Feature selection and fuzzy classification for the prediction of CO2 ratings

Abstract

The rapid growth of technology has led to the generation of large amounts of data in all possible sectors. There are varieties of algorithms that can be used to analyze and learn about the data in the machine-learning world. It is a best practice to preprocess the data before it undergoes any kind of learning. Feature selection is one of the steps done in the data preprocessing. In this study, we will be focusing on the fuzzy entropy-based feature selection method together with 3 different fuzzy classifiers. The features selected from the dataset will be tested in FkNN, Similarity-based, and ANFIS classification models. The accuracies were obtained from each model. Based on the highest accuracies FkNN 98%, Similarity 95%, and ANFIS 97% obtained, respectively. The performance measures such as sensitivity, specificity, and precision values were examined for each model and it was found that the classes were imbalanced and all 3 models failed to predict the positive incidents, while the models predicted the negative incidents more accurately.