Data Overview
Profile the AllLife Bank customer dataset, classify every column, and flag any anomalies before modeling.
Unique customer ID — dropped before modeling.
Years.
Years of professional experience (some negative — anomalous).
Family size (1–4).
1 = Undergrad, 2 = Graduate, 3 = Advanced/Professional.
Geographic — treated as categorical, dropped for baseline modeling.
Annual income ($K).
Average monthly credit-card spend ($K).
Mortgage value, $K (0 = no mortgage).
Binary 0/1.
Binary 0/1.
Binary 0/1.
Binary 0/1.
Binary target — 1 = accepted.
• ZIPCode is geographic and high-cardinality — treated as categorical, dropped for the baseline tree to avoid overfitting on noise.
• Binary 0/1 columns (Securities, CD, Online, CreditCard) are not one-hot encoded — they are already model-ready.
• Negative Experience values are flagged for treatment, not silently ignored.
• Class imbalance (0.0% positives) means the train/test split is stratified on Personal_Loan.
Click Generate to produce an evidence-based observation, insight, recommendation, and business benefit for this section.