Comparative Analysis of Machine Learning Models for Buy / Sell Signals in High-Frequency Stock Trading

Aarav Magesh

Abstract: This research compares the performance of three machine learning-based forecasting models Simple Moving Average (SMA), Exponential Moving Average (EMA), and Holt?s Linear Trend Model in generating buy/sell signals within high-frequency stock trading. Using historical stock data from ADANIENT, ICICIBANK, and TATAMOTORS, the study evaluates the predictive power of each model through simulated investment strategies. Performance is assessed based on return trends over specific intervals, visualized using time-series plots. The results indicate that Holt?s model consistently generates superior returns across various time frames, outperforming both SMA and EMA, which exhibit higher volatility and reversion to mean behaviors. The findings underscore the value of robust trend-based models in designing data-driven trading strategies.

Keywords: Machine learning, Stock Prediction, Algorithmic Trading, Time Series Forecasting, Holt's model

How to Cite?: Aarav Magesh, "Comparative Analysis of Machine Learning Models for Buy / Sell Signals in High-Frequency Stock Trading", Volume 14 Issue 8, August 2025, International Journal of Science and Research (IJSR), Pages: 126-135, https://www.ijsr.net/getabstract.php?paperid=SR25801185454, DOI: https://dx.doi.org/10.21275/SR25801185454