Inspiration
This Hack-A-Thon project was very appealing because of the chance to use applied machine learning models. We found the data to be meaningful as it related to cancer research in the medical field.
What It Does
Our application connects to a .csv where survey results are stored on a server. As the results are brought in to our app, our model predicts which risk category patients fall under where that field is left blank. The model is then published on the web at http://www.olhsoftware.com/Dash in a web application where analysts can interact with the generated reports in a wiki fashion.
How We Built It
The application was hacked together using C# and Microsoft Core .Net technology. The underlying architecture is MVC, and the data sources and infrastructure are housed on the cloud. We used Bootstrap for our front end, DataFlow for the data cleaning and prediction, and PowerBI for the published reports.
Challenges We Ran Into
We had to create new Microsoft organization accounts to access the infrastructure necessary to solve this challenge. Additionally, we had to utilize a high capacity virtual machine, which is prohibitively expensive, for training our models. For our demonstration, we applied $200 worth of free credit towards training. This step can cost upward of $5,000 a month.
Accomplishments We're Proud Of
Our wiki web application has edit capabilities, so anybody who is interested or passionately curious can come interact with the model and try to visualize the data in new and insightful ways.
What We Learned
We learned more about the problems that practitioners run into involving data cleaning! Much time was spent parsing the raw data into a more usable form.
What's Next for Team Reest Cancer Data
Work can be done to improve the model by populating the data with more survey entries. The web application can be used from our git-lab to publish a more permanent solution to a dedicated domain.
Log in or sign up for Devpost to join the conversation.