Reference implementation, provider level claims anomaly detection

Description #

The working code behind the healthcare provider fraud detection methodology paper: a single end to end notebook covering exploratory analysis of the beneficiary data, feature engineering including the derived Age and Alive fields and the chronic condition indicators, construction of the discharge and admission fields from the inpatient and outpatient claims, merging beneficiary, claims, and provider level fraud labels into one training set, and training and evaluating three classifiers, Random Forest, Support Vector Machine, and K Nearest Neighbors.

Files #

Each file's SHA-256 is listed above. To confirm a download is unmodified: shasum -a 256 filename

Available to verified institutions at no cost. Verification confirms institutional identity, not eligibility.