Prediksi Harga Saham Menggunakan Metode K-Nearest Neighbors (KNN) Regression Dan Long Short Term Memory (LSTM)

Marwany, Michelle Cintania Antontte (2025) Prediksi Harga Saham Menggunakan Metode K-Nearest Neighbors (KNN) Regression Dan Long Short Term Memory (LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002201068-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (10MB) | Request a copy

Abstract

Faktor kunci yang mendukung pertumbuhan pasar modal meliputi kemudahan pada akses teknologi dan upaya penyediaan edukasi digital yang diinisiasi oleh pemerintah dan sektor keuangan. Hal ini merupakan sebuah tren yang dipicu oleh kemudahan akses informasi perdagangan online yang semakin berkembang. Salah satu aspek yang semakin menonjol dalam konteks pasar modal adalah prediksi harga saham yang mampu bekerja pada data fluktuatif. Pada penelitian ini telah dilakukan prediksi harga saham menggunakan metode K-Nearest Neighbors Regression (KNNR), Long Short Term Memory (LSTM), dan KNNR-LSTM. Metode KNNR-LSTM merupakan metode gabungan dengan KNNR akan menjadi input awal metode yang hasil prediksinya akan digunakan sebagai fitur baru bagi pengujian model LSTM. Penelitian ini menggunakan data historis saham harian PT. Bank Jago Tbk dari tahun 2016 hingga 2025. Hasil dari pengujian menggunakan metode KNNR dengan time step=5 dan k=15 menghasilkan nilai MAE sebesar 165.82, RMSE sebesar 217.56, dan MAPE sebesar 6.67%. Untuk metode LSTM pada time step 10 dan epoch 1000 menghasilkan nilai MAE = 77.26, RMSE = 108.53, dan MAPE = 2.97%. Pendekatan hybrid KNNR-LSTM pada time step 5 dan epoch 500 menghasilkan MAE = 73.10, RMSE = 103.30, dan MAPE = 2.86%. Dari pernyataan berikut, metode KNNR-LSTM memberikan hasil prediksi dengan hasil terbaik.
====================================================================================================================================
A key factor supporting the growth of the capital market involves the ease of access to technology and efforts to provide digital education initiated by the government and financial sectors. This reflects a growing trend driven by increasingly accessible online trading information. One aspect that has become more prominent in the capital market context is stock price prediction, particularly for stocks with high volatility. This study aims to predict stock prices using the K-Nearest Neighbors Regression (KNNR), Long Short-Term Memory (LSTM), and KNNR-LSTMmethods. The KNNR-LSTM is a hybrid method where the output of KNNR becomes the input for LSTM, serving as a new feature for testing the LSTM model. This research uses historical daily stock data of PT Bank Jago Tbk from 2016 to 2025. The KNNR method with a time step of 5 and k = 15 produced MAE = 165.82, RMSE = 217.56, and MAPE = 6.67%. The LSTM method with a time step of 10 and 1000 epochs produced MAE = 77.26, RMSE = 108.53, and MAPE = 2.97%. The hybrid KNNR-LSTM approach with a time step of 5 and 500 epochs yielded MAE = 73.10, RMSE = 103.30, and MAPE = 2.86%. Based on these results, the KNNR-LSTM method produced the best prediction performance

Item Type:	Thesis (Other)
Uncontrolled Keywords:	K-Nearest Neighbors Regression, Long Short Term Memory, Prediksi Saham, K-Nearest Neighbors Regression, Long Short Term Memory, Stock Prediction.
Subjects:	H Social Sciences > HG Finance H Social Sciences > HG Finance > HG4915 Stocks--Prices
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics
Depositing User:	Michelle Cintania Antontte M
Date Deposited:	05 Aug 2025 04:03
Last Modified:	05 Aug 2025 04:03
URI:	http://repository.its.ac.id/id/eprint/126098

Actions (login required)

View Item