MODEL PERAMALAN UNTUK PERENCANAAN PERSEDIAAN MATERIAL PESAWAT B737-800 DI PT GARUDA MAINTENANCE FACILITY AERO ASIA TBK MENGGUNAKAN LSTM DAN GRU

• Main Author: Kafa Atriantio, Marchiano
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77509
• Institution: Institut Teknologi Bandung

As the COVID-19 epidemy sweeps by, PT Garuda Maintenance Facility Aero Asia Tbk (GMF) was experiencing a net loss of $328 million and $127 million in 2020 and 2021 respectively. This was due to the incorrect amount of order that subsequently resulted in inventory circulation dropping the in value below one. In accordance to the aforementioned case, this research intends to develop an effective forecasting model to help expendable material planning for Boeing 737- 800 in GMF’s inventory. The research methods used in this paper involves classification of materials based on Fast, Slow, Non-Moving (FSN) analysis technique, material demand data pattern recognition through data decomposition, and implementation of Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural network model to forecast material demand. Evaluation is done with Mean Squared Error (MSE), and order method with Economic Order Quantity (EOQ) which will be utilized to determine the optimal order quantity. Research results reveal that as many as 26 materials are fast moving, and all material shows a scrambled data pattern with trends, seasonal and residual. From the tested forecasting method, LSTM is used for 5 materials, GRU for 14 materials, and Moving Average (MA) method for 7 materials. Evaluation results shows that EOQ method provides lower cost compared to actual order method. The implications from the research shows that using the suggested order method can decrease cost of material acquisition by 41% compared to actual ordering method. In other words, forecasting method with LSTM and GRU in material planning can have significant effects on material planning and cost management.