This paper presents a novel approach to feature selection voting strategies and a combination of hyperparameter optimization techniques to improve the performance of boosting classification. As a case study, the paper focuses on developing robust predictive models capable of accurately classifying different degrees of damage in concrete structures. By leveraging boosting algorithms and optimization strategies, the proposed methodology aims to enhance the efficiency and accuracy of damage classification processes. The findings contribute to advancing the field of structural health monitoring and maintenance by providing efficient and reliable methods for assessing damage in reinforced concrete structures. Experimental results demonstrate the effectiveness of the approach in accurately identifying damage levels in reinforced concrete frames. Moreover, this work improves the performance of the boosting classification models and identifies the most relevant sensors. After the feature selection process and hyperparameter optimization, the best experimental result reaches an F1-score of 0.919, identifying as best accelerometers those located closer to the ground.