Data Science2021Data Scientist
Best Binary Classifier Bake-off
KNN, trees, SVM, and logistic regression compared
Visualised, preprocessed, and normalised a classification dataset, then compared K-nearest neighbours, decision trees, SVM, and logistic regression using F1, Jaccard, and log-loss.
The Problem & Engineering Constraint
The Core Challenge
Picking a classifier from accuracy alone hides precision/recall and probability-quality differences.
Technical Architecture & Approach
Engineering Solution & Implementation
Normalised features, trained four classical models, and ranked them on F1, Jaccard, and log-loss.
Measured Production Impact
Verified Outcomes & Deliverables
Documented a side-by-side metric comparison and selected the best model.
Technologies & Components
System Tooling & Technologies
PythonPandasNumPyScikit-Learn