Classifying Wastes Using Random Forests, Gaussian Naïve Bayes, Support Vector Machine and Multilayer Perceptron

Abstract

As the population of the Philippines increases in size, so does the amount of garbage that it generates. A lot of this garbage is not properly segregated when dumped andmany citizens are not aware of the effects of unsegregated garbage which leads to them simply disposing mixed waste in dump sites. A method to help classify garbage can help address this issue. This paper presents a comparison of the efficiency of four commonly used machine learning models, Random Forests, Gaussian Naïve Bayes, Support Vector Machines and Multilayer Perceptron in classifying biodegradable and non-biodegradable waste. The data used for training and testing was collected from various sources and then augmented to increase the size of the dataset resulting in 15,000 images of biodegradable and non-biodegradable trash used for training. The four models were trained using the K-Fold cross validation technique with the dataset being split ten times. The results indicate that the four models were able to achieve a relatively high accuracy in classifying images of biodegradable and non-biodegradable trash with the Random Forests, Gaussian Naïve Bayes, Support Vector Machines and Multilayer Perceptron models achieving an accuracy of 97.49%, 81.46%, 89.51% and 96.44% respectively.