The inspiration for our project was a classic machine learning program used to teach the basics of machine learning in several courses. We took the traditional house price estimator application and integrated real time and historic stock price information to aid in our estimation of house value.
What it does
Our application takes variables that describe a house, as well as a time, and tries to create an estimate for the value of the house. The information used as our training data was gathered from a few towns surrounding LA, but we can apply the same model to several different data sets.
How we built it
We used python and the machine learning python package Sickit Learning. The primary function of the application uses Stochastic Gradient Descent to find the most probable solution.
Challenges we ran into
Cleaning the data set that we got to ensure that we were using reasonable values as training data was very important, and proved to be time consuming. Discovering how a new python package worked was difficult at first but we tried to work quickly to understand how sickit could aid our project.
Accomplishments that we're proud of
Integrating stock price information into the housing market data was very interesting to us.
What we learned
Machine learning practices and methods for estimating data based off of large training data sets.
What's next for Market Analytics with Machine Learning
Integrating other market data into the housing estimator, or changing the markets that the program works with to analyze data across multiple markets.