sultanbeishenkulov / home-credit-default-risk

The dataset train.csv contains the some information about the loans; the source is Kaggle https://www.kaggle.com/c/home-credit-default-risk. The aim is to build a model that can predict the payment performance for the next loan requests, based on some predictors.

Home Credit Default Risk Prediction

Dataset Description

The dataset train.csv contains information about loans, sourced from Kaggle: Home Credit Default Risk.

Each record corresponds to a specific dossier, uniquely identified by the ID variable. The TARGET variable represents the payment performance with values:

• 1: Customer with difficulties in repaying debt
• 0: Otherwise

The goal is to build a model predicting payment performance for the next loan requests based on various predictors.

Predictors

Socio-demo Info

• GENDER/ NAME_FAMILY_STATUS/ CNT_CHILDREN/ CNT_FAM_MEMBERS: Client's gender / family status / number of children / number of family members
• OCCUPATION_TYPE/ ORGANIZATION_TYPE: Client's occupation and employer's organization type
• DAYS_AGE: Client's age in days at the time of application
• DAYS_EMPLOYMENT: Days before the application the person started their current employment
• DAYS_ID_PUBLISH: Days before the application client changed identity document
• NAME_EDUCATION_TYPE: Level of the highest education the client achieved
• FLAG_PHONE/ FLAG_MOBIL/ FLAG_WORK_PHONE/ FLAG_EMP_PHONE: If client provided home / mobile / work / employer's phones
• FLAG_CONT_MOBILE: Was the mobile phone reachable
• FLAG_EMAIL: Did client provide an email
• REG_REGION_NOT_LIVE_REGION/ REG_REGION_NOT_WORK_REGION/ LIVE_REGION_NOT_WORK_REGION/ REG_CITY_NOT_LIVE_CITY/ REG_CITY_NOT_WORK_CITY/ LIVE_CITY_NOT_WORK_CITY: Various region and city-related flags

Income Info

• NAME_HOUSING_TYPE: Housing situation of the client (renting, living with parents, etc.)
• AMT_INCOME_TOTAL and NAME_INCOME_TYPE: Income of the client and client's income type
• FLAG_OWN_CAR (OWN_CAR_AGE) and FLAG_OWN_REALTY: Flag if the client owns a car (and its age) or a house (flat)

Behavioral Info

• NAME_TYPE_SUITE: Who was accompanying the client when applying for the loan
• WEEKDAY_APPR_PROCESS_START / HOUR_APPR_PROCESS_START: Day of the week / hour the client applied for the loan

Residence Info

The various _TAG suffixes correspond to the average (_AVG), modus (_MODE), and median (_MEDI) of:

• APARTMENTS_TAG: Apartments size
• YEARS_BUILD_TAG: Age of buildings
• ELEVATORS_TAG: Number of elevators
• ENTRANCES_TAG: Number of entrances
• FLOORSMIN_TAG / FLOORSMAX_TAG: Number of floors
• HOUSETYPE_MODE: House type
• TOTAL_AREA_MODE: Total area
• WALLSMATERIAL_MODE: Walls material
• REGION_RATING_CLIENT / REGION_RATING_CLIENT_W_CITY: Home Credit's rating of the region where the client lives / accounting for the city

Loan Info

• AMT_CREDIT: Amount of the loan
• AMT_ANNUITY: Loan annuity
• NAME_CONTRACT_TYPE: Indicates if the loan is cash or revolving
• AMT_GOODS_PRICE: For consumer loans, it is the price of the goods for which the loan is given

Bureau Scores

• EXT_SOURCE_1 / EXT_SOURCE_2 / EXT_SOURCE_3 / EXT_SOURCE_4: Normalized scores from an external data source
• AMT_REQ_CREDIT_BUREAU_tag: Number of enquiries to Credit Bureau about the client one hour (day, week, etc.) before application
• OBS_30_CNT_SOCIAL_CIRCLE / DEF_30_CNT_SOCIAL_CIRCLE: Number of observations of the client's social surroundings observed 30 days past due and how many defaulted
• OBS_60_CNT_SOCIAL_CIRCLE / DEF_60_CNT_SOCIAL_CIRCLE: Number of observations of the client's social surroundings observed 60 days past due and how many defaulted

Tools and Libraries Used

• numpy, pandas, seaborn, matplotlib, scikit-learn, imblearn