Slope Entropy Characterisation: The Role of the δ Parameter

Area of Science:

• * Signal Processing
• * Data Science
• * Machine Learning

Background:

• * Time series entropy calculation methods are crucial for signal classification across scientific fields.
• * Slope Entropy (SlpEn) is a recently proposed method utilizing consecutive sample differences and two parameters: gamma (γ) and delta (δ).
• * The role of the delta (δ) parameter, typically set to small values like 0.001, in SlpEn's classification performance has not been rigorously quantified.

Purpose of the Study:

• * To investigate the influence of the delta (δ) parameter on Slope Entropy's time series classification performance.
• * To assess if removing δ or optimizing its value can maintain or improve classification accuracy.
• * To determine if a simplified SlpEn offers a practical alternative without significant performance loss.

Main Methods:

• * Experimental evaluation of SlpEn with δ removed from the calculation.
• * Grid search optimization of the δ parameter to identify potentially better values.
• * Comparison of classification accuracy across different SlpEn configurations using various datasets.

Main Results:

• * Removing the δ parameter from SlpEn calculation resulted in minimal loss of classification accuracy.
• * Optimizing δ through grid search showed only marginal improvements (up to 5%) in classification performance.
• * The experimental results indicate that the δ parameter's contribution to accuracy gains is limited.

Conclusions:

• * The delta (δ) parameter in Slope Entropy (SlpEn) has a limited impact on time series classification accuracy.
• * Simplifying SlpEn by removing or optimizing δ is a feasible approach, offering a potential reduction in computational complexity.
• * A simplified SlpEn calculation presents a viable alternative for signal classification tasks without substantial performance degradation.