Jadwal Sholat

Memuat jadwal sholatโ€ฆ

Ilmu Komputer & AI editorial

Open AccessOA2023

An Overview of Machine Learning, Deep Learning, and Reinforcement Learning-Based Techniques in Quantitative Finance: Recent Progress and Challenges

This review surveys machine learning, deep learning, reinforcement learning, and deep reinforcement learning applications in quantitative finance and stock market forecasting, highlighting recent progress and outlining future research directions.
S. Sahu; A. Mokhade; N. Bokdeยท Applied Sciencesยท 2023ยท DOI 10.3390/app13031956

The core problem

Forecasting stock market behavior is a classic yet difficult problem that has attracted economists and computer scientists alike. Over the past two decades, researchers have explored linear models and models based on machine learning (ML), deep learning (DL), reinforcement learning (RL), and deep reinforcement learning (DRL) to build accurate predictive models. The paper aims to explain the utility of these techniques in quantitative finance (QF) and the stock market, and to outline potential future research paths. The authors note that ML algorithms can now extract high-level financial market data patterns, and investors are using DL models to anticipate and evaluate stock and foreign exchange markets due to AI advantages. Recent years have seen proliferation of DRL in algorithmic trading, with DRL agents combining price prediction and trading signal production to construct fully automated trading systems or strategies.

Innovation

As a review paper, the methodology involves a systematic overview of the literature on ML, DL, RL, and DRL in QF. The authors synthesize recent progress, categorize techniques, and discuss challenges. The paper likely covers:
- **ML models**: linear models, support vector machines, tree-based methods, etc.
- **DL models**: recurrent neural networks (RNNs), long short-term memory (LSTM), convolutional neural networks (CNNs), transformers.
- **RL and DRL**: Q-learning, policy gradients, deep Q-networks (DQN), actor-critic methods.
The review also outlines future research directions based on the presented overview. No specific equations or diagrams are provided in the raw content, but typical formulations include:
- Price prediction:

where is a feature vector.
- RL objective: maximize expected return
.

Introduction
Forecasting stock market behavior is a classic yet difficult problem that has attracted economists and computer scientists alike. Over the past two decades, researchers have explored linear models and models based on machine learning (ML), deep learning (DL), reinforcement learning (RL), and deep reinforcement learning (DRL) to build accurate predictive models. The paper aims to explain the utility of these techniques in quantitative finance (QF) and the stock market, and to outline potential future research paths. The authors note that ML algorithms can now extract high-level financial market data patterns, and investors are using DL models to anticipate and evaluate stock and foreign exchange markets due to AI advantages. Recent years have seen proliferation of DRL in algorithmic trading, with DRL agents combining price prediction and trading signal production to construct fully automated trading systems or strategies.

Why it matters

The authors discuss the utility of ML, DL, RL, and DRL in QF and the stock market, and outline potential future research paths. Challenges likely include data quality, overfitting, non-stationarity, interpretability, and computational complexity. The review aims to enable interested researchers to stay current and easily imitate earlier findings. Future directions may involve hybrid models, improved generalization, and integration of alternative data. The paper concludes by summarizing the progress and challenges, providing a foundation for further research.

Who should read this

Pembaca Research Digest dan praktisi rumpun ini.

Opening member contentโ€ฆ