Exploring the Relationship among Predictability, Prediction Accuracy and Data Frequency of Financial Time Series.

Area of Science:

• Quantitative Finance
• Financial Market Analysis
• Information Theory

Background:

• Understanding stock price predictability is crucial for financial market analysis.
• Data frequency is a potential factor influencing stock market predictability and prediction accuracy.
• Previous work introduced the Ensemble Empirical Mode Decomposition-Filtered Frequency-domain (EEMD-FFH) method for financial forecasting.

Purpose of the Study:

• To investigate the relationship between data frequency and the predictability of stock closing price changes.
• To quantify the predictability of stock price changes using different data frequencies (five-minute and daily).
• To evaluate if higher sampling frequency in financial data leads to improved prediction accuracy.

Main Methods:

• Proposed a new information-theoretic estimator, Plz (derived from Lempel-Ziv entropy), to measure predictability.
• Applied the previously developed EEMD-FFH prediction method to assess prediction accuracy.
• Utilized five-minute and daily closing price data for the SSE 50 index from the Chinese stock market.
• Performed linear regression analysis to examine the correlation between predictability and prediction accuracy.

Main Results:

• Intraday five-minute stock data demonstrated higher predictability than daily data.
• Five-minute stock data also yielded higher prediction accuracy compared to daily data.
• A positive correlation was found between predictability and prediction accuracy across both data frequencies.

Conclusions:

• Data frequency significantly impacts both the predictability and prediction accuracy of stock returns.
• Higher frequency financial data offers superior predictability and accuracy for stock price forecasting.
• Predictability and prediction accuracy in stock markets are positively related, with higher predictability associated with better forecasting outcomes.