My major contribution towards this project would be on the Deep Learning and NLP aspects of the project. I was responsible to build the classifier models from ground up.
My college curriculum experience has given me the chance to take courses that offer ML and DL application-based experiences inside the classroom. Some courses that stand out are CS 4033 Artificial Intelligence, CS 5173 Deep Learing, CS 2021 Python Programming and CS 5168 Parallel Programming. Python Programming was the base and stepping stone for my knowledge in machine learning and artificial intelligence. In Artificial Intelligence, I was exposed to interesting and powerful concepts in computer programming, such as CSP and MinMax Algorithms. Deep learning introduced me to the powerful concept of neural nets. Mainly CNN, RNN and Reinforcement Learning.
I was able to build upon on my Machine learning skills I acquired from classes and co-op experiences. I was able to combine the contextual learning of a 1D CNN along with the memory based learning of LSTMs. I also experimented with learned a new concept of Bidirectionality of LSTMs. All these combined lead me to create an optimum classifier model which was efficiently able to identify democrats with a 90% accuracy and republicans with 70% accuracy.
Our group was successfully able to create a holistic political prediction platform which is able to classify twitter users based on their complete profile. Compared to all other political classification or prediction systems, our model provides a more whole rounded prediction.
We were able to efficiently split the work load between the two of us, and since this is a research project, most of our time was spent with identifying and scouring for research papers. We were both able to learn a lot about deep learning models from these academic papers. The final stages of integration was harder due to the current pandemic situations. But we were able to work around that and we built a complete research project.