Smart Healthcare in Smart Cities: Leveraging Machine Learning for Disease Detection

Machine Learning techniques are leveraged by Smart cities for early identification and intervention of critical diseases. Diabetes and heart diseases are the most common diseases and they have adverse impact on one’s health. Early detection of these diseases can severely reduce their impact. We are applying machine learning algorithms such as Logistic Regression, Decision Trees, KNN, Naïve Bayes, SVM and various ensemble learning techniques such as Random Forest, Bagging Meta Estimator, XGBoost, AdaBoost, Gradient Boost, CatBoost for early detection of respected disease. We have done a comparative study between all the algorithms and chosen the best one by assessing the model in terms of accuracy, recall, f1 score, precision and roc_auc score..Our method can determine whether a patient is detected with a heart disease or diabetes based on selected features of the datasets which were obtained from Kaggle. In case of heart disease, Gradient Boost and CatBoost outperform with Gradient Boost having an accuracy of 88.6% and precision of 88.5% while CatBoost having an accuracy of 89.1% and precision of 89.2%. In case of Diabetes, KNN performs nearly well with an accuracy of 77.9% and precision of 69.2%.