DESCRIPTION

For safe and secure lending experience, it's important to analyze the past data. In this project, you have to build a deep learning model to predict the chance of default for future loans using the historical data. As you will see, this dataset is highly imbalanced and includes a lot of features that make this problem more challenging.

Objective: Create a model that predicts whether or not an applicant will be able to repay a loan using historical data.

Domain: Finance

Analysis to be done: Perform data preprocessing and build a deep learning prediction model.

Steps to be done:

⦁ Load the dataset that is given to you ⦁ Check for null values in the dataset ⦁ Print percentage of default to payer of the dataset for the TARGET column ⦁ Balance the dataset if the data is imbalanced ⦁ Plot the balanced data or imbalanced data ⦁ Encode the columns that is required for the model ⦁ Calculate Sensitivity as a metrice ⦁ Calculate area under receiver operating characteristics curve

In financial year 2024, banks in India advanced over *** trillion Indian rupees in housing loans. This was an increase compared to the previous year. This reflected renewed homebuyer sentiment, as an increasing number of Indians were investing in buying residential property. Growth of home loans market Forty years ago, home loans were an alien concept. People would direct their provident fund savings and retirement benefits toward buying a home. However, three key institutions: HDFC, ICICI Ltd, and the State bank of India with their new lending concepts led to significant changes in the home loan market. Currently different commercial banks, NBFCs, and housing finance companies have flooded the mortgage market, and giving prospective home buyers from diverse strata of society with bargaining power and a chance at affording a home. Inflation and home loans   India is not untouched by global inflation. To address the problem, the Reserve Bank of India hiked the repo rate **** times since April 2022 to *** percent. Consequently, leading banks and housing finance companies raised their lending rates. For a prospective homebuyer, this meant a rise in tenure for home loans. In other words, equivalent monthly payments (EMIs)for homebuyers have lengthened and become more expensive. In financial year 2022, banks in India advanced around *** trillion Indian rupees in housing loans almost reaching pre-COVID levels.

Key mortgage statistics for second homes in 2024, including loan size, interest rates, and fees.

The Home Mortgage Disclosure Act (HMDA) database (Consumer Financial Protection Bureau, 2022) has compiled mortgage lending data since 1981, but the collection and dissemination methods have changed over time (Federal Financial Institutions Examination Council, 2018), creating barriers to conducting longitudinal analyses. This HMDA Longitudinal Dataset (HLD) organizes and standardizes information across different eras of HMDA data collection between 1981 and 2021, enabling such analysis. This collection contains two types of datasets: 1) HMDA aggregated data by census tract for each decade and 2) HMDA aggregated data by census tract for individual years. Items for analysis include borrower income values, mortgages by loan type (e.g., conventional, Federal Housing Administration (FHA), Veterans Affairs (VA), refinances), and mortgages by borrower race and gender.

HERA Section 1212k requires FHFA to prepare a Public Use Database containing information on their loan purchases at the Census Tract level.

Overview

This dataset comprises new residential mortgage statistics sourced from the National Mortgage Database (NMDB), which provides comprehensive details on mortgage market characteristics in the U.S. The data encapsulates a wide temporal span, starting from 1998 and extending up to 2022.

Acknowledgment

The information within this dataset has not been crafted or manipulated by any third party. It has been extracted directly from the official resources provided by NMDB. Users are requested to acknowledge the original authors and NMDB while using this dataset for their research or projects.

Citation: National Mortgage Database (NMDB), U.S. Federal Housing Finance Agency. New Residential Mortgage Statistics (1998-2022).

License

This dataset is shared under the original terms and permissions as provided by NMDB. It is essential for users to review any associated licenses or terms of use from the NMDB's official website before deploying the data in their projects.

Files

The dataset is divided into various files, each providing a distinct perspective on mortgage statistics:

1. Annual Data (1998-2021)

   • All Geographic Areas
   • National, Census Regions, and Census Divisions
   • States

2. Quarterly Data (1998 Q1 - 2022 Q3)

   • National, Census Regions, and Census Divisions
   • Files by Year

3. Monthly Data (January 1998 - September 2022)

   • National, Census Regions, and Census Divisions
   • Files by Year

4. Alternate Wide Format Files

For comprehensive information regarding each field within the dataset, users can refer to the Data Dictionary and Technical Notes provided by NMDB.

Dataset Image

The dataset's cover image is sourced from IDFC FIRST Bank's article on home loan eligibility and benefits.

The US Home Loan Market Report is Segmented by Loan Purpose (Purchase, Home Improvement/Renovation, Others), Provider (Banks, Housing Finance Companies, Others), Interest Rates (Fixed Interest Rates, Floating Interest Rates), and Loan Tenure (Less Than or Equal To 10 Years, 11 – 20 Years, and Longer Than 20 Years). The Market Forecasts are Provided in Terms of Value (USD).

These data relate to new mortgage lending on residential property in Ireland on an annual basis. Data relates to those institutions [(banks and non-bank mortgage lenders)] who issue at least €50 million of new mortgage lending in a six-month period and are subsequently required to submit loan-level information to the Central Bank for the purposes of the macroprudential mortgage measures. The value and volume of new lending is provided, by borrower type, along with the distribution of lending by Loan-to-value and Loan-to-income ratio. Average characteristics are also provided. These data do not constitute official statistics. These data are published to support transparency and understanding of market developments.

As of March 2025, ICICI Bank provided the lowest interest rates for its home loans in India, with an average of **** percent. Bank of Maharashtra accounted for the highest interest rate with an average of **** percent.

