DIABETES PREDICTION USING F-TEST RANDOM FOREST TECHNIQUES

Abstract

One of the deadliest diseases regarded as the eighth leading cause of high mortality rate in developing countries is diabetes. It's the most dangerous chronic disease that if not properly monitored could lead to other serious complicating diseases such as heart attack, hypertension blindness, deterioration of the kidney, and nervous system, and many more health-related problems. Deficiency or ineffective production of insulin by the pancreas can be traced to be the major cause. Medical practitioners do advise a healthy balanced routine, exercise, and early treatment. In recent times machine learning algorithms have been helpful for this purpose. However, these algorithms differ in their ability to handle complex and nonlinear relationships between predictors and outcomes. Also, the performance of these algorithms may depend on the characteristics of the dataset, such as sample size, feature selection, and missing data overfitting, and lower classification accuracy is often noticed due to the problem of high-dimensional diabetes datasets. This has made it difficult for researchers to identify the most promising technique for the diabetes prediction process. Therefore to develop a suitable model for the diabetes detection process. Hence this study aims to develop an F-test Random Forest (F-test-RF) Model for the diabetes detection process. F-test was used to select the optimal features subset which was then used to build a Random Forest classifier for the diabetes prediction process. A seven (7) Month old diabetes dataset was obtained from the Kaggle repository, The dataset contains Nine (9) attributes and 10001 responses. Python programming language was used to implement the study. The study was analyzed based on different performance matrixes. Results were obtained in terms of percentage for accuracy, Precision, specificity, and sensitivity where 96.83%, 90.68%, 99.33%, and 70.08% respectively. Results showed that the performance of the developed model was highly impressive. F-test-Rf can be recommended for the diabetic detection process. The Developed models show confidence in their knowledge discovery from the dataset used. Keywords: Filter Technique. Diabetes Detection. Random Forest. F-Test.