Inside the forecaster
"A Random Forest trained on 73 days of real Aditya-L1 data. Detects solar flares from SoLEXS soft X-ray and HEL1OS hard X-ray features at 1-second cadence."
Validation accuracy
61.0%
Trained on 73 days of real Aditya-L1 satellite data (July 2024 – June 2026). Validated using cross-block temporal holdout on unseen solar periods.
Per-class metrics
| Class | Precision | Recall | F1 |
|---|---|---|---|
| Quiet | 0.790 | 0.690 | 0.730 |
| Pre-flare | 0.160 | 0.250 | 0.190 |
| Flare | 0.440 | 0.610 | 0.510 |
Architecture
Algorithm Random Forest Classifier
Estimators 200 trees · max depth 12 · min samples leaf 5
Class weight Balanced (handles flare rarity ~11% of data)
Imputation Median strategy (handles instrument gaps)
Input → (1, 20) // 20 engineered features per second
// SoLEXS + HEL1OS · rolling stats · derivatives
Output → P(quiet), P(pre-flare), P(flare)
Training data 73 days · 6.3M rows · July 2024 – June 2026
Aditya-L1 SoLEXS + HEL1OS · 1-second cadence
Labels GOES XRS flare catalog
Validation Cross-block temporal (leave-one-block-out)
Peak result flare_prob = 0.999 at 19:28 UTC June 21 2026