Plastic-type prediction based on digital image using multinomial Naïve Bayes method

Plastic is a solid waste that can harm the environment. Recycling is a way to reduce plastic waste by reusing rather than throwing it away. This way is also helpful for reducing pollution due to greenhouse gas emissions generated in processing new plastic from raw materials. Nowadays, plastic recycling activities continue to increase. The initial step in the plastic waste recycling process involves sorting plastic to distinguish different types of material. Accurately identifying the plastic-type is very useful for sorting system building in the recycling industry. This paper aims to investigate the plastic-type prediction performance using multinomial naive Bayes. Multinomial naive Bayes is a statistical learning method that takes decisions as the basis for predictions for the target variable based on Bayes’ theorem. This method extends the naïve Bayes method, which assumes that predictor variables have a normal distribution. Multinomial naive Bayes method to accommodate violations of this assumption or if the variable is not of continuous type. This method frequently has adequate performance compared to other methods in many applications. The validation of the prediction result is evaluated based on the k-fold cross-validation method. The results showed that the multinomial naive Bayes method has a satisfactory performance on all performance measures in identification the plastic type. The performance measures are 97.34% accuracy, 96% recall-µ, 96.07 % recall-M, 98 % specificity-µ, and specificity-M 95.57%.