Analisis Prediksi Persediaan Bahan Baku Bijih Plastik Menggunakan Kombinasi Kausalitas Dan Deret Waktu : Studi Kasus Di Perusahaan Industri Tekstil

Rawung, Frangky (2024) Analisis Prediksi Persediaan Bahan Baku Bijih Plastik Menggunakan Kombinasi Kausalitas Dan Deret Waktu : Studi Kasus Di Perusahaan Industri Tekstil. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025211001-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (3MB) | Request a copy

Abstract

Manajemen persediaan bahan baku bijih plastik menjadi tantangan penting bagi perusahaan industri tekstil untuk menjamin kelancaran produksi dan mengurangi biaya operasional. Ketepatan jumlah dan waktu pengadaan bahan baku esensial dalam memastikan efisiensi dan pengurangan biaya. Namun, prediksi kebutuhan bahan baku sering terhambat oleh fluktuasi permintaan pasar dan ketidakpastian pasokan serta tidak adanya stok minimal bahan baku yang sesuai dengan persediaan. Mengingat kesulitan ini, penelitian ini dirancang untuk mengembangkan model prediksi persediaan bijih plastik yang memanfaatkan pendekatan kausalitas dan analisis deret waktu. Model ini bertujuan untuk mengatasi ketidakpastian pasokan dan permintaan, memastikan stok bahan baku optimal, mengurangi risiko kekurangan stok, dan mengoptimalkan biaya persediaan di industri tekstil. Penelitian ini berfokus pada pengembangan model deret waktu dengan menerapkan pendekatan Bidirectional Long Short-Term Memory (BiLSTM) dan pengembangan model kausalitas menggunakan pendekatan Multiple Linear Regression (MLR). Harmonic mean digunakan untuk menggabungkan hasil prediksi dari kedua model tersebut. Penelitian mengenai BiLSTM pada bahan baku PP menunjukkan bahwa Root Mean Square Error (RMSE) adalah 18.95 dan R-squared (R²) adalah 0.91. Di sisi lain, model MLR menghasilkan RMSE 39.69 dan R² 0.76. Hasil kombinasi BiLSTM dan MLR menunjukkan RMSE 25.65 dan R² 0.83. Penelitian ini juga melakukan perbandingan nilai RMSE dan R² untuk bahan baku MB dan PET. Untuk bahan baku MB, BiLSTM menghasilkan RMSE 3.82 dan R² 0.80, sedangkan MLR memberikan RMSE 39.69 dan R² 0.76. Kombinasi BiLSTM dan MLR menghasilkan RMSE 6.97 dan R² 0.78. Pada bahan baku PET, BiLSTM menghasilkan RMSE 6.53 dan R² 0.93, sedangkan MLR menghasilkan RMSE 6.91 dan R² 0.91. Kombinasi kedua model ini menghasilkan RMSE 6.71 dan R² 0.92. Hasil ini menunjukkan bahwa bahan baku PET memiliki nilai yang lebih tinggi dibandingkan dengan bahan baku lain, baik pada model BiLSTM, MLR, maupun kombinasi BiLSTM dan MLR.
=================================================================================================================================
The management of raw plastic ore inventory becomes a crucial challenge for textile industry companies to ensure smooth production and reduce operational costs. The accuracy of the quantity and timing of raw material procurement is essential in ensuring efficiency and cost reduction. However, predictions of raw material needs are often hindered by market demand fluctuations, supply uncertainty, and the absence of a minimal raw material stock that aligns with inventory. Considering these difficulties, this research is designed to develop a plastic ore inventory prediction model that utilizes causality approaches and time series analysis. This model aims to address supply and demand uncertainties, ensure optimal raw material stock, reduce the risk of stock shortages, and optimize inventory costs in the textile industry. This research focuses on developing a time series model by applying the Bidirectional Long Short-Term Memory (BiLSTM) approach and a causality model using the Multiple Linear Regression (MLR) approach. The harmonic mean is used to combine the prediction results from both models. Research on BiLSTM for PP raw materials shows that the Root Mean Square Error (RMSE) is 18.95, and the R-squared (R²) value is 0.91. Conversely, the MLR model yields an RMSE of 39.69 and an R² of 0.76. The combination of BiLSTM and MLR models shows an RMSE of 25.65 and an R² of 0.83. This study also compares the RMSE and R² values for MB and PET raw materials. For MB raw materials, BiLSTM produced an RMSE of 3.82 and an R² of 0.80, while MLR yielded an RMSE of 39.69 and an R² of 0.76. The combination of BiLSTM and MLR resulted in an RMSE of 6.97 and an R² of 0.78. For PET raw materials, BiLSTM produced an RMSE of 6.53 and an R² of 0.93, whereas MLR produced an RMSE of 6.91 and an R² of 0.91. The combination of these two models resulted in an RMSE of 6.71 and an R² of 0.92. These results indicate that PET raw materials have higher values compared to other materials, in both the BiLSTM model, the MLR model, and their combination.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Persediaan Bahan Baku, Prediksi, Kausalitas, Deret Waktu, Machine Learning, Raw Material Inventory, Prediction, MLR, BiLSTM, Causality, Time Series
Subjects:	T Technology > T Technology (General) > T174 Technological forecasting
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Frangky Rawung
Date Deposited:	28 Feb 2024 02:42
Last Modified:	28 Feb 2024 02:42
URI:	http://repository.its.ac.id/id/eprint/107724

Actions (login required)

View Item