Implementasi Principal Component Analysis (PCA) dan Gap Statistic untuk Clustering Kanker Payudara pada Algoritma K-Means

Abstract

Breast cancer is one of the most common causes of death worldwide. Data mining can be utilized to detect breast cancer, where information is extracted from data to provide valuable insights. Clustering of breast cancer is conducted to assist medical professionals in grouping the characteristics of each cancer type. However, multicollinearity in breast cancer data can impact clustering results. To address this issue, dimensionality reduction through Principal Component Analysis (PCA) is employed. PCA can effectively handle multicollinearity issues and enhance computational efficiency. Additionally, the K-Means method has limitations in determining the optimal number of clusters. Therefore, the Gap Statistic method is employed to find the optimal K value suitable for breast cancer data. This study compares the evaluation results of the K-Means clustering model, the combined PCA-KMeans clustering model, and the combined PCA-GapStatistic-KMeans clustering model. The findings indicate that the evaluation results for the K-Means model with PCA dimensionality reduction and optimal Gap Statistic K are superior to the K-Means model without dimensionality reduction. The Gap Statistic suggests 2 clusters as the optimal number, with an evaluation result of 1.195513.