Learning Objectives

• Apply predictive models to business data
• Validate model with business metrics
• Create what-if scenarios

python
Run Inline
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score
import pandas as pd
# Load processed data
df = pd.read_csv('data/processed/sales_features.csv')
# Prepare features
feature_cols = ['price', 'discount', 'marketing_spend', 'competitor_price',
                'day_of_week', 'month', 'is_holiday', 'lag_sales_7d']
X = df[feature_cols]
y = df['sales']
# Train/test split (temporal)
split_idx = int(len(df) * 0.8)
X_train, X_test = X.iloc[:split_idx], X.iloc[split_idx:]
y_train, y_test = y.iloc[:split_idx], y.iloc[split_idx:]
# Model
model = RandomForestRegressor(n_estimators=200, random_state=42)
model.fit(X_train, y_train)
# Predict
y_pred = model.predict(X_test)
# Business metrics
mae = mean_absolute_error(y_test, y_pred)
rmse = np.sqrt(mean_squared_error(y_test, y_pred))
mape = np.mean(np.abs((y_test - y_pred) / y_test)) * 100
r2 = r2_score(y_test, y_pred)
print(f"MAE: ${mae:.0f}")
print(f"RMSE: ${rmse:.0f}")
print(f"MAPE: {mape:.1f}%")
print(f"R2: {r2:.3f}")

Q1: How to evaluate a forecasting model for business?

MAPE (mean absolute percentage error) - interpretable (% error). Compare to naive baseline (persist last value). RMSE penalizes large errors more. Q2: What is lag feature engineering?

Use past values as features. Sales(t-1), Sales(t-7) to predict Sales(t). Captures seasonality and autocorrelation. Common for time series forecasting. Q3: What is the difference between forecasting and prediction?

Forecasting: predicting FUTURE values in time series (ordered). Prediction: general term for estimating any unknown value. Forecasting requires temporal validation. Q4: How to create what-if analysis?

Modify input features (what if price +10%?), run model predictions, compare to baseline. Helps business understand impact of decisions before making them. Q5: What is the business value of a predictive model?

Less than 5% MAPE: excellent for operational planning. 5-10%: good for strategic decisions. >15%: useful for direction but not precision. Context matters. Q6: Create what-if scenario:

python
Run Inline
# What if we increase price by 10%?
scenario = X_test.copy()
scenario['price'] = scenario['price'] * 1.10
scenario_pred = model.predict(scenario)
baseline_pred = model.predict(X_test)
impact = scenario_pred - baseline_pred
print(f"Revenue impact: ${impact.sum():.0f}")