Comparison of different Artificial Intelligence techniques to predict Diabetic Kidney Disease (Preprint)
BACKGROUND Diabetic kidney disease (DKD) is a progressive disease that leads to loss of kidney function. As early intervention improves patient outcomes, it is essential to identify the patients who are at high risk of developing DKD. Artificial Intelligence methods apply different machine learning classification techniques to identify high-risk patients by building a predictive model from a given dataset. OBJECTIVE This study aims to find an accurate classification technique for predicting DKD by comparing different classification techniques applied to a DKD dataset using WEKA machine learning software. METHODS We analyzed the performance of nine different classification techniques on a DKD dataset with 410 instances and 18 attributes. 66% of the dataset was used to build a model, and 33% of the data was used for evaluating the model. The performance of classification techniques were assessed based on their execution time, accuracy, correctly and incorrectly classified instances, kappa statistics (K), mean absolute error, root mean squared error and true values of the confusion matrix. RESULTS Random Forest classifier was found to be the best performing technique with an accuracy of 76.5854% and a higher K value (0.5306) in comparison to other classifiers. Besides, it also showed the lowest root mean squared error rate (0.4007). From the confusion matrix, it was found that there were 46 false-positive instances and 50 false-negative instances from the Random Forest technique. CONCLUSIONS This study identified the Random Forest classification technique as the best performing classifier and accurate prediction method for DKD. CLINICALTRIAL NA