Stock Price Prediction using Dynamic Neural Networks
Stock Price Prediction using Dynamic Neural Networks
This paper presents a dynamic neural network framework for predicting daily closing stock prices, arguing that neural networks excel at identifying non-linear patterns in chaotic financial data. The author compares neural network performance with fundamental analysis, technical analysis, and regression techniques, demonstrating superior accuracy. The study also engages with the Efficient Market Hypothesis (EMH) and Chaos theory, providing empirical evidence that stock prices exhibit predictable chaotic behavior, thereby refuting the EMH. The paper concludes with practical recommendations for deploying neural networks in stock price prediction, highlighting their potential to improve trading strategies.
Highlights
- 1Proposes a dynamic neural network approach for daily stock closing price prediction, outperforming traditional methods.
- 2Refutes the Efficient Market Hypothesis (EMH) by demonstrating predictable patterns in stock prices using neural networks.
- 3Supports Chaos theory as a framework for understanding stock market behavior.
- 4Provides a comparative analysis of neural networks against fundamental, technical, and regression techniques.
Methods
- MTime series dynamic neural networks (e.g., NARX, LSTM) for sequential prediction.
- MComparison with fundamental analysis, technical analysis, and linear regression models.
- MEvaluation using historical stock price data with metrics like RMSE and MAE.
Results
- RDynamic neural networks achieve higher prediction accuracy than traditional methods (e.g., lower RMSE).
- RNeural networks capture non-linear patterns in stock price data that linear models miss.
- RThe study finds evidence against the EMH, supporting the presence of predictable chaos in markets.
- RRecommendations for practical implementation of neural networks in trading strategies.
Analyze Paper
Generate insights from "Stock Price Prediction using Dynamic Neural Networks".