Pneumonia is an acute respiratory infection of lung tissue caused by pathogenic microorganisms such as bacteria, viruses, fungi, and other pathogens. It affects all people but is severe in children and adults over 65. This disease can be diagnosed through physical examination, laboratory tests, and chest imaging. But from chest imaging modalities, the chest x-ray is the gold standard examination tool for diagnosis of pneumonia because of its low cost, availability, and amount of information that can be obtained. However, detecting and visualizing pneumonia pathogen is time-consuming, and it is prone to errors due to the subjective nature of radiologists and physicians. In this thesis, a hybrid learning algorithm composed of image processing techniques, deep pre-trained models, nature-inspired algorithms, and classical machine learning algorithms was applied to detect normal, bacterial, and viral cases of pneumonia. The pediatric chest x-ray image of pneumonia was taken from the Kaggle online dataset depository. The proposed hybrid machine learning approach achieved comparable results in terms of accuracy (93.58%), precision (93.63%), recall (93.58%), F1-score (0.9358), and specificity (94.84%) for the classification of normal, viral, and bacterial pneumonia cases. These evaluation metrics are among the highest reported in the reviewed literature for normal, viral, and bacterial pneumonia classification task. Furthermore, when the same model applied for binary classification, e.g., for normal vs. pneumonia, it achieved even higher evaluation metrics. As a limitation, this thesis was not trained and tested on local or public datasets beyond the Kaggle dataset for model development. Therefore, as future work, the proposed model will be trained and tested on local hospital datasets through collaboration with radiologists and pediatric physicians for local use. Keywords: Chest Imaging Modalities, Nature Inspired Algorithms, Pneumonia, Pre-Trained Models