Praedixa / Active research
Praedixa Research — demand forecasting
Research repository for Praedixa's wedge: anticipating demand and operational needs in perishable food businesses.
Problem
Food operators make production, inventory and staffing decisions under volatile demand with underused operational data.
Approach
Local bronze/silver/gold pipeline, canonical training matrix, XGBoost/TFT models, temporal evaluations and economic readout separated from predictive error.
Evidence
Warehouse source of truth: gold_training_matrix_d1.
Strict temporal backtests and explicit latent demand vs observed sales distinction.
Current result: economic gain under the cost function, while global MAE still trails the robust baseline.
Limitations
Still active research: results should be read as evidence of pipeline and evaluation discipline, not as product traction.