Alzheimer’s Disease Detection using Weighted KNN Classifier in Comparison with Medium KNN Classifier with Improved Accuracy
Manbir Kaur, Chintan Thacker, Laxmi Goswami, T. R. Thamizhvani, Imad Saeed Abdulrahman, A. Stanley Raj
Abstract
This research work aims at developing Alzheimer’s disease detection and classification using KNN classifiers. This study involves two groups and they are Weighted KNN (N=30) classifiers and Medium KNN classifiers (N=30) with sample size for each group is 30. For the purpose of calculating sample size, the pre-test power value is set at 80%, the threshold value is set at 0.05, and the confidence interval is set at 95%. The data set for Alzheimer dementia and non-dementia conditions is downloaded from kaggle.com. The classifiersperformanceis measured by precision, accuracy, and recall. Weighted KNN classifiers result in mean accuracy of 96.59 % (p=0.058) and precision of 93.68 % (p=0.178) and recall of 98.90 % (p=0.528). Medium KNN results in mean accuracy of 94.68% and precision of 90.69% and recall of 97.80%. For the identification and classification of Alzheimer’s disease, Weighted KNN classifiers outperformed Medium KNN classifiers substantially.