Peramalan Disbursement Intermiten Belanja Anggaran Investasi Non-Rutin Berbasis TSB – CROSTON - ADIDA dengan Rekonsiliasi Hierarkis: Studi Kasus Disbursement PLN UIP Sulawesi

Poluan, Pricilia Maureen (2026) Peramalan Disbursement Intermiten Belanja Anggaran Investasi Non-Rutin Berbasis TSB – CROSTON - ADIDA dengan Rekonsiliasi Hierarkis: Studi Kasus Disbursement PLN UIP Sulawesi. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032241067-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Pola realisasi disbursement belanja investasi non-rutin pada PT PLN (Persero) Unit Induk Pembangunan (UIP) Sulawesi bersifat intermiten, ditandai dominasi periode tanpa realisasi yang diselingi lonjakan pembayaran terkonsentrasi pada akhir tahun. Periode nol ini umumnya bersumber dari kendala administratif dan lapangan proses pengadaan lahan, perizinan, serta kelengkapan dokumen pembayaran dan bukan ketiadaan kebutuhan riil, sehingga menimbulkan ketidakakuratan perencanaan kas dan idle cash akibat over-forecasting dalam penyusunan Anggaran Kas Bulanan (AKB). Penelitian ini mengevaluasi tiga metode Intermittent Demand Forecasting (IDF) yaitu Croston, Teunter-Syntetos-Babai (TSB), dan Aggregate-Disaggregate Intermittent Demand Approach (ADIDA) pada data disbursement bulanan periode 2021–2025 dengan struktur hierarki tiga level (UIP–UPP–PRK). Klasifikasi pola berdasarkan Average Demand Interval (ADI) dan koefisien variasi (CV) menunjukkan dominasi kategori intermittent dan lumpy, membenarkan penggunaan kerangka IDF. Evaluasi menerapkan pendekatan tiga tahap analitis sebagai triangulasi bukti: validasi in-sample (29 PRK), evaluasi out-of-sample dengan rolling origin (180 PRK), dan production forecast 2026 (959 PRK), menggunakan metrik MASE, MAE, dan ME. Koherensi prakiraan lintas level dijamin melalui rekonsiliasi Weighted Least Squares (WLS) yang dipilih atas pertimbangan stabilitas estimasi pada data intermiten, dengan Bottom-Up (BU) dan Top-Down (TD) sebagai pembanding. Hasil penelitian mengidentifikasi ADIDA sebagai metode IDF terbaik dengan dominasi win-rate konsisten (58,6% pada Tahap 1 dan 64,2% pada Tahap 2) serta efisiensi mendekati batas teoretis Oracle, didukung keunggulan robustness dan kecepatan komputasi. Rekonsiliasi WLS menghasilkan total prakiraan disbursement 2026 sebesar Rp 2.781,89 Miliar yang koheren di seluruh level dan terdistribusi proporsional antar-UPP. Pada validasi hold-out 2025, WLS terbukti lebih akurat dibanding BU (MASE 1,244 vs 1,473) dengan total lebih konservatif sehingga menekan potensi idle cash, serta menawarkan trade-off operasional yang menguntungkan dibanding TD. Temuan ini memberikan landasan empiris bagi perencanaan kas berbasis data dan mendukung kesiapan implementasi SAP S/4HANA Single Source of Truth (SSoT) di lingkungan PLN UIP Sulawesi.
====================================================================================================================================
The disbursement realization pattern of non-routine investment expenditure at PT PLN (Persero) Sulawesi Main Project Unit (UIP) is intermittent, characterized by a predominance of zero-realization periods interspersed with payment surges concentrated at year-end. These zero periods generally stem from administrative and field constraints — land acquisition, permits, and payment document completeness — rather than the absence of actual demand, causing cash planning inaccuracy and idle cash due to over-forecasting in preparing the Monthly Cash Budget (AKB). This study evaluates three Intermittent Demand Forecasting (IDF) methods — Croston, Teunter-Syntetos-Babai (TSB), and Aggregate-Disaggregate Intermittent Demand Approach (ADIDA) — using monthly disbursement data for the 2021–2025 period within a three-level hierarchy (UIP–UPP–PRK). Pattern classification based on the Average Demand Interval and the coefficient of variation reveals a predominance of intermittent and lumpy categories, justifying the IDF framework. The evaluation employs a three-stage analytical approach as evidence triangulation: in-sample validation (29 PRK), out-of-sample evaluation via rolling origin (180 PRK), and 2026 production forecast (959 PRK), assessed using MASE, MAE, and ME metrics. Forecast coherence across organizational levels is ensured through Weighted Least Squares (WLS) reconciliation, selected for its estimation stability on intermittent data, with Bottom-Up (BU) and Top-Down (TD) as comparators. The results identify ADIDA as the best IDF method, demonstrating consistent win-rate dominance (58.6% in Stage 1 and 64.2% in Stage 2) and efficiency approaching the theoretical Oracle bound, supported by superior robustness and computational speed. WLS reconciliation yields a coherent total 2026 disbursement forecast of IDR 2,781.89 billion, proportionally distributed across UPPs. In the 2025 hold-out validation, WLS proves more accurate than BU (MASE 1.244 vs 1.473), producing a more conservative total that suppresses potential idle cash, while offering an operational trade-off favorable to TD. These findings provide an empirical foundation for data-driven cash planning and support readiness for the SAP S/4HANA Single Source of Truth (SSoT) implementation at PLN UIP Sulawesi.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	disbursement intermiten, peramalan permintaan intermiten, perencanaan kas, rekonsiliasi hierarkis, Weighted Least Squares.
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HD Industries. Land use. Labor > HD30.27 Business forecasting
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Pricilia Maureen Poluan
Date Deposited:	29 Jul 2026 02:25
Last Modified:	29 Jul 2026 02:25
URI:	http://repository.its.ac.id/id/eprint/139311

Actions (login required)

View Item