<p>Implementing machine learning (ML) methods in performance level assessment of reinforced concrete (RC) shear walls can considerably enhance the speed and accuracy of performance evaluations, which are traditionally time-consuming and require extensive expertise. However, due to the effects of many input features in the seismic behavior of RC shear walls, several challenges, e.g. overfitting, increased complexity and computational cost, sparsity of data, and difficulty in feature selection can lead to low effectiveness of ML models for real-world examples. Therefore, this research is focused on reducing the number of input features effectively while maintaining high prediction accuracy. According to the results, optimizing extreme gradient boosting (XGBoost) and random forest (RF) models with the adaptive tree of Parzen estimators can boost the performance of ML models without the need for hyperparameter tuning. However, by providing the ML models based on two main input features of wall length and period of structure, the coefficients of A, B, and C can be determined for predicting the median of incremental dynamic analysis (M_IDA) curve of RC shear wall based on the maximum interstory drift ratio threshold. Proposed ML models predicted the M_IDA values of a 5-story case study RC shear wall with an accuracy of 97.2% and 96.5% for near-field records, then, the graphical user interface (GUI) is introduced to be used by researchers for preliminary seismic performance assessment.</p>