Pneumonia detection from Chest X-Ray (CXR) images is critical for timely and accurate diagnosis to ensure effective treatment. But it remains as a challenging work due to the complexity of image features and variation in pneumonia presentations. This study presents a hybrid deep learning framework to effectively address these difficulties by the integration of multi-scale feature extraction, deep feature selection (DFS), and ensemble classification. A deep Convolutional Neural Network (CNN) with residual connections is used to learn the multi-scale feature presentation for various patterns associated with pneumonia and handles the challenges during training. The DFS layer selects the most relevant features and Principal Component Analysis (PCA) reduces dimensionality to improve model performance. For classification process the framework employs an ensemble approach combines the Support Vector Machine (SVM) with Radial Basis Function (RBF) kernel along with Random Forest (RF) classifiers. The ensemble voting mechanism takes advantage of the strengths of both classifiers to improve overall performance. Evaluated on the public CXR dataset the proposed framework achieved a good accuracy of 94.6%, with precision 95%, F1-score of 94.6% and recall of 94.4%. These results demonstrate a marked improvement over traditional deep learning models and standalone machine learning classifiers, highlighting the framework's effectiveness in enhancing pneumonia detection accuracy and reliability.