LQ45 STOCK PRICE FORECASTING: A COMPARISON STUDY OF ARIMA(P,D,Q) AND HOLT-WINTER METHOD

R. Gunawan Santosa and Antonius Rachmat Chrismanto and Willy Sudiarto Raharjo and Yuan Lukito (2024) LQ45 STOCK PRICE FORECASTING: A COMPARISON STUDY OF ARIMA(P,D,Q) AND HOLT-WINTER METHOD. International Journal of Information Technology and Computer Science Applications (IJITCSA), 2 (2). pp. 115-129. ISSN 2964-3139

[img] Text (Artikel Jurnal)
LQ45 Stock Price Forecasting.pdf - Published Version

Download (639kB)

Abstract

The Holt-Winter method and ARIMA(p,d,q) are two frequently used forecasting techniques. When using ARIMA, errors are expected to be connected with earlier errors because it is based on data correlation with prior data (autoregressive) (moving average). The Holt-Winter model comes in two forms: Multiplicative and Additive Holt-Winter. No one has ever attempted to compare combined time series and cross-section data, despite many prior studies on ARIMA and Holt-Winter. This study will compare the Holt-Winter and ARIMA accuracy rates (p,d,q) in a combined time-series and cross-section dataset. LQ45 stock prices are used because they track the performance of 45 stocks with substantial liquidity, sizable market caps, and solid underlying businesses. We use dataset LQ45 stocks as training data in the range 2016–2021. We use data from January - February 2022 for the testing. In terms of time series data analysis, the terms indata are used for training data and outdata for forecasting test data. Daily stock closing data is used in this case: indata of 1458 and outdata of 39.The Mean Absolute Percentage Error (MAPE) method is used to gauge accuracy. This study contributes to MAPE exploration using a Boxplot diagram from cross-sectional data. The Boxplot diagram shows the MAPE spread, the MAPE's center point, and the presence of outliers from the MAPE of LQ45 stock. According to the findings of this empirical study, the average error rate for predicting LQ45 stock prices using ARIMA is 7,0390%, with a standard deviation of 7,7441%; for multiplying Holt-Winter, it is 29,3919%, with a standard deviation of 25,7571%; and for additive Holt-Winter, it is 18,0463%, with a standard deviation of 18,3504%. Apart from numerical comparisons, based on the Boxplot diagram, it can also be seen visually that the ARIMA MAPE (p,d,q) is more focused than Holt-Winter. In addition, in terms of accuracy distribution, it can be seen that the MAPE accuracy of the ARIMA method produces four outliers. Based on the MAPE accuracy rate, we conclude that Holt-Winter has a bigger error based on the MAPE value than ARIMA(p,d,q) at forecasting LQ45 stock prices.

Item Type: Article
Uncontrolled Keywords: ARIMA(p,d,q); boxplot; Holt-Winter; LQ45; time-series
Subjects: Q Ilmu Pengetahuan > QA Matematika > QA76 Perangkat Lunak (Software) Komputer
T Teknologi > T Teknologi (Umum)
Divisions: Fakultas Teknologi Informasi
Depositing User: Beatrix Stefany
Date Deposited: 26 Sep 2024 01:29
Last Modified: 26 Sep 2024 01:29
URI: http://katalog.ukdw.ac.id/id/eprint/9429

Actions (login required)

View Item View Item