10th International Congress on Information and Communication Technology

Authors - Lili Ayu Wulandhari , Aditya Kurniawan, Nathania Christy Nugraha, Natasha Hartanti Winata
Abstract - The stock market, as a fundamental component of a nation's economic structure, profoundly impacts the course of economic progress. The stock market's performance can influence foreign investment and the growth of the real sector. Stock fluctuations are essential for providing valuable insights to market participants and investors, allowing them to develop strategies that foster growth and macroeconomic stability. Nonetheless, precisely forecasting stock values is challenging. This work investigates statistical feature extraction for stock price prediction utilizing machine learning approaches on Indonesia's leading five LQ45 companies: BBCA, ASII, BBRI, BMRI, and TLKM. Daily stock price data from 2019 to 2024 is employed, emphasizing statistical characteristics such as mean, standard deviation, skewness, kurtosis, and interquartile range. The features are examined utilizing decision trees, random forests, and multilayer perceptrons (MLPs) with hyperparameter optimization. The MLP model demonstrates superior accuracy, achieving an average 𝑅2 𝑆𝑐𝑜𝑟𝑒 of 0.95 and a MAE that is 92.52% lower than the standard deviation of the actual stock price. The results indicate the effectiveness of statistical feature-based machine learning for accurate stock price prediction, implying possible uses in financial decision-making and market analysis.