Open marcdotson opened 5 years ago
Currently we have tried Logistic Regression (~80% accuracy) and Random Forest (~84% accuracy). Still working on testing out the CNN and we hope to have everything set for initial modeling by Saturday, October 18th.
Next step would be to build out the classifier for complaint types.
Let's build out a table comparing the predictive accuracy of the different techniques as a markdown table in the paper, @AdrielC.
Using the labeled test data, we can make sure we're using the best approach to classifying tweets as complaints.
Identifying different types of complaints will come next.