RealPredictions predicts the number of real estate sales in a particular month based on economic factors including interest rate and consumer confidence. We implemented a machine learning technique in which we applied linear regression to calculate the predicted number.
How we built it
Using Microsoft Azure's Machine Learning Studio, we constructed the environment for how we wanted the prediction to go. Using Visual Studio and SSMS (SQL Server Management Studio), we sorted through various mock datasets to highlight the data we ultimately needed. We then imported the data, cleaned up our dataset, split it 75% to 25%, and trained the model. We tested several datasets in our framework. Our final product was trained based on city and federal public records. Lastly, we deployed our app so it is accessible as a web service. We also visualized the data using Python's Matplotlib library to give us a first impression and baseline of our data.
Challenges we ran into
The platform we used to help build our app, Microsoft Azure, was relatively new to us since neither of us had solid previous experience, however, we quickly learned how intuitive the platform was for developers. Furthermore, machine learning itself as a skill was new to us so we had to learn the components of how to train and use the data properly and correctly. Finding relevant data was the most challenging part of our project.
Accomplishments that we're proud of
We learned a great amount of machine learning topics and how they would apply to real-world data. We were glad to have had a chance to learn a powerful tool that offers a wide variety of services for developers that we have not yet fully explored.
What we learned
We learned more about how machine learning works and how it can carry out various kinds of regressions. During the process of solidifying our project idea, we explored various pathways of developing our product including sorting through our data using SQL on SSMS and Visual Studio. We learned how to sort, clean, and train our data, and convert our model into a usable web service.