Seasonality

Intermediate Predictive Analytics in Python

Nele Verbiest

Senior Data Scientist @PythonPredictions

Seasonal effects (1)

Seasonal effects (2)

Mean number of donations

Seasonal effects (3)

Mean donation amount

Seasonality and the timeline (1)

Seasonality and the timeline (2)

Seasonality and the timeline (3)

Dealing with seasonality

Check for seasonality

gifts.groupby("month")["amount"].mean()
gifts.groupby("month").size()

Use appropriate timeline in history

Seasonality and predictive models

Model timeline May 2018

logreg = linear_model.LogisticRegression()
logreg.fit(X_may2018, y_may2018)

predictions = logreg.predict_proba(X_jan2019)[:,1]
auc = roc_auc_score(y_jan2019, predictions)
print(round(auc,2))

0.53

Model timeline January 2018

logreg = linear_model.LogisticRegression()
logreg.fit(X_jan2018, y_jan2018)

predictions = logreg.predict_proba(X_jan2019)[:,1]
auc = roc_auc_score(y_jan2019, predictions)
print(round(auc,2))

0.56

Let's practice!

Intermediate Predictive Analytics in Python