Electronic Theses and Dissertation

Kecerdasan buatan telah menjadi alat yang semakin penting dalam berbagai aspek kehidupan manusia, termasuk dalam analisis peramalan saham. Penelitian ini mengkaji implementasi kecerdasan buatan dengan menggunakan Recurrent Neural Network (RNN) berupa Long Short-Term Memory (LSTM) dalam konteks peramalan harga saham Bank Central Asia (BCA). Data historis penutupan harga saham BCA sebanyak 993 data harian digunakan untuk merumuskan model peramalan yang efektif. Model LSTM yang diusulkan dalam penelitian ini terdiri dari dua lapisan LSTM dengan 100 unit pada setiap lapisan, serta dua lapisan dense dengan 25 dan 1 unit masing-masing. Hasil peramalan menggunakan model ini dievaluasi terhadap
data aktual dari 26 November 2021 hingga 21 September 2022. Dari hasil penelitian, ditemukan bahwa model LSTM yang diimplementasikan memiliki kinerja peramalan yang baik. Nilai rata-rata kesalahan prediksi (RMSE) model pada saat uji prediksi model lebih kecil yaitu 13.32 dibandingkan dengan model yang dicoba pada saat uji validasi model sebesar 7.733,06 serta diperoleh nilai uji MAPE sebesar 5.13 %, artinya model LSTM memiliki tingkat kesalahan rata-rata sebesar 5.13% dalam memprediksi penutupan harga saham BCA.

Kata Kunci: LSTM, Peramalan, Saham, RMSE, MAPE

Artificial intelligence (AI) has become an increasingly important tool in various aspects of human life, including in stock forecasting analysis. This study examines the implementation of artificial intelligence using a Recurrent Neural Network (RNN) in the form of Long Short-Term Memory (LSTM) in the context of Bank Central Asia (BCA) stock price forecasting. Historical data on the closing price of BCA's shares of 993 daily data are used to formulate an effective forecasting model. The LSTM model proposed in this study consists of two LSTM layers with 100 units in each layer, as well as two dense layers with 25 and 1 units respectively. Forecasting results using this model were evaluated against actual data from 26 November 2021 to 21 September 2022. From the research results, it was found that the implemented LSTM model had good forecasting performance. The average value of the prediction error (RMSE) of the model during the model prediction test was smaller, namely 13.32 compared to the model tested during the model validation test of 7,733.06 and the MAPE test value was obtained of 5.13%, meaning that the LSTM model has an average error rate an average of 5.13% in predicting the closing price of BCA shares. Keywords: LSTM, Forecasting, Stocks, RMSE, MAPE