Peramalan Penjualan Motor Niaga Roda Tiga Menggunakan Metode Dekomposisi STL dan Regresi Lasso Dengan Indikator Makroekonomi

Prasetya, Agung Budi (2026) Peramalan Penjualan Motor Niaga Roda Tiga Menggunakan Metode Dekomposisi STL dan Regresi Lasso Dengan Indikator Makroekonomi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026221176-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

PT Surya Mas Citra Niaga, sebagai salah satu diler utama motor niaga roda tiga Viar di Jawa Timur, menghadapi tantangan dalam memprediksi permintaan secara akurat. Penelitian ini mengembangkan model peramalan berbasis dekomposisi STL (Seasonal-Trend decomposition based on Loess) dan regresi LASSO yang mengintegrasikan lima indikator makroekonomi nasional sebagai leading indicators. Komponen tren diregresikan terhadap fitur lag makroekonomi dan autoregresif menggunakan LASSO, sedangkan komponen musiman diprediksi menggunakan metode seasonal naive. Kinerja model dievaluasi menggunakan metrik MAPE, RMSE, MAE, dan MASE (Mean Absolute Scaled Error), serta dibandingkan terhadap varian STL-LASSO tanpa indikator makroekonomi, Regresi Linear, Prophet, dan Exponential Smoothing Holt-Winters. STL-LASSO mencatat MASE terendah atau setara terendah pada ketiga seri periode uji (0,559 pada New Karya 150, 0,562 pada New Karya 200, dan 0,500 pada New Karya BIT), mengungguli seluruh model pembanding. LASSO mengidentifikasi uang beredar dalam arti sempit M1, Produk Domestik Bruto, dan nilai tukar sebagai indikator paling berpengaruh terhadap komponen tren, dengan momentum penjualan bulan sebelumnya konsisten terpilih sebagai prediktor signifikan di ketiga seri. Performa yang kurang optimal pada seri dengan perubahan struktural pada periode uji dijustifikasi melalui pengujian tambahan pada data IHSG, yang menghasilkan MASE 0,393, mengonfirmasi bahwa keterbatasan tersebut berasal dari kondisi data, bukan kelemahan model.
===============================================================================================================================
PT Surya Mas Citra Niaga, as one of the main dealers of Viar three-wheeled commercial vehicles in East Java, faces challenges in accurately predicting demand, leading to inventory inefficiencies. This study develops a forecasting model based on STL (Seasonal-Trend decomposition based on Loess) decomposition and LASSO regression that integrates five national macroeconomic indicators as leading indicators. The trend component is regressed against macroeconomic lag and autoregressive features using LASSO, while the seasonal component is predicted using the seasonal naive method. Model performance is evaluated using MAPE, RMSE, MAE, and MASE (Mean Absolute Scaled Error) metrics, and compared against a STL-LASSO variant without macroeconomic indicators, Linear Regression, Prophet, and Exponential Smoothing Holt-Winters. STL-LASSO achieves the lowest or tied-lowest MASE across all three series on the test period (0.559 on New Karya 150, 0.562 on New Karya 200, and 0.500 on New Karya BIT), outperforming all comparison models. LASSO identifies money supply (M1), Gross Domestic Product, and exchange rate as the most influential indicators on the trend component, with the previous month's sales momentum consistently selected as a significant predictor across all series. The suboptimal performance on series affected by structural shifts in the test period is justified through an additional test on IHSG data, which yields a MASE of 0.393, confirming that this limitation stems from data conditions rather than a weakness in the model.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Peramalan Permintaan, Motor Niaga Roda Tiga, Dekomposisi STL, Regresi LASSO, Indikator Makroekonomi, : Demand Forecasting, Three-Wheeled Commercial Vehicle, STL Decomposition, LASSO Regression, Macroeconomic Indicators
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Agung Budi Prasetya
Date Deposited:	29 Jul 2026 03:17
Last Modified:	29 Jul 2026 03:17
URI:	http://repository.its.ac.id/id/eprint/138980

Actions (login required)

View Item