Pendeteksian Anomali Pada Kasus Material KRL KCI Menggunakan Metode Isolation Forest

Pangayom, Danang Luhur (2025) Pendeteksian Anomali Pada Kasus Material KRL KCI Menggunakan Metode Isolation Forest. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002211073_Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Salah satu tantangan dalam produksi KRL adalah memastikan material dan sistem berada dalam kondisi optimal untuk menghindari gangguan operasional serta potensi risiko keselamatan. Dalam industri manufaktur, anomali material dapat berupa pengendalian material yang tidak tepat, kesalahan perhitungan komponen, atau penyimpangan dari spesifikasi teknis yang dapat berdampak negatif terhadap performa dan daya tahan kereta. Penelitian ini menerapkan metode Isolation Forest sebagai teknik machine learning untuk mendeteksi anomali pada data material KRL, dengan didahului oleh proses reduksi dimensi menggunakan Principal Component Analysis (PCA). Hasil penelitian menunjukkan bahwa kombinasi PCA dan Isolation Forest efektif dalam mengidentifikasi anomali, dengan mendeteksi 8 komponen berjenis steel plate (0,65% dari total data) sebagai outlier. Visualisasi hasil proyeksi dua dimensi menunjukkan pemisahan yang cukup jelas antara data normal dan data anomali. Skor anomali yang dihasilkan juga menunjukkan bahwa material dengan skor lebih rendah cenderung memiliki penyimpangan terhadap standar kualitas. Berdasarkan evaluasi dengan confusion matrix, model Isolation Forest memiliki nilai akurasi sebesar 93,44%, precision 62,50%, recall 35,71%, dan F1-score 45,45%. Hal ini menunjukkan bahwa meskipun Isolation Forest bersifat konservatif dalam mendeteksi anomali, metode ini unggul dalam menghasilkan prediksi yang akurat dan minim kesalahan positif palsu. Dengan demikian, pendekatan ini dinilai relevan dalam mendukung sistem kontrol kualitas material dan berpotensi meningkatkan efisiensi serta keselamatan dalam proses produksi KRL di Indonesia.
====================================================================================================================================
One of the key challenges in KRL (commuter rail) manufacturing is ensuring that materials and systems remain in optimal condition to avoid operational disruptions and potential safety risks. In the railway manufacturing industry, material anomalies may occur due to improper material handling, miscalculation of components, or deviations from technical specifications, all of which can negatively affect the performance and durability of the train. This study implements the Isolation Forest method, a machine learning technique, to detect anomalies in EMU material data, preceded by dimensionality reduction using Principal Component Analysis (PCA). The results show that the combination of PCA and Isolation Forest is effective in identifying anomalies, detecting 8 components of the steel plate type (0.65% of the total data) as outliers. The two-dimensional projection visualization shows a clear separation between normal data and anomalies. The anomaly scores also indicate that materials with lower scores tend to deviate from quality standards. Based on evaluation using a confusion matrix, the Isolation Forest model achieved an accuracy of 93.44%, a precision of 62.50%, a recall of 35.71%, and an F1-score of 45.45%. These results indicate that although Isolation Forest is conservative in detecting anomalies, it excels in producing accurate predictions with minimal false positives. Therefore, this approach is considered relevant to support material quality control systems and has the potential to improve efficiency and safety in the KRL (commuter rail) production process in Indonesia.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kereta Rel Listrik, Deteksi Anomali, PCA, Machine Learning, Isolation, Forest, Kualitas Material, Confusion Matrix, Electric Train, Anomaly Detection, Machine Learning, Isolation, Material Quality
Subjects:	Q Science > QA Mathematics Q Science > QA Mathematics > QA75 Electronic computers. Computer science. EDP
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Danang Luhur Pangayom
Date Deposited:	22 Jan 2026 07:16
Last Modified:	22 Jan 2026 07:16
URI:	http://repository.its.ac.id/id/eprint/129518

Actions (login required)

View Item