/Healtchare_Fraud_Detector

Healthcare provider fraud detector is a classification machine learning project to predict the potential fraudulent providers based on the healthcare claims filed.

Primary LanguageJupyter Notebook

Title

Classification - predict the potentially fraudulent providers based on the healthcare claims filed

  • Data Source Kaggle - Healthcare Provider Fraud Detection Analysis
  • Features Claim Dates, Doctor IDs, Patient info (DOB, DOD, Gender, Chronicle Conditions..), Procedure Codes, Diagnosis Codes, Reimbursement Amount, Deductible Amount, Geo Locations, Length of Hospital Stay
  • Target y for each provider (0 - Not Potential Fraud, 1 - Potential Fraud)
  • Outcomes
    • Identify important variables in detecting the behavior of potentially fraud providers
    • Label potential fraud providers
    • Study fraudulent patterns in the provider's claims
  • Performance Metrics
    • F1-score
    • ROC AUC

EDA

Data Model - merge and union data to prepare the dataframe

data model

Bench Model

model flowchart

Model

App

WebApp API Github URL