Peramalan Permintaan Teh Serbuk Menggunakan Metode Support Vector Machine (Studi Kasus di PT. Dua Sahabat Mendunia)

Putu Yunita, Seliyanti and Wike Agustin Prima Dania,, STP.M.Eng.Ph.D and Ir. Usman Effendi,, MS (2023) Peramalan Permintaan Teh Serbuk Menggunakan Metode Support Vector Machine (Studi Kasus di PT. Dua Sahabat Mendunia). Sarjana thesis, Universitas Brawijaya.

Abstract

Tanaman teh (Camellia sinensis.L) merupakan salah bahan baku minuman yang paling populer, salah satu jenis olahannya adalah teh hitam yang bahkan dianggap sebagai minuman pengganti air mineral. Ditinjau dari bentuk sediaannya, jenis teh yang populer adalah teh celup dan teh serbuk sebab lebih fleksibel. PT. Dua Sahabat Mendunia, merupakan salah satu produsen minuman instan dengan produk andalan berupa teh serbuk varian melati (DS Jasmine tea). Penjadwalan produksi yang didasarkan atas sistem pre-order membatasi perusahaan dalam menjangkau pasar yang lebih luas dikarenakan sistem yang kurang fleksibel dan sulit memenuhi permintaan yang mendadak. Berdasarkan permasalahan tersebut, penelitian ini bertujuan untuk melakukan peramalan permintaan DS Jasmine Tea. Metode peramalan permintaan DS Jasmine Tea untuk periode 2023 yang digunakan dalam penelitian ini adalah metode support vector machine dan time series. Metode support vector machine dipilih untuk mengatasi kekurangan metode time series yang sulit digunakan pada data dengan karakteristik acak, sebab model training lebih akurat dan stabil. Peramalan permintaan DS Jasmine tea dilakukan dengan input data interval harian, mingguan dan bulanan. perbandingan dilakukan untuk metode time series dan support vector machine dengan input data harian dan bulanan. Penelitian dilakukan pada bulan Januari 2023 hingga Mei 2023. Tahap peramalan dengan metode support vector machine meliputi preprocessing, pelatihan dan pengujian. Hasil peramalan terbaik diperoleh untuk input data bulanan baik untuk metode time series (simple seasonal) dengan nilai MAPE 27,179% maupun support vector machine dengan nilai MAPE data training 14,99% dan data testing 19,51%. Hal ini menunjukkan hasil peramalan dengan support vector machine memiliki tingkat akurasi baik (MAPE 10-20%), dikarenakan metode support vector machine memiliki kemampuan pembacaan pola atau pembelajaran yang lebih baik dibanding metode time series. Pada penelitian selanjutnya disarankan untuk mempertimbangkan input variabel biaya distribusi maupun biaya promosi/ frekuensi pengiklanan, serta mempertimbangkan metode lain seperti jaringan saraf tiruan atau random forest. Bagi perusahaan, dapat menggunakan hasil penelitian ini sebagai acuan untuk perencanaan produksi.

English Abstract

Tea (Camellia sinensis L.), particularly black tea, is a popular beverage and is often used as a mineral water substitute. Teabags and powdered tea are two of the most popular ways to serve and package tea since they are more versatile. PT. Dua Sahabat Mendunia is one of the producers of jasmine tea (DS Jasmine Tea). The company's production scheduling is based on a pre-order system. Because the system is less flexible and makes it difficult to accommodate unanticipated requests, it prevents the company from accessing a larger market. Based on these issues, this study is aimed at forecasting demand for DS Jasmine Tea. The support vector machine and time series methods were used in this work to forecast demand for DS Jasmine Tea. Because the training model is more precise and stable, the support vector machine approach was chosen to solve the limitations of the time series method, which is difficult to utilize on data with random features. Demand for DS Jasmine Tea is forecasted by inputting data at daily, weekly, and monthly intervals with year of 2023 as target output. Comparisons were made between the time series approach and the support vector machine. This study was done from January 2023 until May 2023. The forecasting stage using the support vector machine method includes preprocessing, training, and testing. The best forecasting results were obtained for input monthly data for both the time series method (simple seasonal) with a MAPE value of 27.179% and a support vector machine with a MAPE value of 14,99% for data training and 19.51% for data testing. This shows that the results of forecasting with the support vector machine have a good level of accuracy (10-20% MAPE), which also showed that support vector machine method has better forecasting accuracy due to better pattern reading or learning ability than the time series method. Future studies are advised to considering other variable inputs such as distribution costs, promotion costs/advertising frequency, as well as other methods such as artificial neural networks or random forests. Companies can use the results of this study as a reference for production planning.

Item Type:	Thesis (Sarjana)
Identification Number:	052310
Uncontrolled Keywords:	Peramalan Permintaan, Support Vector Machine, Time Series Demand Forecasting, Support Vector Machine, Time Series
Divisions:	Fakultas Teknologi Pertanian > Teknologi Industri Pertanian
Depositing User:	Unnamed user with username verry
Date Deposited:	17 Jan 2024 01:59
Last Modified:	17 Jan 2024 01:59
URI:	http://repository.ub.ac.id/id/eprint/211236

Text (Dalam Masa Embargo) PUTU YUNITA SELIYANTI.pdf Restricted to Registered users only until 31 December 2025. Download (2MB)

Actions (login required)

View Item