Studi Komparatif Kinerja Model ARIMA Kalman Filter dan Average-based Fuzzy Time Series dalam Memprediksi Harga Saham BBCA.JK

Abstract

Saham merupakan salah satu jenis investasi yang menawarkan berbagai keuntungan. Meskipun begitu, pergerakan harga saham dipengaruhi oleh banyak faktor eksternal sehingga diperlukan model yang tepat dalam memprediksi harga saham. Model yang dapat digunakan dalam memprediksi harga saham adalah Autoregressive Integrated Moving Average Kalman Filter (ARIMA-KF) dan Average-based Fuzzy Time Series (Average-based FTS). Penelitian ini membandingkan hasil prediksi model ARIMA-KF dan Average-based FTS. Data penelitian yang digunakan merupakan harga penutupan saham PT Bank Central Asia Tbk pada rentang waktu 10 Oktober 2023 hingga 10 Oktober 2024. Data dibagi menjadi 70% data training dan 30% data testing. Model ARIMA(2,1,0) dipilih sebagai model terbaik dengan nilai MAPE sebesar 4,25948%. Model tersebut selanjutnya digunakan dalam membangun state space model Kalman Filter dan menghasilkan model ARIMA-KF dengan nilai MAPE sebesar 0,91759%. Model Average-based FTS menghasilkan nilai MAPE sebesar 1,23416%. Model ARIMA-KF menunjukkan kinerja model terbaik dengan nilai MAPE terendah, sehingga dapat dijadikan sebagai pilihan terbaik dalam prediksi harga penutupan saham BBCA.JK.

Stock investment is highly popular due to the potential returns it offers. However, stock price movements are influenced by multiple external factors. Accurate forecasting models are essential to support optimal investment decisions. Two models that can be used for stock price prediction are the Autoregressive Integrated Moving Average Kalman Filter (ARIMA-KF) and the Average-based Fuzzy Time Series (Average-based FTS). This study aims to compare the prediction results using ARIMA-KF and Average-based FTS. Dataset used in this research consists of the closing prices of PT Bank Central Asia Tbk from October 10, 2023, to October 10, 2024. Dataset divided into 70% data training and 30% data testing. ARIMA(2,1,0) model was identified as the best model with a MAPE of 4,25948%. Incorporating the Kalman Filter improved the prediction accuracy, reducing the MAPE to 0,91759%. Additionally, the Average-based FTS model achieved a MAPE of 1,23416%. These results indicate that the ARIMA-KF model provides superior predictive accuracy, making it preferred choice for forecasting BBCA.JK stock prices.