Tensorflow model training
Histogram of Twitter search for "nike"
Application GUI

SharkBait

RamHacks Sentiment Analysis

We got a running GUI for a sentiment analysis platform, and we separately made and trained a dense, 6-layer neural network to work with the input comments, tweets, etc... It is currently able to pull data from reddit and twitter, with a working instagram implementation. That being said, we lack the access key to pull data off of instagram at this time.

It was trained using IMDB's movie review dataset, equalling out to an accuracy of around 87%. We felt this was adequate as there was not a whole lot of time nor processing power to fiddle around with the model. If we did have the time, we would have written a custom genetic algorithm to evolve the neural network to benefit the performance and adaptability of the program.

With both of these up and running, we were unable to adapt the application to support the neural network in time, however with a little more time (and some sleep), it would be up and running in no time.

Team Members:

Aaron Causey: causeyac20@mail.vmi.edu; Kathleen Yates: yateskm20@mail.vmi.edu; Robert DuPont: dupontrw20@mail.vmi.edu; Zachary Harris: harriszj20@mail.vmi.edu;

API usage: Tweepy, Tensorflow, Pickle, Enchant, MatPlotLib, tkinter, seaborn, nltk, pandas, textblob, praw