Sistem Peramalan Jumlah Penumpang di Bandara Banyuwangi Menggunakan Metode LSTM Berbasis Website

Abstract

Indonesia is an archipelagic country that relies on air transportation as an efficient means of connecting regions. However, the Covid-19 pandemic caused a significant decline in passenger numbers and the suspension of thousands of aircraft operations. Banyuwangi Airport was among the affected airports, experiencing a drastic decrease in passenger numbers in 2018–2019 and losing its international status based on the Ministry of Transportation Decree No. 31/2024. The data used in this study consist of passenger arrivals and departures from 2011 to 2022 obtained from the Central Bureau of Statistics (BPS). The research stages include data preprocessing Preprocessing was carried out through normalization and splitting the data into two parts, namely training data and testing data., model development using the Long Short-Term Memory (LSTM) method, and evaluation using Mean Absolute Percentage Error (MAPE) and Root Mean Squared Error (RMSE). Hyperparameter optimization was carried out using ParameterGrid to obtain the best combination of values for epoch, batch size, and units. The LSTM model with 400 epochs, batch size of 16, and 16 units achieved the best performance with a MAPE value of 0.1693 and an RMSE value of 741.4489. The model was implemented into a web-based application using the Streamlit framework and deployed through cloud computing, allowing the application to run on an internet server with prediction data securely stored in Firebase. In addition, the use of cloud computing simplifies the deployment and maintenance process through integration with the GitHub repository. This system is expected to assist airport management in decision-making and to anticipate fluctuations in passenger numbers more effectively.