Proceedings of the International Conference on Financial Innovation, FinTech and Information Technology, FFIT 2022, October 28-30, 2022, Shenzhen, China

Research Article

Comparison and Analysis of Prediction for High Volatility Stocks using Different Machine Learning Models

Download2152 downloads
  • @INPROCEEDINGS{10.4108/eai.28-10-2022.2328453,
        author={Minglong  Chen and Mohan  Ren and Yujie  Zhou},
        title={Comparison and Analysis of Prediction for High Volatility Stocks using Different Machine Learning Models},
        proceedings={Proceedings of the International Conference on Financial Innovation, FinTech and Information Technology, FFIT 2022, October 28-30, 2022, Shenzhen, China},
        publisher={EAI},
        proceedings_a={FFIT},
        year={2023},
        month={4},
        keywords={high-volatility; machine learning; lstm; stock prediction},
        doi={10.4108/eai.28-10-2022.2328453}
    }
    
  • Minglong Chen
    Mohan Ren
    Yujie Zhou
    Year: 2023
    Comparison and Analysis of Prediction for High Volatility Stocks using Different Machine Learning Models
    FFIT
    EAI
    DOI: 10.4108/eai.28-10-2022.2328453
Minglong Chen1, Mohan Ren2,*, Yujie Zhou3
  • 1: School of Economics and Management Tongji University Shanghai
  • 2: Department of Computer Science The University of Manchester Manchester
  • 3: Department of Mathematical Sciences The University of Liverpool Liverpool
*Contact email: mohan.ren@student.manchester.ac.uk

Abstract

As the global situation is facing many unpredictable risks, like the covid pandemic and the wars between some countries, a high level of volatility is disturbing the stock market in a world range. Besides, due to the development of machine learning techniques, more and more of these techniques have been used in predicting the price of the stock to get profits, and some of them like Long Short-Term Memory (LSTM) have also shown an extraordinary performance. However, these applications should be reevaluated because of nowadays market volatility while in the past, most experiments were built based on generic volatility. In this experiment, therefore, this study chose two most volatile and two least volatile stocks in the past two years and utilized some popular machine learning models, such as k-Nearest Neighbor (KNN), Support Vector Machine (SVM) and Long Short-Term Memory to predict their prices. The results have shown that all of them will suffer from a roughly 4 to 5 times performance reduction in terms of the prediction ability, but when ignoring how the volatility is, LSTM will always give the best performance.