Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/166697
Title: Model Prediksi Suhu Berbasis Machine Learning pada Proses Pengukusan Bolu Lapis Kukus di PT XYZ
Other Titles: Machine Learning-Based Temperature Prediction Model for the Steaming Process of Layered Sponge Cake at XYZ Co.
Authors: Marimin
Ramadhani, Ahmad Ghifari
Issue Date: 2025
Publisher: IPB University
Abstract: Konsistensi suhu pengukusan merupakan faktor kritis dalam produksi bolu lapis kukus untuk mencegah cacat produk. PT XYZ sebagai produsen makanan terkemuka membutuhkan solusi prediktif guna mengatasi keterbatasan pendekatan konvensional yang bersifat reaktif. Penelitian ini bertujuan mengembangkan solusi untuk meningkatkan keandalan pengendalian proses pengukusan produk. Penelitian ini mengembangkan model prediksi suhu berbasis machine learning menggunakan algoritma KNN dan LSTM berdasarkan pendekatan supervised learning dengan variasi metode pengolahan data. Hasil pengujian menunjukkan model KNN-4 mencapai akurasi prediksi tertinggi dengan r²=0,93 pada data test, disertai nilai eror MSE=0,08; MAE=0,21; dan RMSE=0,29 dengan kombinasi metode pengolahan data berupa normalisasi dan pembagian data terurut. Hasil pengembangan model ensemble dengan pendekatan weighted voting regressor menunjukkan bahwa model ENS-6 tidak memberikan peningkatan kinerja (r²=0,91) dari model tunggal sehingga KNN-4 dipilih sebagai solusi akhir. Integrasi dengan sistem IoT dan validasi lebih lanjut di lingkungan produksi aktual diperlukan guna memastikan keandalan model dalam kondisi operasional nyata.
Consistent steaming temperature is a critical factor in layered sponge cake production to prevent product defects. As a leading food manufacturer, XYZ Co. requires a predictive solution to overcome the limitations of conventional reactive approaches. This study aims to develop a solution to enhance the reliability of the steaming process control for products. This study develops a machine learning-based temperature prediction model using KNN and LSTM algorithms with variations in data processing methods. Test results show that the KNN-4 model achieves the highest prediction accuracy with r²=0.93 on test data, along with error values of MSE=0.08; MAE=0.21; and RMSE=0.29, using a combination of data normalization and ordered data splitting. The development of an ensemble model with a weighted voting regressor approach indicates that the ENS-6 model does not improve performance (r²=0.91) compared to the single model, leading to the selection of KNN-4 as the final solution. Integration with an IoT system and further validation in an actual production environment are necessary to ensure model reliability under real operational conditions.
URI: http://repository.ipb.ac.id/handle/123456789/166697
Appears in Collections:UT - Agroindustrial Technology

Files in This Item:
File Description SizeFormat 
cover_F3401211106_256b1dbd59274bd2970eb86aed09cdb3.pdfCover1.93 MBAdobe PDFView/Open
fulltext_F3401211106_283f8df46e1e4569af9d1de4afb8f628.pdf
  Restricted Access
Fulltext2.53 MBAdobe PDFView/Open
lampiran_F3401211106_4482315f1f754c3db24706cf3ecaa689.pdf
  Restricted Access
Lampiran1.9 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.