Nachfrageprognose & Personalplanung

Die Herausforderung verstehen

Roberto's Pizzeria needs to plan staff schedules a week in advance. Too many cooks on a quiet Monday wastes money; too few on a busy Friday evening means long wait times and unhappy customers. Your task is to build a data-driven staffing recommendation system.

Datenquellen erkunden

You have access to several data files that will drive your predictions:

# Fetch historical order data from the API
curl -s http://localhost:8080/menu | python3 -m json.tool

# Weather data for the region
cat /data/exercise4/weather.json | python3 -m json.tool | head -30

# Calendar with holidays and school breaks
cat /data/exercise4/calendar.json | python3 -m json.tool

# Local sports events (football matches drive pizza orders!)
cat /data/exercise4/sports_events.json | python3 -m json.tool

# Menu item preparation times
cat /data/exercise4/menu.json | python3 -m json.tool

Bestellmuster analysieren

Start by understanding the baseline demand patterns from historical order data:

# Load and explore order history
python3 -c "
import pandas as pd
import json

# Load order history
orders = pd.read_json('/data/exercise4/order_history.json')
print(f'Total orders: {len(orders)}')
print(f'Date range: {orders.date.min()} to {orders.date.max()}')

# Orders by day of week
orders['dow'] = pd.to_datetime(orders['date']).dt.day_name()
print(orders.groupby('dow')['order_count'].mean())

# Peak hours
orders['hour'] = pd.to_datetime(orders['timestamp']).dt.hour
print(orders.groupby('hour')['order_count'].sum())
"

Nachfrageprognose-Modell erstellen

Combine order history with external data to predict future demand:

# Feature engineering
python3 build_features.py \
    --orders /data/exercise4/order_history.json \
    --weather /data/exercise4/weather.json \
    --calendar /data/exercise4/calendar.json \
    --sports /data/exercise4/sports_events.json \
    --output features.csv

# Train prediction model
python3 train_model.py \
    --features features.csv \
    --model random_forest \
    --output demand_model.pkl

# Predict next week's demand
python3 predict.py \
    --model demand_model.pkl \
    --start-date 2026-03-23 \
    --days 7 \
    --output predictions.json

Personalempfehlungen generieren

Convert demand predictions into actionable staff schedules:

# Staffing rules
# - 1 cook can handle ~8 orders/hour
# - Minimum 2 staff at all times
# - Maximum 6 staff in the kitchen
# - Shifts are 4 or 8 hours

python3 generate_schedule.py \
    --predictions predictions.json \
    --cook-capacity 8 \
    --min-staff 2 \
    --max-staff 6 \
    --output weekly_schedule.json

# Visualize the schedule
python3 visualize.py \
    --schedule weekly_schedule.json \
    --output schedule_chart.png

Ergebnisse präsentieren

Prepare a presentation covering:

• Which factors most influence order volume? (feature importance)
• Visualization of predicted vs. actual demand
• Your generated weekly staff schedule with justifications
• Cost analysis: what would over/under-staffing cost Roberto?
• How could this system integrate with the API in production?