Perbandingan Peramalan Menggunakan Metode ARIMA Berbasis Prosedur dan ARIMA Otomatis Berbasis Akurasi

Mardhiyah, Shofiyah (2025) Perbandingan Peramalan Menggunakan Metode ARIMA Berbasis Prosedur dan ARIMA Otomatis Berbasis Akurasi. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111840000106-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Peramalan data penjualan obat merupakan aspek penting dalam pengelolaan stok apotek agar dapat memenuhi kebutuhan pelanggan secara optimal. Penelitian ini akan membandingkan dua pendekatan dalam peramalan deret waktu, yaitu metode ARIMA klasik berbasis prosedur statistik (Box-Jenkins) dan metode Auto ARIMA berbasis akurasi yang diotomatisasi menggunakan Python. Studi kasus dilakukan pada delapan jenis obat terlaris di Apotek Rizieq Farma berdasarkan data penjualan harian selama 77 hari. Metode ARIMA klasik meliputi tahapan identifikasi stasioneritas, penentuan orde model melalui fungsi autokorelasi (ACF) dan fungsi autokorelasi parsial (PACF), estimasi parameter, validasi residual dengan uji Ljung-Box, dan pemilihan model terbaik berdasarkan nilai Akaike’s information criterion (AIC). Sementara itu, Auto ARIMA secara otomatis memilih model terbaik berdasarkan kombinasi parameter (p,d,q) yang menghasilkan nilai AIC terendah. Hasil penelitian menunjukkan bahwa untuk data CTM dan Promag baik model ARIMA klasik maupun Auto ARIMA menghasilkan model dengan tingkat akurasi yang sama, diukur dengan AIC menghasilkan masing-masing 608,202 untuk CTM, dan 771,407 untuk Promag. Sedangkan data lainnya yaitu Paracetamol, Demacolin, AMF, Sanmol, Decolgen, dan Amlodipine memberikan hasil Auto ARIMA lebih akurat dibandingkan ARIMA klasik dengan selisih AIC secara berturutan adalah 4,266; 12,007; 10,215; 11,302; 0,028; dan 1,146. Enam data penjualan obat harian yaitu dengan nilai AIC lebih rendah pada sebagian besar obat dibandingkan pendekatan klasik, menunjukkan keunggulan dalam hal akurasi dan efisiensi waktu. Namun, pendekatan klasik tetap relevan karena memberikan kontrol dan pemahaman mendalam terhadap proses pemodelan. Dengan demikian, Auto ARIMA dapat menjadi alternatif yang efektif untuk peramalan cepat dan akurat dalam pengelolaan stok obat, terutama di lingkungan operasional yang dinamis.
=================================================================================================================================
Forecasting medicine sales data is an important aspect of pharmacy stock management to optimally meet costumer’s needs. This study will compare two approaches to time series forecasting, which is the classical ARIMA method based on the statistical procedures (Box-Jenkins) and the accuracy-based Auto ARIMA method automated using Python. The case study was conducted on eight best-selling drugs at Rizieq Farma Pharmacy based on 77 days of daily sales data.
The classical ARIMA method includes the stages of identifying stationarity, determining model order through the autocorrelation function (ACF) and partial autocorrelation function (PACF), estimating parameters, validating residuals with the Ljung-Box test, and selecting the best model based on the Akaike’s information criterion (AIC) value. Meanwhile, Auto ARIMA automatically selects the best model based on the combination of parameters (p, d, q) that produces the lowest AIC value. The research findings indicate that for CTM and Promag datasets, both classical ARIMA and Auto ARIMA models produced comparable levels of forecasting accuracy, as measured by the Akaike Information Criterion (AIC), yielding values of 608.202 for CTM and 771.407 for Promag, respectively. In contrast, datasets for Paracetamol, Demacolin, AMF, Sanmol, Decolgen, and Amlodipine demonstrated superior accuracy with Auto ARIMA, with sequential AIC improvements over classical ARIMA of 4.266, 12.007, 10.215, 11.302, 0.028, and 1.146. Across these six daily pharmaceutical sales datasets, the consistently lower AIC values observed under the Auto ARIMA approach suggest advantages in terms of forecasting accuracy and computational efficiency. Nevertheless, the classical ARIMA model remains relevant due to its transparency and greater interpretability of the modeling process. Accordingly, Auto ARIMA may serve as an effective alternative for fast and accurate demand forecasting in dynamic operational environments, particularly for pharmaceutical inventory management.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	ARIMA, Auto ARIMA, Runtun Waktu, Peramalan, AIC, ARIMA, Auto ARIMA, Time Series, Forecasting, AIC.
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting T Technology > T Technology (General) > T58.8 Productivity. Efficiency
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Shofiyah Mardhiyah
Date Deposited:	01 Aug 2025 01:32
Last Modified:	01 Aug 2025 01:32
URI:	http://repository.its.ac.id/id/eprint/123346

Actions (login required)

View Item