Faculty Sponsor
Azim Ahmadzadeh
Final Abstract for URS Program
Traditional metrics for evaluating binary classifiers, such as Accuracy, F1 Score, and True Skill Statistic (TSS), often obscure the underlying tradeoffs between true positive and true negative performance—particularly in imbalanced or high-stakes domains. This poster introduces the Contingency Space, a two-dimensional representation of classifier behavior defined by true positive rate (TPR) and true negative rate (TNR). Within this space, scalar performance metrics become geometric surfaces, revealing how scores vary across the entire landscape of possible classifier outputs.
We present a Python package that implements this framework, enabling users to map model predictions into the Contingency Space, visualize metric surfaces (including the effects of class imbalance), and trace classifier trajectories across it. Critically, the tool allows users to define custom performance metrics as analytic functions of TPR and TNR, empowering researchers to design evaluations that reflect domain-specific priorities. This work reframes classifier evaluation as a geometric and exploratory task—offering a visual, flexible, and intuitive alternative to one-size-fits-all metrics.
Presentation Type
Visual Presentation
Document Type
Article