Diabetes Prediction Using Blood Sample Data with Novel Voting Classifier over Random Forest

Abstract

This study focuses on how to predict diabetes using blood sample data and machine learning algorithms like the Voting Classifier over the Random Forest technique. The proposed prediction models were trained and evaluated on a dataset that included seven variables: glucose level, diastolic blood pressure, blood thickness, insulin levels, BMI, age, and skin. The new Voting classifier (VC) and Random Forest (RF) algorithms are used on a diabetes dataset of 1495 records with 10 features, sample size=5, and two groups with a g-power value of 80%. With a threshold of 0.05, a confidence interval of 95 percent, and a standard deviation of one standard deviation, the patients’ information was acquired from a variety of websites. The framework was built using blood sample data and the VC over Random Forest machine learning algorithm, resulting in a successful research of diabetes prediction using blood sample data and the Voting Classifier (95%) over Random Forest machine learning technique (85 percent). With a 95 percent confidence interval, the two-tailed t-test revealed a statistical significance value of 0.001 (p0.05). This research shows that the VC algorithm’s results are more accurate than the RF approach, which was written in Python.