Deteksi Anomali Pada Produksi Crude Palm Oil (CPO) Menggunakan Diagram Kendali EWMA Berbasis Residual Model Temporal Convolutional Network (TCN)

Fahriyah, Adristy Rizki (2025) Deteksi Anomali Pada Produksi Crude Palm Oil (CPO) Menggunakan Diagram Kendali EWMA Berbasis Residual Model Temporal Convolutional Network (TCN). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003211082-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Crude Palm Oil (CPO) merupakan komoditas utama dalam pasar minyak nabati global, dan Indonesia memiliki peran strategis dalam produksinya. Namun, sejak 2020, produksi CPO mengalami fluktuasi signifikan yang memerlukan analisis lebih lanjut. Penelitian ini bertujuan mendeteksi anomali produksi CPO di Indonesia serta mengidentifikasi penyebabnya menggunakan diagram kendali Exponentially Weighted Moving Average (EWMA) berbasis residual dari model Temporal Convolutional Network (TCN). Data yang digunakakn adalah produksi bulanan Januari 2007–Desember 2023 dibagi menjadi dua fase, yaitu Fase I untuk pelatihan dan Fase II untuk pengujian. Hasil menunjukkan bahwa model TCN dua lapisan berhasil mendeteksi anomali pada Fase I, khususnya pada bulan Juli, Oktober, dan November 2018 serta Januari, Mei, Agustus, dan September 2019, yang dikategorikan sebagai lonjakan sesaat (spike) akibat faktor seperti La Niña, perluasan lahan, dan kebijakan B20/B30. Pada Fase II, sebagian besar residual tetap berada di luar batas kendali tanpa titik spesifik yang dominan, mengindikasikan perubahan pola produksi signifikan yang dipengaruhi pandemi COVID-19 dan dinamika global. Model TCN menunjukkan keunggulan dibanding ARIMA dalam mengikuti pola baru, dan diagram EWMA terbukti lebih sensitif dalam mendeteksi perubahan pola secara bertahap.
=============================================================================================================================================
Crude Palm Oil (CPO) is one of the main commodities dominating the global vegetable oil market, with Indonesia playing a strategic role in its supply. However, since 2020, CPO production has shown significant fluctuations, requiring further analysis to identify these changes. This study aims to detect anomalies in Indonesia’s CPO production and identify their causes using an Exponentially Weighted Moving Average (EWMA) control chart based on residuals from a Temporal Convolutional Network (TCN) model. Monthly production data from January 2007 to December 2023 is divided into two phases: Phase I for model training and Phase II for testing. The results show that the two-layer TCN model successfully detected anomalies in Phase I, particularly in July, October, and November 2018 as well as January, May, August, and September 2019. These anomalies are classified as sudden spikes, caused by factors such as La Niña, land expansion, and B20/B30 policy interventions. In Phase II, most residuals remained outside the control limits without dominant specific points, indicating significant changes in production patterns driven by the COVID-19 pandemic and global market instability. Compared to ARIMA, the TCN model performed better in capturing new patterns, while the EWMA chart proved more sensitive in identifying gradual shifts in the production process.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Produksi Crude Palm Oil, Deteksi Anomali, Temporal Convolutional Network (TCN), Fluktuasi Produksi CPO, Diagram Kendali EWMA Crude Palm Oil Production, Anomaly Detection, Temporal Convolutional Network (TCN), CPO Production Fluctuations, EWMA Control Chart
Subjects:	H Social Sciences > HD Industries. Land use. Labor H Social Sciences > HD Industries. Land use. Labor > HD9980.5 Service industries--Quality control.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Adristy Rizki Fahriyah
Date Deposited:	31 Jul 2025 08:42
Last Modified:	31 Jul 2025 08:42
URI:	http://repository.its.ac.id/id/eprint/124980

Actions (login required)

View Item