Performance of plastic-type prediction using decision tree approaches.

Currently, plastic is used for various purposes, and the waste generated by that use is increasing rapidly. Recycling is a technique for reducing plastic waste by reusing rather than discarding it. Plastic recycling activities continue to grow in popularity today. This activity is beneficial for resolving environmental issues caused by plastic waste and increasing the perceived value of plastic's benefits. The first step in the plastic waste recycling process is sorting plastic to determine its composition. Accurately predicting the plastic-type is highly beneficial when developing sorting systems in the industry. Decision Tree is a statistical learning approach widely used to predict tasks when the target variable is categorical. This method has two popular types; Decision Tree ID3 and Decision Tree C45. Decision Tree C45 can handle all variables, but Decision Tree ID3 can handle categorical type variables only. The purpose of this study is to evaluate the performance of plastic-type prediction using both of the Decision Tree types approach based on a k-fold cross-validation resampling method. For ID3, the numerical variables were transformed into categorical type variables using discretization. The performances are assessed using some metric measures for assessing the performance; accuracy, the micro and macro proportion of plastic-type with correctly predicted (recall), and the micro and macro the proportion of the plastic-type into other types predicted correctly (specificity). The results indicate that the prediction performance of both Decision Trees types differs, and the Decision Tree C45 performs admirably on all performance metrics. [ABSTRACT FROM AUTHOR]