Term-deposit classification pipeline

An end-to-end data mining workflow for the Data Mining module, built around a bank-marketing-style classification problem: predicting term-deposit subscription from a real-shaped dataset with the class-imbalance problem that kind of data always carries.

Preprocessing came first, then SMOTE to handle the imbalance directly rather than let the model quietly learn to just predict the majority class, then PCA for dimensionality reduction before modeling. The comparison at the center of the module was Logistic Regression against SVM on the same prepared dataset, evaluating not just raw accuracy but how each model handled the imbalance-corrected data differently.

It's a compact project next to the module's siblings, but a complete pipeline in miniature, preprocessing through model comparison, the same shape a production classification system would follow at much larger scale.