The assessment of seismic vulnerability in buildings is a pressing global concern, as earthquakes pose a significant threat to both structural integrity and human lives. Diverse nations employ varying strategies and techniques to mitigate the potentially catastrophic effects of seismic events. Despite a comprehensive understanding of structural analysis, conducting thorough nonlinear analyses on every building within a target area is impractical. Earthquakes exhibit variability in size, intensity, and duration, necessitating a more efficient approach to preventative safety measures. This study explores the application of accurate prediction models, specifically leveraging various machine learning (ML) methods to predict seismic vulnerability. The primary objective is to identify key features through ML models that contribute most effectively to classifier performance when predicting alert categories. By harnessing the power of machine learning, this research aims to enhance the precision and efficiency of seismic vulnerability assessments. Through the development and optimization of predictive models, the study seeks to provide valuable insights that can inform targeted and proactive measures to safeguard buildings against the detrimental impacts of earthquakes. In this study, the prediction of seismic vulnerability was conducted employing particle swarm optimization (PSO) and genetic algorithms (GA) as feature extraction techniques. These methods were applied in conjunction with various classifiers to identify the most impactful features for optimizing the performance of each classifier. The performance was tested using accuracy. It was observed that AdaBoost classifier gave a prominent change in accuracy after using PSO and GA. The classifier performed better with PSO. The accuracy obtained was 90.2%.