hashojaei / shinyhome

Real Estate Market Forecasting and Analytics

shinyHome

Real Estate Market Forecasting and Analytics

Introduction

shinyHome allows the user to explore real estate market statistics and to employ the most acknowledged time series forecasting algorithms to predict home values for up to 10 years, for over 20,000 markets. Users will:

• Explore current and historical median home value data,
• Analyze price trends using time series decomposition techniques,
• Create and evaluate prediction modules using eight time series forecasting algorithms, and
• Forecast home values using the prediction algorithms.

Configuration

The application was written on R 3.3.1 for Windows.

Requirements

This application requires the following R packages.

• datasets
• dplyr
• forecast
• ggplot2
• plotly
• plyr
• rCharts
• shiny
• shinydashboard
• TTR
• xlsx

File Manifest

The file manifest is as follows:

• currentCity: Current home value index and growth rates by city
• currentCounty: Current home value index and growth rates by county
• currentState: Current home value index and growth rates by state
• currentZip: Current home value index and growth rates by zip code
• geo: State, county, city and zip code cross-reference file
• hviAllCity: Historical home value data by city
• hviAllCounty: Historical home value data by county
• hviAllState: Historical home value data by state
• hviAllZip: Historical home value data by zip code
• models: Descriptions of forecasting algorithms employed

Copyright

©John James, 2016

Contact

Developer: John James [email protected]

Known Bugs

When running the structural model for time series by maximum likelihood (StructTS) the application occasionally throws the following error: Error in optim(start, f, method = method, hessian = TRUE, ...) : L-BFGS-B needs finite values of 'fn'

Credits and Acknowledgements

• Huge acknowledgement to Zillow Research for the data.
• Ramnath Vaidyanathan for a beautiful charting package
• Joe Cheng for his active support in the githubsphere
• Avril Coghlan for the not so little book on R for time series.
• The universe of shiny programmers that seem to have already asked and answered all my questions before I knew I had them.