Application of supervised learning algorithm to determine the quality of slippers in WEKA

Abstract

This study is driven by the objective of evaluating the effectiveness of various regression algorithms in the prediction of slipper quality. The selected regression algorithms were implemented within the Waikato Environment for Knowledge Analysis. The assessment of their performance was conducted through the analysis of correlation coefficients, providing insights into their predictive capabilities. Notably, the Random Forest algorithm demonstrated the highest predictive power with an impressive correlation coefficient (r=0.76), surpassing other models in the analysis. Following Random Forest, the k-nearest neighbor algorithm achieved a substantial correlation coefficient of (r=0.65), followed by the Decision Tree (r=0.53), Linear regression (r=0.51), and the Multi-layer perceptron (r=0.51). In contrast, the Support Vector Machine showed a notably lower correlation coefficient (r=0.51), indicating its comparatively weaker predictive performance. Furthermore, this study uncovered two variables, "Easy to Wash" and "Water Resistance," which displayed significant correlations of (r=0.49) and (r=-0.35), respectively, in relation to the predictive performance of the regression model. However, no significant correlation was observed for other variables. In light of these findings, future research endeavors may explore alternative predictive models to further assess and compare their performance against the outcomes presented in this study, contributing to the ongoing enhancement of slipper quality prediction methodologies.