Integrating big data analytics and machine learning algorithms has become increasingly important in the fast-changing landscape of stock market investment. The numerical findings showcase the tangible impact of our methodology on the accuracy and efficiency of stock market trend predictions. Identifying and selecting the most salient features (technical indicators) is critical in predicting the trend direction of exchange-traded funds (ETFs) in emerging markets, leveraging financial and economic indicators. Our methodology encompasses an array of statistical techniques strategically employed to identify critical technical indicators with significant implications for time series problems. We improve the efficacy of our model by performing systematic evaluations of statistical and machine learning methods across multiple sets of features or technical indicators, resulting in a more accurate trend prediction mechanism. Notably, our approach not only achieves a substantial reduction in the computational cost of the proposed neural network model by selecting only 5% of the total technical indicators for predicting ETF trends but also enhances the accuracy rate by approximately 2%.