<![CDATA[The growing demand for remote healthcare solutions has increased the importance of efficient disease diagnosis based on textual symptom descriptions. This study explores the application of machine learning models Multinomial Naive Bayes, Random Forest, and Support Vector Machine (SVM) to classify 24 different diseases from natural language symptom inputs. Utilizing a dataset of 1,200 balanced samples and TF-IDF for feature extraction, we trained and evaluated the models using both accuracy and cross-validation metrics. Among the models, SVM achieved the highest test accuracy of 97.5% and demonstrated consistent performance across all disease categories. These findings underscore the potential of classical machine learning approaches in enhancing digital diagnostic tools, particularly for early screening in telemedicine applications. Future work could extend this study by integrating deep learning architectures and multilingual capabilities to accommodate broader and more diverse healthcare scenarios.]]>
