Feature selection plays a crucial role in improving the performance of predictive models, especially in complex and dynamic environments such as the Brazilian financial market. This study evaluates the impact of different feature selection methods—OneR, Symmetrical Uncertain, Gain Ratio, Pearson Correlation, mRMR, and Correlation-based Feature Subset Selection—on financial market prediction using Random Forest models. A dataset consisting of 194 technical analysis indicators was analyzed, and progressive attribute removal was applied, with model performance assessed through cross-validation. The results demonstrate that feature selection significantly enhances model efficiency by reducing dimensionality while maintaining predictive accuracy. Moreover, our findings provide valuable insights into the relevance of different financial indicators, offering a methodological contribution to financial market analysis. Future work includes leveraging the selected features in advanced machine learning architectures, such as deep learning, to further refine prediction accuracy and develop more reliable models for financial forecasting.