In cancer treatment, model interpretability is crucial. Random Forest (RF) models are well-suited for this task due to their ability to deliver accurate predictions while highlighting feature importance, aiding biological understanding of kinase inhibitor resistance. However, RF performance relies heavily on optimal hyperparameter tuning. To address this, Bayesian optimization algorithm was used to select the best values for the hyperparameters and maximize AUROC through systematic tuning. The used dataset is unbalanced, and to avoid degrading the performance of the proposed approach, Synthetic Minority Oversampling Technique (SMOTE) was applied to balance the training data and improve detection of resistant mutations. The proposed approach includes an integrated pipeline combining RF modeling, SMOTE balancing, Bayesian hyperparameter optimization, and feature selection. variants were trained using different subsets of 56 molecular descriptors. The proposed approach was evaluated using multiple metrics such as accuracy, precision, recall, and F1-score. The obtained results show the efficiency of the proposed approach, as it yields an accuracy of 0.91, especially in identifying resistant mutations. The proposed approach achieved a recall of 0.6 and a precision of 0.75 on test data, The performance of the proposed approach was also compared with the performance of state-of-the-art models, and the comparison results show the superior of the proposed approach.