In 2024, the value of personal housing loans in China amounted to ************** yuan, representing a slight drop of *** percent compared to the previous year. The overall value of outstanding mortgages more than doubled between 2016 and 2021 before it plateaued afterwards. A key factor to the growth of the real estate market China's personal housing loan market emerged in the 1990s in tandem with the marketization of the country's real estate sector. Its subsequent expansion also mirrored the growth in the property industry. Thanks to the dramatic rise in home prices across China since the early 2000s, substantial capital has poured into the market through real estate development loans and personal housing credits. For almost two decades, many Chinese middle class citizens accumulated their personal wealth through the considerable appreciation of their properties, which they financed with the help of mortgages. Risks The persistently high level of outstanding personal mortgage is becoming increasingly concerning amidst China’s current economic and market situation. With the country’s economic slowdown and the oversupply in the property sector, the housing market is losing steam, resulting in elevated risks of bad debts to financial institutions. At the same time, the household debt in China is now staying above ** percent of the country’s GDP, undermining the ability to consume and invest in the Chinese population.

Home Loan Program Information Update Frequency: Monthly

The National Mortgage Database (NMDB®) is a nationally representative five percent sample of residential mortgages in the United States.

Dataset contains the percent of denied mortgages based on the purpose of the application and disaggregated by race. Each cell represents the denial rate within that column's race/ethnicity category's total applications. Data pulled from the Consumer Financial Protection Bureau, collected by the Home Mortgage Disclosure Act, which requires many financial institutions to maintain, report, and publicly disclose information about mortgages.

China Consumer Loan: Residential Housing Mortgage Loan data was reported at 25,750.000 RMB bn in 2018. This records an increase from the previous number of 21,860.500 RMB bn for 2017. China Consumer Loan: Residential Housing Mortgage Loan data is updated yearly, averaging 2,473.416 RMB bn from Dec 1997 (Median) to 2018, with 20 observations. The data reached an all-time high of 25,750.000 RMB bn in 2018 and a record low of 13.100 RMB bn in 1997. China Consumer Loan: Residential Housing Mortgage Loan data remains active status in CEIC and is reported by The People's Bank of China. The data is categorized under China Premium Database’s Money and Banking – Table CN.KB: Loan: Consumer Loan.

30 Year Mortgage Rate in the United States decreased to 6.23 percent in November 26 from 6.26 percent in the previous week. This dataset includes a chart with historical data for the United States 30 Year Mortgage Rate.

The period from March to August 2020 saw noticeably weaker demand for home loans in Poland. This state results from the outbreak of the coronavirus (COVID-19) pandemic in Poland in March 2020. The highest decline in the demand for housing loans was recorded in August 2022 due to the high inflation. One of the main reasons for the decline in demand was also tightening and rising interest rates that reduced the creditworthiness of potential borrowers. As of June 2025, the demand for mortgage loans continues to evolve, with the latest data indicating a value of **** percent. Loan market in Poland In 2023, the net value of loans to households in Poland was *** billion zloty, representing a decrease of *** percent over the previous year. In this period, in terms of credit type, the highest value was achieved by real estate loans granted to households. Their value amounted to over *** billion zloty, followed by consumer loans with a total value of nearly *** billion zloty. Mortgage loans in Poland In 2023, there has been a continuing trend of decreasing the popularity of housing loans up to 100,000 zloty. A growing number of Poles have taken out loans of the value of over ******* zloty. And the average value of a mortgage loan in 2023 increased by ** percent compared to the previous year and amounted to nearly ******* zloty. In this period, Poland’s number of active mortgage loans also decreased, reaching nearly *** million.

The Housing and Economic Recovery Act of 2008 (HERA) requires the Federal Housing Finance Agency (FHFA) to submit an annual report to Congress on the collateral pledged to the FHLBanks, including an analysis of collateral by type and by Bank district.3 FHFA’s Report on Collateral Pledged to Federal Home Loan Banks provides the required information as well as additional analysis of data on the types and amounts of collateral pledged to the Banks to secure advances and other collateralized products offered by the Banks to their members. The information in this report uses data collected through a quarterly data collection conducted by FHFA’s Division of Federal Home Loan Bank Regulation (DBR).

United States Home Mortgage: Loan-to-Price Ratio data was reported at 78.400 % in Oct 2018. This records an increase from the previous number of 78.000 % for Sep 2018. United States Home Mortgage: Loan-to-Price Ratio data is updated monthly, averaging 75.800 % from Jan 1973 (Median) to Oct 2018, with 550 observations. The data reached an all-time high of 80.800 % in Dec 1994 and a record low of 68.640 % in May 1982. United States Home Mortgage: Loan-to-Price Ratio data remains active status in CEIC and is reported by Federal Housing Finance Agency. The data is categorized under Global Database’s United States – Table US.KB011: Home Mortgage Terms.

Fair lending statistics from 100+ million mortgage applications showing approval rates and demographic patterns by Homebuyer.com.

Home Mortgage Disclosure Act (HMDA) requires many FIs to maintain, report, and publicly disclose information about applications for and originations of mortgage loans. HMDA s purposes are to provide the public and public officials with sufficient information to enable them to determine whether institutions are serving the housing needs of the communities and neighborhoods in which they are located, to assist public officials in distributing public sector investments in a manner designed to improve the private investment environment, and to assist in identifying possible discriminatory lending patterns and enforcing antidiscrimination statutes.