Classification Analysis of 2 K-Nearest Neighbor (KNN) and Decision Tree Algorithms Using Rapidminer in Breast Cancer

Abstrak

Cancer remains the primary cause of mortality in developed nations and ranks second in developing countries. According to 2018 data from the Global Cancer Observatory, Indonesia recorded a cancer incidence rate of 136.2 per 100,000 people, placing it 8th in Southeast Asia and 23rd across Asia. Among women, breast cancer had the highest incidence rate at 42.1 per 100,000, with a mortality rate averaging 17 per 100,000, while cervical cancer followed at 23.4 per 100,000 incidence and 13.9 per 100,000 mortality. Many patients assume that once breast cancer is no longer detectable in the body, it will not return yet recurrence is a real possibility that remains poorly understood by the public. Breast cancer is among the most common cancers affecting Indonesians, especially women, and is generally classified into two forms: benign (non-cancerous) and malignant (cancerous). This study aims to identify an accurate algorithm for analyzing breast cancer datasets, generate insights into patterns or models that support early detection, and compare the performance of two classification algorithms Decision Tree C4.5 and K-Nearest Neighbor (KNN) in classifying breast cancer cases. The research follows the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. The central question guiding this study is which of the two algorithms delivers higher classification accuracy and greater utility for identifying patterns relevant to early breast cancer detection. Findings from the CRISP-DM analysis revealed that KNN outperformed the Decision Tree C4.5 model, achieving an accuracy of 97.14% and an AUC score of 0.976, indicating excellent classification performance (AUC values between 0.90 and 1.00 are considered very good).

Identifier

• DOI: https://doi.org/10.59890/ijels.v4i7.44
• OpenAlex: https://openalex.org/W7172286782