Prediksi Return Saham PT. Petrindo Jaya Kreasi Tbk. (CUAN) Dengan Pendekatan Two-stage AR-ARCH Dan Multilayer Perceptron

Ningrum, Narsisha Sekar (2026) Prediksi Return Saham PT. Petrindo Jaya Kreasi Tbk. (CUAN) Dengan Pendekatan Two-stage AR-ARCH Dan Multilayer Perceptron. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221163-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Peramalan return saham merupakan salah satu aspek penting dalam pengambilan keputusan investasi karena mampu memberikan gambaran mengenai potensi keuntungan dan risiko yang akan dihadapi investor. Data return saham umumnya memiliki karakteristik berupa volatilitas yang berubah-ubah dari waktu ke waktu sehingga memerlukan metode yang mampu menangkap dinamika rata-rata dan varians secara simultan. Penelitian ini bertujuan untuk menganalisis karakteristik return saham PT Petrindo Jaya Kreasi Tbk (CUAN) serta mengembangkan model two-stage AR-ARCH dan Multilayer Perceptron (MLP) untuk meningkatkan akurasi peramalan return saham. Data yang digunakan berupa data harga penutupan harian saham CUAN yang ditransformasikan menjadi return logaritmik. Analisis diawali dengan identifikasi karakteristik data melalui statistika deskriptif, uji stasioneritas, analisis autokorelasi, uji efek ARCH, dan uji nonlinieritas Teräsvirta. Hasil analisis menunjukkan bahwa data return bersifat stasioner dan mengandung efek ARCH sehingga pemodelan volatilitas dilakukan menggunakan model ARCH(1). Informasi return historis dan volatilitas hasil estimasi ARCH kemudian digunakan sebagai fitur masukan pada model MLP. Penentuan arsitektur terbaik dilakukan menggunakan Grid Search berdasarkan nilai validation loss terkecil. Hasil penelitian menunjukkan bahwa model ARCH(1) menghasilkan nilai Root Mean Squared Error (RMSE) sebesar 0,048607, sedangkan model two-stage ARCH–MLP menghasilkan RMSE sebesar 0,043699. Hasil tersebut menunjukkan bahwa penambahan MLP mampu meningkatkan akurasi peramalan sebesar 10,10% dibandingkan model ARCH tunggal. Dengan demikian, model two-stage ARCH–MLP terbukti lebih efektif dalam meramalkan return saham CUAN dibandingkan pendekatan ARCH saja. Penelitian ini menunjukkan bahwa integrasi metode statistika dan machine learning dapat menjadi alternatif yang baik untuk meningkatkan akurasi peramalan return saham pada data yang memiliki karakteristik volatilitas yang dinamis.
=====================================================================================================================================
Stock return forecasting is an important aspect of investment decision-making as it provides insights into the potential returns and risks faced by investors. Stock return data generally exhibit time-varying volatility, requiring forecasting methods capable of capturing both mean and variance dynamics simultaneously. This study aims to analyze the characteristics of PT Petrindo Jaya Kreasi Tbk (CUAN) stock returns and to develop a two-stage ARCH–Multilayer Perceptron (MLP) model to improve stock return forecasting accuracy. The data used in this study consist of daily closing prices of CUAN stock, which were transformed into logarithmic returns. The analysis began with identifying data characteristics through descriptive statistics, stationarity testing, autocorrelation analysis, ARCH effect testing, and the Teräsvirta nonlinearity test. The results indicate that the return series is stationary and exhibits ARCH effects; therefore, volatility modeling was conducted using an ARCH(1) model. Historical returns and conditional volatility estimated from the ARCH model were subsequently used as input features for the MLP model. The optimal network architecture was determined through Grid Search based on the lowest validation loss. The results show that the ARCH(1) model achieved a Root Mean Squared Error (RMSE) of 0.048607, while the two-stage ARCH–MLP model achieved an RMSE of 0.043699. These findings indicate that incorporating MLP improved forecasting accuracy by approximately 10.10% compared to the standalone ARCH model. Therefore, the two-stage ARCH–MLP model proved to be more effective in forecasting CUAN stock returns than the ARCH approach alone. This study demonstrates that integrating statistical methods and machine learning can serve as an effective alternative for improving stock return forecasting accuracy in data characterized by dynamic volatility.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	AR-ARCH, MLP, Peramalan, Return Saham, Two-stage. AR-ARCH, MLP, Forecasting, Stock Return, Two-stage.
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HG Finance > HG4915 Stocks--Prices Q Science > QA Mathematics > QA280 Box-Jenkins forecasting Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Narsischa Sekar Ningrum
Date Deposited:	05 Aug 2026 05:50
Last Modified:	05 Aug 2026 05:50
URI:	http://repository.its.ac.id/id/eprint/143929

Actions (login required)

View Item