Code and dataset for "F10.7 Index Prediction: A Multiscale Decomposition Strategy with Wavelet Transform for Performance Optimization"

project brief This project has implemented the innovative prediction scheme proposed in the paper titled "F10.7 Index Prediction: A Multiscale Decomposition Strategy with Wavelet Transform for Performance Optimization". For the first time, we have applied wavelet decomposition to the prediction of the F10.7 index. By extracting multi-scale signal features and combining them with the iTransformer model, we have significantly improved the prediction performance. Core innovation points 🌊 First application of wavelet decomposition: Utilizing wavelet decomposition to extract the approximate and detail signals of the F10.7 index 🔄 Multi-scale feature fusion: Using both the original signal and decomposed signals as model inputs ☀️ Multi-source data integration: Incorporating the International Sunspot Number (ISN) and its decomposed signals 📊 Comprehensive performance comparison: Systematically comparing with the latest research and prediction models from international institutions 🔬 Strong generalization ability: Validating the effectiveness of the optimal combination model on different datasets and model architectures project structure ├── 1_Wavelet-decomposed_data/ # Section 4.1 Wavelet-decomposed data │ ├── code file # The implementation of wavelet decomposition │ └── folder # The decomposed data and the visualization results ├── 2_F10.7_wavelet_signals/ # Section 4.2 Performance comparison: integrating wavelet-decomposed F10.7 signals │ └── code file # model training │ └── results/ # Trained model and experimental result files ├── 3_ISN_wavelet_signals/ # Section 4.3 Performance comparison: integrating wavelet-decomposed ISN signals │ ├── code file # model training │ └── results/ # Trained model and experimental result files ├── 4_Comparison_study/ # Section 4.4 Comparison with the latest research and international institutions │ ├── 1Compare.py # Comparison between different years │ ├── 2Compare.py # Comparison between different conditions │ ├── period_average_metrics.csv # The comparison results between different years │ └── four_conditions.csv # The comparison results between different conditions └── 5_Generalization_test/ # Section 4.5 Generalization performance on Dataset B ├── code file # model training └── results/ # Trained model and experimental result files Data sources • F10.7 index and International Sunspot Number(ISN):DRAO • Langfang data:L&S observations environmental requirement python==3.12 torch==2.8.0 numpy==1.26.4 pandas==2.2.2 scipy==1.13.0 scikit-learn==1.7.1 matplotlib==3.8.4 seaborn==0.13.2 PyWavelets==1.9.0 Last update by Ma, xuran: January 24th, 2026