Real Estate Prediction in App
Chat Functionality [work in progress]
Real Estate Prediction Visualized with MatPlot
Real Estate Prediction Boxplot
Real Estate Histograph
Real Estate Data HeatMap
As we come from diverse backgrounds, including low-income and underrepresented minority groups, we are passionate about promoting equal opportunities for success though our love of programming.
"Over 99 percent of America's 28.7 million firms are small businesses. The vast majority (88 percent) of employer firms have fewer than 20 employees". [ JPMorgan Chase ]
With regards to creating equal opportunity, we often neglect the importance of providing a means for vertical movement regardless of one's background. We believe that these kinds of groups not only need opportunities to create a small business, but access to tools essential for expanding their business and growing it into something more profitable.
What it does
Our application would allow users to have access to various data in order to make informed decisions when trying to expand their businesses. It utilizes machine learning in order to predict the price of real estate, which is essential data for the long term when trying to expand ones business.
How we built it
We utilized web scraping with nltk to get housing data, used beautiful soup to extract data from the html, and loaded it into pandas for data analysis. From this, we removed information that was not a number and/or unreadable. Numpy was used for scientific computing purposes, and we mapped the latitude and longitude to zip codes 1000-1100 using geopy. We then utilized a regressive machine learning model with scikit-learn to form real estate preditions. Lastly, we visualized the data using matplot and tableau to allow those with varying levels of financial literacy to access and understand the same data. The results of the algorithm were put onto a web mock up of what our app would look like.
Challenges we ran into (╯°□°)╯︵ ┻━┻
Utilizing machine learning in conjunction with many different libraries including: numpy, pandas, scikit-learn, nltk, matplot, and geopy with the Google Maps API. Also we'd never done web scraping or machine learning outside of a decision tree model, so it certainly was a learning curve. We'd planned to use Android Studio and Xamarin, however due to time constraints we decided to create a web mock up instead. We attempted to implement the Google Cloud Platform, but had issues hosting the domain from domain.com properly due to ownership verification issues. We originally had difficulties with deciding the direction of the machine learning prediction model, but the mentors art CoStar and Capital One helped us to come up with a unique solution that we're truly proud of.
Accomplishments that we're proud of (ﾉ◕ヮ◕)ﾉ*:･ﾟ✧
We were able to use many libraries in conjunction with our machine learning algorithm, and made visualizations in tableau and matplot to better convey vast quantities of data. We also got the opportunities to use technologies we were interested in, and experiment despite not having used them before.
What we learned
We fought through several issues that took hours to fix, and had to revise our idea several times. We learned not only how to use various new technologies, but also how to be more adaptable in the face of unforeseen issues.
What's next for Mercury Launchpad
Allowing users to share ratings and reviews with others, streaming data through the machine learning pipeline to generate a property’s value estimation based on monthly economy listings.Also a series of visuals including a map representation of affordable listings allowing small businesses to view nearby available locations and budget easily when purchasing new property