Diabetes prediction with machine learning methods

Artificial intelligence, which is widely used in many fields today, helps people to solve complex problems more easily and quickly. Artificial intelligence can learn, improve and innovate through the data it collects by imitating human intelligence. Machine learning techniques, which are also widely used in the field of health, provide convenience by providing early diagnosis and on-site diagnosis. Within the scope of this study, common machine learning techniques such as Logistic Regression, XGBoost, Naive Bayes, Nearest Neighbor, Support Vector Machines, Decision Tree were applied to Diabetes data by using Python programming language on the Spyder tool and diabetes detection was made. By applying and comparing machine learning methods for the detection of diabetes, the most effective methods for early diagnosis and prediction of future diagnosis are determined. The aim of this study is to find the algorithm that provides the highest accuracy rate for diabetes diagnosis by applying artificial intelligence algorithms to the data collected from a total of 768 patients, the results obtained from these techniques are compared. Artificial intelligence algorithms used in this thesis are Logistic Regression, KNN, XGBoost, Naive Bayes, Random Forest, Support Vector Machines, Decision Tree. When the results obtained from the algorithms are compared, the highest accuracy rate was obtained by the Random Forest and KNN algorithm with 96%.