Model Machine Learning Untuk Klasifikasi Persediaan Material Pada Industri Pengolahan Gas Alam

Elian, Endra Hilman (2025) Model Machine Learning Untuk Klasifikasi Persediaan Material Pada Industri Pengolahan Gas Alam. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5010211014_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Perusahaan objek amatan mengalami beberapa masalah seperti tingginya jumlah material yang mengalami understock, overstock, dan potential dead stock, yang menunjukkan bahwa kebijakan Min-Max yang digunakan perusahaan tidak lagi mencerminkan kebutuhan aktual. Nilai batas persediaan tidak diperbarui secara berkala, sementara karakteristik permintaan dan lead time terus berubah. Selain itu, tidak adanya klasifikasi material yang sistematis menyulitkan perusahaan dalam menetapkan prioritas pengendalian stok. Penelitian ini menerapkan analisis klasifikasi ABC, pola permintaan, dan pola pergerakan dengan mempertimbangkan perbedaan antara suku cadang consumable dan non-consumable. Metode peramalan disesuaikan dengan pola permintaan masing-masing, seperti penggunaan metode Single Exponential Smoothing untuk pola smooth dan erratic, serta Croston TSB untuk pola lumpy. Selanjutnya, dilakukan pelatihan model klasifikasi status stok menggunakan XGBoost berdasarkan fitur-fitur seperti inventory level, lead time, dan forecast. Hasil menunjukkan bahwa sebagian besar material berada dalam kategori lumpy dan non-moving, serta dominasi potential dead stock pada kedua jenis suku cadang. Model XGBoost menghasilkan akurasi terbaik pada kategori understock, dengan akurasi keseluruhan mencapai 81.75%. Model ini dapat digunakan sebagai alat pendukung pengambilan keputusan yang objektif dan berbasis data, untuk membantu mengevaluasi kondisi stok dan mengidentifikasi kebutuhan penyesuaian tanpa menggantikan kebijakan Min-Max.
===================================================================================================================================
The observed company faces several issues, including a high number of materials experiencing understock, overstock, and potential dead stock, indicating that the current Min-Max policy no longer reflects actual needs. Stock level thresholds are not updated regularly, while demand patterns and lead times continue to change. Furthermore, the absence of a systematic material classification complicates the prioritization of inventory control. This study implements ABC classification, demand pattern analysis, and movement pattern classification by considering the differences between consumable and non-consumable spare parts. Forecasting methods are tailored to each demand pattern, such as the use of Single Exponential Smoothing for smooth and erratic patterns, and Croston TSB for lumpy patterns. Furthermore, a stock status classification model is trained using XGBoost based on features such as inventory level, lead time, and forecast. The results show that most materials fall into the lumpy and non-moving categories, with a dominance of potential dead stock in both consumable and non-consumable items. The XGBoost model achieved the best performance in identifying understock conditions, with an overall accuracy of 81.75%. This model can be used as a data-driven decision support tool to evaluate stock conditions and identify adjustment needs without replacing the existing Min-Max policy.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Analisis ABC, Continuous Review (s,Q) dan (s,S), Forecasting, Klasifikasi ADI-CV, Klasifikasi Pola Pergerakan, K-Means Clustering, Machine Learning, XGBoost, ABC Analysis, ADI-CV Classification, Continuous Review (s,Q) and (s,S), Forecasting, K-Means Clustering, Machine Learning, Movement Pattern Classification, XGBoost
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD38.5 Business logistics--Cost effectiveness. Supply chain management. ERP Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > TS Manufactures > TS161 Materials management.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Endra Hilman Elian
Date Deposited:	22 Jul 2025 06:01
Last Modified:	22 Jul 2025 06:01
URI:	http://repository.its.ac.id/id/eprint/120468

Actions (login required)

View Item