In order to improve diagnostic accuracy in clinical practice, machine learning techniques are being investigated extensively, including pediatric respiratory diseases. Prior research has shown that ensemble models like Random Forest and XGBoost can effectively improve predictions for conditions such as bronchopulmonary dysplasia and bronchiolitis. However, distinguishing between similar respiratory conditions, like bronchitis, bronchopneumonia, and URTIs in children, remains challenging due to symptom overlap. In this work, we analyze the accuracy of seven ML classifiers in distinguishing between these respiratory diseases in children under the age of five; the results provide 98.75% accuracy using the Random Forest classifier. Key diagnostic characteristics including pulse rate and duration of fever are identified, Pulse Rate (bpm), Serum Creatinine (mg/dL), Crepitations, Past History and Fever (days) among which contribute to better classification. The data were collected from FMMC Hospital, a tertiary care hospital in Mangalore, India, such as Medical Records and Registration Departments. EMRs of 591 children aged 0–5 years diagnosed of respiratory diseases, which included URTI, brochopenumonia and bronchitis, were evaluated. Access to the data has been permitted by the Scientific and Ethical Committee of FMMC. This knowledge, reveal the contribution of Machine Learning technology.