부산대학교 도서관

In this study, Feature importance and model evaluation for the classification of the target variable are presented based on the dataset enriched with statistical and wavelet characteristics. The first step involves applying the Random Forest classifier to rank the features based on the significance with which they predict the target variable. We further validated the importance of these features by a set of very rigorous cross-validation procedures, and, in a word, the stability and reliability of our findings came out across different splits of the data. To improve model performance, we explored advanced feature engineering techniques using new domain knowledgebased features that influence model accuracy. The analysis thus highlights the relevance of key feature characteristics concerning classification performance. It gives a beneficial insight into their behavior and interaction. It also enables an all-inclusive approach toward model interpretability. This helps to build a strong base for future work and practical deployment of developed predictive models. This work constitutes an important contribution by putting a significant emphasis on the role of feature selection and engineering in machine learning pipelines, creating a solid framework for further research and application.