tested to classify fruit shelf-life. The most accurate classifier is the Decision Tree Classifier which has an accuracy around 52%. The combination of machine learning algorithms and big data analysis becomes a powerful tool in working to reduce food waste.