Inspiration
The inspiration comes from the fact that most modern machine learning platform is inherently black box (for a low dimensional problem, one could still theoretically show how the algorithm works with worked examples). On the other hand, a Bayesian Network model, while still in the category of machine learning, leverages transparent probability theory to make predictions.
What it does
The model output is represented as an expected total COVID 19 case in the US for both the Bayesian Network model and the Neural Network Model.
How we built it
We used the American Community Survey to stream the data of the relevant variables that we used as input for the models. The data is transformed into the probability space to inform the Bayesian Network model and its Conditional Probability Table. For the neural network, the data is streamed as a 1-D vector as training set and test set.
Challenges we ran into
We did this challenge as part of the Microsoft Cloud challenge project. We wanted to leverage the power of Azure to host our webapp that we have built to showcase our results but we did not manage to get the webapp up and running.
Accomplishments that we're proud of
The dashboard and the fact that we managed to build two independent models and benchmark them.
What we learned
The predictive capability for the two network is comparable albeit further deep dive analysis is required.
What's next for Modeling Benchmark - A Case Study of COVID 19 Cases
Model validation and scenarios simulation
Log in or sign up for Devpost to join the conversation.