Seasonal Time Series Forecasting by F¹-Fuzzy Transform

Area of Science:

• Meteorology and Atmospheric Sciences
• Data Science and Computational Statistics

Background:

• Seasonal time series forecasting is crucial for climate prediction and resource management.
• Existing fuzzy transform methods require performance improvements for seasonal weather data.

Purpose of the Study:

• To develop and validate a novel seasonal forecasting method using the F1-transform (fuzzy transform of order 1).
• To enhance the predictive accuracy of fuzzy transform-based approaches for seasonal weather time series.

Main Methods:

• Polynomial fitting to determine the time series trend.
• Partitioning data into seasonal subsets for F1-transform component calculation.
• Utilizing inverse F1-transforms for future weather parameter prediction.

Main Results:

• The proposed F1-transform method demonstrated superior performance compared to ARIMA, ADANN, and standard seasonal F-transform methods.
• Accurate heat index predictions were achieved for July and August in the Campania Region, Italy (2003-2017).

Conclusions:

• The enhanced F1-transform method offers a significant improvement for seasonal weather forecasting.
• This approach provides a robust tool for predicting weather parameters in seasonal time series analysis.