VIVES University of Applied Sciences · Kortrijk

See what's coming.
Then play with it.

The VIVES Forecasting research group turns historical data into reliable predictions for companies and public organisations. Below you find three real use cases, each with a small game. Flip the switches, make your call and see how you do against the model.

model.predict(horizon=12)● live

1Flip the switches

Every demo has the same drivers a real model uses: weather, promotions, lead times, threat signals.

2Make your call

Decide how much to cook, when to reorder or where to put your defences.

3Beat the model

Compare your gut feeling with a data driven forecast. Most people lose. Some win.

01
Demand forecasting · TETRA Smart Meal Planning

How many meals do we cook tomorrow?

Every day, kitchens in schools, hospitals, company restaurants and catering guess how many people will show up. Cook too much and food ends up in the bin. Cook too little and guests leave hungry. In the TETRA project Smart Meal Planning, VIVES builds machine learning models (gradient boosted trees such as LightGBM) together with foodservice partners. The models learn from past sales, the calendar, the weather and the menu to predict demand per day and per dish, so kitchens can plan with confidence and cut food waste.

  • INPUTSSales history, weekday, holidays, exams, weather, menu, events
  • MODELGradient boosting (LightGBM) with prediction intervals
  • OUTPUTMeals per day and per dish, with an uncertainty band
  • IMPACTLess food waste, fewer shortages, easier purchasing
Campus restaurant · meal forecaster
Day
Weather
Dish of the day
Calendar & context
Forecast for Monday
0meals± 0
Cook between
0
80% prediction interval
Waste avoided
0 kg
vs. cooking a fixed 480

Every plate is 10 meals forecastcapacity

This week forecast80% interval

Why this number? impact of each driver on the forecast

02
Inventory forecasting · retail & supply chain

Never an empty shelf, never a full warehouse.

Stock is money on a shelf. Too little and customers walk away, too much and you pay for storage, capital and products that expire. Classic reorder rules (order Q units when stock drops below a fixed level) ignore what is coming next: a promotion, a seasonal peak, a supplier that delivers late. Forecast driven replenishment looks ahead. It predicts demand over the lead time and orders exactly what is needed, with a safety buffer that grows with the uncertainty.

  • INPUTSSales per day, promotions, seasonality, lead times
  • MODELProbabilistic demand forecast plus order policy
  • OUTPUTWhen to order, how much, and the safety stock
  • IMPACTHigher service level with less stock on hand
Store 042 · oat milk 1L · 6 week simulation
What's coming
Reorder when stock < 80
Order quantity 200
Your best runs
No runs yet
Day
0

Stock on hand stockdemandstockout

Score
0
out of 100
Service level
0%
demand served
Lost sales
0
units
Avg stock
0
units on hand
03
Predictive cybersecurity

Stop the attack before it happens.

Security teams drown in alerts, and small organisations often have no security team at all. Predictive cybersecurity uses the same forecasting logic as demand planning: learn from signals such as failed logins, phishing waves, unpatched systems and employee behaviour, then estimate where and when an incident is most likely. With limited people and budget, you protect the right systems first. This work links to our research on phishing and digital resilience in Flemish SMEs, where the human factor is often the weakest link.

  • INPUTSLog events, patch status, phishing reports, time of day
  • MODELRisk scoring and time series anomaly forecasting
  • OUTPUTProbability of an incident per system, per hour
  • IMPACTScarce defences go where the risk is highest
SME network · threat radar
Threat context
Day 1 of 5€0k damage
Your shields 3 left

6 attacks will hit in the next 24 hours. A shield protects only the system you put it on. Protect where the risk and the damage are highest.

Damage prevented
Gut feeling–
With AI radar–

Incident risk, next 24 hours predicted probability

VIVES Forecasting research group

Your data already knows what happens next week.

We help organisations turn historical data into forecasts they can act on: from pilots and proof of concepts to funded research projects (TETRA, VLAIO) with companies, care organisations and public partners. Got a planning, stock or risk question? Let's test it on your data.

Get in touch
01Explore. We map your decision and the data behind it.
02Model. We build and validate a forecast on your own history.
03Decide. We turn predictions into rules, dashboards and actions.
04Share. Results flow back into teaching and applied research.