APPLICATION OF ADASYN OVERSAMPLING TECHNIQUE ON K-NEAREST NEIGHBOR ALGORITHM

Abstract

The K-Nearest Neighbor Algorithm is a commonly used data mining algorithm for classification due to its effectiveness with large datasets and noise. However, class imbalance may impact classification results, where data with unbalanced classes may classify new data based on the majority class and ignore minority class data. The research analyzed whether applying the Adaptive Synthetic (ADASYN) oversampling technique in the K-Nearest Neighbor Algorithm can handle data imbalance problems. The study looks at the resulting accuracy, specificity, and sensitivity values. ADASYN oversamples the minority class data based on the model's difficulty level of data learning using distribution weights. This research uses the Pima Indian Diabetes Dataset from the Kaggle website. The dependent variable was diabetes mellitus status, while the independent variables were number of pregnancies, glucose levels, diastolic blood pressure, insulin levels, Body Mass Index (BMI), and age. The study found that the accuracy, specificity, and sensitivity values were 72.88%, 73.42%, and 71.79%, respectively. Based on the results of the analysis, it can be concluded that using ADASYN in the K-Nearest Neighbor Algorithm to classify diabetes mellitus in Pima Indian women is good enough to address imbalanced data. It is shown that the ADASYN oversampling technique can help the K-Nearest Neighbor Algorithm to classify new data without ignoring the data of the minority class.