Overview

• This repository contains a data science project focused on understanding the key factors influencing US home prices over the last 20 years.

• The project utilizes the S&P Case-Schiller Home Price Index as a proxy for home prices and explores various economic indicators to build a predictive model.

• The chosen model is a Random Forest Regressor, achieving a remarkable R2 score of 99.87%

Problem Statement

• The objective is to identify and analyze factors that significantly impact home prices in the United States.

• By leveraging publicly available data, we aim to build a robust predictive model that explains the variations in the S&P Case-Schiller Home Price Index over the past two decades.

Main Features

The following features were selected for the predictive model:

Zillow Home Value Index (USAUCSFRCONDOSMSAMID):

• Median market value of all homes.
• Reflects overall trends in home values, providing insights into market conditions and price movements.

Average Sales Price for New Houses Sold (ASPNHSUS):

• Reflects the average cost of newly constructed homes.
• Influences perceptions of the affordability of new housing.

Median Sales Price for New Houses Sold (MSPNHSUS):

• Provides insights into the typical price of newly sold houses.
• Helps understand the distribution of new home prices.

Total Construction Spending: Residential (TLRESCONS):

• Reflects the level of investment in residential construction.
• Influences housing supply, potentially affecting home prices.

Gross Domestic Income (GDI):

• Influences economic conditions and affects consumer confidence in buying houses.

Consumer Price Index for All Urban Consumers: Housing (CPIHOSNS):

• Indicates inflation in housing costs, potentially impacting home prices.

Total Population: All Ages (POP):

• Population growth influences housing demand, potentially affecting home prices.

National Totals of State and Local Tax Revenue: Property Taxes (QTAXT01QTAXCAT1USNO):

• Property taxes influence the overall cost of living.

New Privately-Owned Housing Units Authorized (PERMIT):

• Authorized housing units can influence home prices and increase demand.

Monthly Supply of New Houses (MSACSR):

• Reflects the balance between housing supply and demand.

30-Year Fixed Rate Mortgage Average (MORTGAGE30US):

• Provides insights about a fixed interest rate for 30 years that buyers can afford for a loan.

University of Michigan: Consumer Sentiment (UMCSENT):

• Provides insights into consumer sentiment about the economy and the housing market.

Unemployed Population: Aged 25-54 (LFUN25TTUSM647S) and Unemployment Rate (UNRATE):

• Influences job security; more job security improves purchasing ability.

Housing Inventory Estimate: Vacant Units (EVACANTUSQ176N):

• Provides insights into vacant unit availability, market conditions, and supply and demand balance.

Federal Funds Effective Rate (FEDFUNDS):

• Changing in the federal reserve; adjusts the federal funds rate may influence the mortgage rate.

Model Process

Data Cleaning and Imputation:

• Utilized machine learning models to impute missing values in the dataset.

Exploratory Data Analysis (EDA):

• Conducted thorough EDA to understand the relationships between features and the target variable.

Feature Engineering:

• Engineered relevant features to improve model performance.

Model Selection:

• Chose the Random Forest Regressor based on its outstanding R2 score of 99.87%.

Model Evaluation:

• Utilized R2 score, Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE) for model evaluation.

Feature Importance :

• After training the Random Forest Regressor model, the feature importance analysis has done to reveal the predictor variables that contributes more when predicting target variable.

Tools Covered

• Programming Langauge : Python

• Code Notebook : Google Colab-notebook

• Data Collection: FRED (Federal Reserve Economic Data)

• Data Cleaning: Machine Learning Imputation

• Exploratory Data Analysis: Pandas, Matplotlib, Seaborn

• Model Building: Scikit-Learn (Random Forest Regressor)

• Model Evaluation: R2 Score, MAE, RMSE

Results

• Model Fit (R2 Score): 99.87%
• MAE (Mean Absolute Error): 1.39
• RMSE (Root Mean Squared Error): 2.37
• The Random Forest Regressor demonstrated superior performance in minimizing errors and capturing the variance in the target variable compared to other models