An Implementation of a Data-Driven Approach for Forecasting Narrow-Body Aircraft Scheduled Maintenance Workload at GMF Aero-Asia

Abstract

High volatility in GMF Aero-Asia's narrow-body aircraft scheduled maintenance workload demands better forecasting
for more effective capacity planning and operations. Unfortunately, their management used conventional forecasting
approach that resulted in a high error rate > 50 %, making it incapable of capturing the complex and dynamic patterns
of data fluctuation. In this paper, we propose an implementation and comparison of the performance of two data-driven
models, namely ARIMA (Autoregressive Integrated Moving Average) and Artificial Neural Network (ANN), to forecast
the monthly scheduled maintenance workload. By leveraging the data-driven models, we can develop a more accurate
forecasting model to improve resource scheduling efficiency, reduce planning errors, and support the transformation
towards a data-driven company. For evaluation, we used the data scheduled maintenance workload at GMF Aero-Asia
from the 2020-2024 period that implemented using two-models including ARIMA and ANN then comparing it with
conventional approach. The model performance measured using key metrics indicators. The result showed that ANN
(lag=9) method proved to be best model with lowest MAPE of 29.02%, outperforming the ARIMA (1,1,1) model around
35.38%, and the conventional approach with MAPE 55.25%. This result demonstrates that the ANN is capable of
capturing non-linear patterns in the data-driven at GMF Aero-Asia, producing smoother predictions in response to
fluctuations.