Analytics Provecta

Data Science That Drives Real Decisions

From predictive modelling to scenario simulations, we help organisations transform data into strategy.

What We Do

Analytics Provecta is a data science consultancy specialising in predictive analytics, forecasting, and operational decision intelligence. We work across energy, infrastructure, retail, and finance to build models that clarify what comes next — and quantify the impact of what-if scenarios before they happen.

Our Services

Predictive ModellingTime-series forecasting with XGBoost, LSTMs, and ensemble methods tailored to your domain.
Scenario SimulationWhat-if analysis and sensitivity modelling to stress-test decisions before they're made.
Explainable AISHAP-based model explanations that make machine learning transparent to stakeholders.
Dashboards & ReportingInteractive visualisations that turn model outputs into actionable intelligence.
Strategic ConsultingEnd-to-end advisory from data architecture to production deployment.

Featured Demo: Split, Croatia — Electricity Consumption Forecast

Split is Croatia's second-largest city and the heart of the Adriatic tourism economy. Its electricity demand profile is the inverse of inland Croatia: mild winters mean almost no heating load, but scorching summers and 1.5 million annual tourists drive heavy air-conditioning demand from June through August. This XGBoost model predicts weekly consumption 24 weeks ahead using Split weather data (temperature, wind, pressure) trained directly on Split-scale consumption.

Avg Forecast
—
GWh / week
Peak Week
—
—
vs Last Year
—
same 24-week period
Model Accuracy
—
test RMSE (GWh/week)

Scenario Explorer

How it works: Croatia shows a double-peak consumption profile: cold winters drive electric heating, while hot Adriatic summers spike air-conditioning demand. The model captures this U-curve around ~18°C — temperatures both well below and well above this point increase weekly consumption.

Procurement Risk Calendar Loading market price…

Low risk — 70% forward cover Medium risk — 80% forward cover High risk — 90% forward cover

Model Validation — Actual vs Predicted (Test Set)

What Drives the Forecast? (SHAP Feature Importance)

Contact

analyticsprovecta@gmail.com