Year: 2026 | Month: May | Volume: 13 | Issue: 5 | Pages: 716-721
DOI: https://doi.org/10.52403/ijrr.20260571
Modeling the Stock Price of PT Bank Negara Indonesia (Persero) Tbk Using Multivariable Local Polynomial Regression
Melani Ayu Azizah1, Suparti2, Diah Safitri3
1,2,3Department of Statistics, Faculty of Science and Mathematics, Diponegoro University, Semarang, Indonesia
Corresponding Author: Diah Safitri
ABSTRACT
Stocks are investment instruments with relatively high risk, and their prices form time series data. Forecasting stock price data helps investors make informed decisions. However, these data often exhibit complex, nonlinear patterns. Nonparametric regression offers a flexible approach that does not assume a specific functional form, allowing it to better adapt to data characteristics. This study models and forecasts the closing stock price of PT Bank Negara Indonesia (Persero) Tbk (BNI) using multivariable local polynomial regression. The data span from 1 December 2024 to 30 December 2025 and are split into 80% in-sample and 20% out-of-sample data. The in-sample data are used for stationarity testing, lag selection via the Partial Autocorrelation Function (PACF), model estimation using local polynomial regression with various bandwidths, and bandwidth selection based on the Generalized Cross Validation (GCV) criterion, followed by assessing model goodness-of-fit using the coefficient of determination. The out-of-sample data are used to evaluate the model's forecasting performance using the Mean Absolute Percentage Error (MAPE). The results show that a first-degree local polynomial regression model with the Epanechnikov kernel achieves the lowest Generalized Cross Validation (GCV) value of 270,710.862, a coefficient of determination of 82.33%, and an out-of-sample MAPE of 1.03%, indicating strong explanatory power and high forecasting accuracy for the BNI stock price.
Keywords: Stock price, Local polynomial regression, Generalized Cross Validation, Epanechnikov kernel, Forecasting.
[PDF Full Text]