Desain Optimalisasi Peramalan Suku Cadang Berbasis Machine Learning
Abstract
Di era perkembangan revolusi industri 4.0 dan era digitalisasi saat ini, persaingan dunia bisnis kian sangat ketat dan sengit. Perusahaan mulai berlomba-lomba menciptakan strategi agar konsumen tidak berpindah ke kompetitor lain. Salah satu strategi agar konsumen tidak berpindah adalah dengan melakukan peramalan suku cadang, karena dengan adanya peramalan maka kegiatan produksi maupun jasa yang ada di perusahaan akan terus berjalan, sehingga berdampak kepada kepuasan pelanggan. Seiring dengan perkembangan zaman, untuk memenangkan persaingan, strategi yang dilakukan perusahaan juga bervariasi, seperti penerapan machine learning yang banyak dilakukan perusahaan besar. Penelitian ini memiliki tujuan untuk mendesain optimalisasi model peramalan dengan menggunakan machine learning. Penelitian ini membandingkan enam jenis model peramalan dengan dataset yang sudah disediakan. Model peramalan dapat dikatakan terbaik, apabila memiliki nilai MAE (Mean Absolute Error) yang terendah. Hasil penelitian menunjukkan model peramalan yang terbaik adalah model hybrid, yaitu model Optimize Neural Network (Autoencoder + LSTM), yang memiliki nilai MAE 0,5528.
In the era of the development of the Industrial Revolution 4.0 and the current era of digitalization, competition in the business world is becoming very tight and fierce. Companies started to compete to create strategies so that consumers do not move to other competitors. One of the strategies to keep consumers from moving is to forecast spare parts, because with forecasting, the production and service activities in the company will continue to run, which has an impact on customer satisfaction. Along with the times, to win the competition, the strategies carried out by companies also vary, such as the application of machine learning that many large companies do. This research aims to design a forecasting model optimization using machine learning. This research compares six types of forecasting models with the provided data set. The forecasting model can be said to be the best if it has the lowest MAE (Mean Absolute Error) value. The results show that the best forecasting model is a hybrid model, namely the Optimize Neural Network (Autoencoder + LSTM) model, which has an MAE value of 0.5528.
Downloads
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.