Eric Bai
05/04/2016
Future Invest: Find best under market value properties
Reason:
It is very time consuming when researching the market in order to make a knowledgable investment into real estate. There is no good analysing tool to help people easily find under valued properties.
Data Pipeline and Modeling Steps
-
scrapped the property information on ziprealty.
-
Select about 15 important features to fit into the model. Here is a summary of most important features. Location and size are the two most important features.
-
Ran a random forest model with selected features with cross validation to find best parameters
-
Returned estimated market prices for all properties in the San Francisco Bay area
-
Created a web app for user to be able to use this tool
Next steps: -
Get data from the whole US
-
Find other possible important features
-
Include photos as features with nerual network because people can tell if the property is good or not based on pictures
-
Create a mobile app
