The Culture: Trend Forecasting Model (XGBoost)
Multi-class XGBoost classifier predicting a fashion trend's lifecycle stage (emerging, growing, peak, stable, declining) from engagement and velocity signals.
Overall accuracy
0.2024
Per-class metrics
- declining: precision=0.219, recall=0.389, f1=0.280
- emerging: precision=0.143, recall=0.083, f1=0.105
- growing: precision=0.333, recall=0.125, f1=0.182
- peak: precision=0.118, recall=0.125, f1=0.121
- stable: precision=0.211, recall=0.286, f1=0.242
Features used
Categorical (one-hot): ['garment_type', 'color', 'archetype_affinity', 'season', 'region', 'trend_velocity', 'price_tier']
Numeric: ['mention_count', 'save_count', 'search_volume_index', 'post_engagement_avg', 'week_over_week_growth_pct', 'thrift_availability_score', 'influencer_adoption_score']
Usage
See the inference / Hugging Face call cells in the training notebook for a full working example.
Part of The Culture ML model suite.
