Back to home

Methodology & Sources

Our Pricing Algorithm: Methodology & Sources

Nexorev combines published forecasting methods, public tourism indicators, and hotel-specific PMS data once a pilot property is integrated. This page documents the methodology without inventing ML researchers, PhDs, or deployed customer results.

Demand Forecasting

Prophet by Meta

Additive time-series forecasting with trend, seasonality, holiday, and analyst-tunable components.

Forecasting at Scale

The published Prophet methodology by Taylor and Letham.

ARIMA via statsmodels

Classical autoregressive integrated moving-average modeling for time-series baselines.

XGBoost

Gradient-boosted trees for tabular demand and pricing features.

XGBoost paper

A Scalable Tree Boosting System by Chen and Guestrin.

Real Data Sources

These sources inform market priors and public-data backtests. Production pricing still requires PMS data from each hotel.

source aliases for verification:
istat
bancaditalia
enit

Algorithm Transparency

The pricing engine is designed as a decision-support system. It should recommend rates inside hotel-defined guardrails, not silently override commercial judgment.

Current and forecast occupancy
Booking lead time and booking pace
Comparable-set rate position
Local events and seasonal demand
Weather and travel disruption signals
Day-of-week and holiday effects
Length of stay and cancellation risk
Hotel-defined floors, ceilings, and blackout rules

Honest Limitations

Currently in pilot stage with synthetic + public data; production deployment requires hotel PMS integration. Forecasts and simulated revenue lift on this site are backtests, not proof of deployed customer performance.

View model performance transparency