Akvelon | Case Study: Real Estate Machine Learning Project
21224
post-template-default,single,single-post,postid-21224,single-format-standard,ajax_fade,page_not_loaded,,qode-theme-ver-8.0,wpb-js-composer js-comp-ver-4.9.2,vc_responsive
 

04 Dec Case Study: Real Estate Machine Learning Project

Business Need

For Akvelon’s Hackathon, a team of Akvelon employees built an Azure Machine Learning demo project for the real estate domain. Our goal was to build a service that could predict the fair market value of residencies based on all vital characteristics of a neighborhood. This project can be integrated with real estate search engines (Trulia, Zillow, etc.) and can help users eliminate properties that are not worth considering. Should the users be aware of upcoming changes to the city development, the service can help users predict future price changes.

Solution

Our team developed an Azure ML service to predict housing prices and cluster results for individuals searching for a new house. The data from Kaggle’s competition was used to create regression and cluster ML models in Azure ML Studio. We used additional information about neighborhood safety and distance to nearest schools, parks, and malls to create more accurate predictions.

The houses are shown in a map and users can select different clusters by:

1) Price and quality

2) Neighborhood safety

3) Distance to schools, parks, malls, etc.

Also, the user can select recommendations for families with children, for which square footage, safety, and distance to schools are important as well as recommendations for young homeowners, for which price and distance to parks and malls are important.

 

Clustering works as the following: the system predicts how suitable a house is for users and clusters the data for each individual family’s preference. The clusters are color coded and each color means something different: green means that the house is the right fit for the user, red means that house is not the right fit for the user, and yellow means the house is somewhere in the middle.

Benefits and Results

We successfully created a prototype application which shows our experience in Azure ML Services and the real estate domain and can be used to demonstrate how versatile our team is to potential customers. Our team was even selected as a grand prize winner at the Akvelon Hackathon.

Results of this system:

  • Created a re-trainable Azure ML Regression model which can be applied to existing data as well as new house data to predict housing prices
  • Created an Azure ML Clustering model to cluster houses by different combinations of parameters
  • Use of Python and SQL scripts for data preprocessing
  • Developed a website which allows user to:
    • register their email
    • search for available houses with descriptive details on each listing
    • select different recommendations based on preferences
    • add region alerts that notify the user through email when a new house appears in a desired region
    • create a new house sale listing, obtain predicted price, and save it in system
  • Web API services allow users to:
    • get data from Azure SQL
    • get additional information on a new house
    • find distances to schools, parks and malls, and neighborhood safety information
    • determine if a new house was created in an existing search region created by the user
    • receive email notifications

 

Products and Technologies used

Azure ML, Azure SQL, Asp.Net MVC, Unify Website Template, Web API, Bing Maps, Leaflet, Python, SendGrid

Tags:
, ,


Akvelon

Akvelon

We Make BIG Software Happen. Contact Us