Background
HomeUnion is a leading online marketplace for real estate investing, with fully managed properties sourced through due diligence and data analysis. It also provides post-purchase market intelligence, portfolio analysis and management oversight for investors.
HomeUnion simplifies real estate investing by removing the guess work out of investing. Using data science techniques, the company aims to empower investors to choose portfolios and acquire assets tailored to their investment goals
Challenges
To improve investment decision making, HomeUnion wanted to build data driven tools that helped investors assess investment options. The key challenges in doing this were:
Handling an extremely large volume of data
HomeUnion had a very large volume of data. Additionally, updated feeds needed reprocessing and made data processing complex
High variability in data sources and availability
For statistical modeling, HomeUnion needed data on multiple factors like crime rate, employment, income, schools, rental trends, etc. This data came from multiple sources and varied in its frequency, creating the need for a Big Data Pipeline Processing Factory
Statistical analysis of data
Each factor identified for quantifying investment decisions, depended on a number of attributes. Building algorithms to rate these attributes and creating statistical models for arriving at the relevant predictive indices was a need
Solutions
Talentica put in place a Data Science based solution that could collect, process, model and analyze data to generate accurate insights. These insights would enable predictive tools that helped investors in making better portfolio investment decisions.
A team of data scientists worked along with our product development team in order to work on this solution.
Raw Data Collection and Processing
Since data for all the identified features was not available from a single source, we developed an automated data ingestion ecosystem to receive data from multiple public/private and free/paid sources
APIs were configured to collect feature data, based on the frequency of its publishing/updation. This data was stored on our Big Data environment
To process such a large volume of data we put in place a Big Data based solution using Map/Reduce, Hadoop, R and Python systems to crunch very large volumes of data growing on a daily basis.
Exploratory Analysis and Feature Engineering
Using the cleaned and prepared data derived from data processing, our data scientists worked on readying the data for use by the algorithms that were to be built
The team brainstormed to come up with
- Factors that drive real estate investment decisions
- Features essential to the statistical model
Exploratory analysis and feature engineering techniques were used to reduce the number of features to only the most integral ones, that would produce the most accurate and relevant insights when modeled
Building Machine Learning Algorithms and Statistical Models
Taking the parameters selected from the feature engineering step, we used open source tools/programming languages R, Python and Hadoop, to build machine learning algorithms
Various regression and machine learning techniques were used to build models for:
- Neighborhood Investment Rating: Rating every US neighborhood for its investment potential
- REALestimate: Forecasting the return on investment on every property
- Predicting the right offer for anchoring an optimal winning bid
- Time-series based Price trends on various US geographies
- Predicting the likely rent for every property in the US
Data-Driven Insights and Analysis
The machine learning algorithms and statistical models were run using an automated flow built on the Big Data environement
The relevant feature data was then modeled to generate trends/forecasts for investment risk, sale price range, rent prediction and price appreciation, etc. at multiple geographic levels
Results
Feature-Rich Product
Partnering with Talentica, HomeUnion was able to launch an end to end solution for real estate investment portfolio management
From Acquisition to property management for buyers, market intelligence for buy/hold /sell decisions, and investment
plans tailored to achieve investor goals, HomeUnion does it all
Easier Investment Decisions Enabled More Deals
Backed by data driven insights and accurate assessment of investment worthiness, portfolio investment decisions became easier
This enabled HomeUnion to close more deals each month than before
Enabling Multiple Revenue Streams
With Talentica as a partner, HomeUnion was able to launch a multifaceted product as per their business plan, with revenue streams such as; property price predictions, property yield, sale price range and rent estimates
Services Provided
- Data Science
- Machine Learning
-
Big Data Engineering
- Predictive Analytics
- Real Estate Analytics
Download PDF
