It is an Hospital Insurance project dealing with insurance claim. The business objective was to check if an health insurance claim was genuine or fraudulent.
For EDA refer = Insurance Claim Project.pdf.
For Data Preprocessing, Feature Engineering, Model Building refer = Final_DT.ipynb.
For Deployment refer = Deployment.py
https://insurance-claim-api.herokuapp.com/
1-All the null values were replaced by Null values.
2-All the duplicate values were removed.
3-Outliers were removed by applying IQR technique.
1-Different visualizations were plotted with the help of seaborn and matplotlib libraries to get a deep insight into the data.
2-Label Encoder was used on the categorical columns to convert them into numeric.
3-Extra trees classifier was used to get the feature scores and the features which had little impact on the model building were eliminated.
Features Abortion, Surg_Description, Emergencydept_yes/No, ethnicity, Admission_type, Area_Service, Weight_baby, Hospital County, Gender, Cultural_group, Payment_Typology, Mortality risk were dropped as they did not have great impact.
4-Smote oversampling technique was used to create the balance in the result column.
