Opposition-Based Learning Binary Bat Algorithm as Feature Selection Approach in Taguchi's T-Method

• Main Authors: Marlan Z.M., Jamaludin K.R., Harudin N.
• Other Authors: 57223885180
• Format: Conference Paper
• Published: Institute of Electrical and Electronics Engineers Inc. 2024
• Subjects: Binary Bat Algorithm; Feature Selection; Opposition-Based Learning; Prediction Model; Taguchi's T-method; Forecasting; Learning algorithms; Learning systems; Bat algorithms; Binary bat algorithm; Climate science; Features selection; Model optimization; Modelling techniques; Opposition-based learning; Orthogonal array; Prediction modelling; Taguchi T-method

Prediction modeling has emerged as a powerful tool in various fields, from healthcare to finance, climate science to marketing. One of the prediction modelling techniques available is known as Taguchi's T-method introduced by Dr. Genichi Taguchi. In the T-method prediction model, optimization of the model's accuracy is performed through feature selection process by utilizing an orthogonal array. However, the outcome yielded a sub-optimal result as the orthogonal array has limitation involving a fixed and limited combination used and lack of higher order feature combination in the analysis. Thus, this study proposed an Opposition-based Learning Binary Bat Algorithm as the feature selection technique in the T-method. Based on the experimental results, the proposed feature selection method successfully found a superior combination that yields a better result in terms of the objective function. The proposed method recorded a 77.8% reduction rate of the number of features from 18 to 4. In terms of prediction accuracy, the new T-method prediction model successfully improved 15.9% as compared to the model without feature selection and the T-method with conventional orthogonal array approach. These results suggest that the new T-method prediction model is better in predicting the output even when only 4 features incorporated in the model. � 2023 IEEE.