Jul 28, 2017 · 1 min read
Redwan Huq: Thanks for the feedback.
The reason i choose Naive Bayes and SVMs are 2:
- Naive Bayes: Often tends to word with clean data for classification and also easy to understand for the starters.
- As i have read in other blogs, SVMs are very good for text classification. But as you saw after optimization of Naive Bayes, SVM didn’t out performed. Again this depends on the dataset.
Also, i will like to see what are the performance results with other machine learning algorithms or any optimization trick. So, feel free to share.
Regarding accuracy, there is no reason TBH. I just didn’t get time to do other stuff, i will probably do in my next blog, for sure :)
The most predictive n-gram as you can see during the grid search was the bi-grams for the given dataset.
