Ensemble machine learning (ML) methods have been established as effective alternatives to individual models for addressing classification problems. These techniques are based on combining the outputs of multiple base learners using a combination rule. The use of ensemble methods has gained notable interest from the stock market movement prediction research community due to the unstable, volatile, and complex patterns in stock market data, which are difficult to forecast using traditional or single ML techniques. This paper presents an empirical evaluation of three meta-ensemble techniques, including Bagging, Boosting, and Random Subspace, using three base classifiers: Decision Trees, Support Vector Machines (SVM), and Logistic Regression. This study aims to forecast the directional movement of the NASDAQ index closing price (upward = 1, downward = 0), which primarily comprises high-technology companies. The predictive model utilizes OHLCV data (Open, High, Low, Close, Volume) along with ten computed technical indicators as input features. Four evaluation metrics (accuracy, recall, precision, and F1-score) were used to assess the predictive performance of proposed ensembles. Hyperparameter tuning was performed using the Grid Search optimization algorithm. The results demonstrate that ensemble methods outperform individual models in predictive performance. Notably, the ensemble combining Bagged SVM with the Random Subspace method achieves a statistically significant improvement. These findings confirm the effectiveness of integrating ensemble learning with feature subspace methods.