sample machine learning for binary text classification

Objective

This sample demonstrate how to use ML to classify user humor in like (1) or dislike (0) using supervised learning.

This is a binary classification but you can extend to multiples categories using some data techniques and changing the model from binary_crossentropy to categorical_crossentropy

What you need

• python 2.7 ~ 3.x (used v 3.7.6)
• keras
• tensorflow OR theano OR cntk (used tensorflow)
• some python libs (see bellow)

install tensorflow

# Current stable release for CPU and GPU
% pip3 install tensorflow --user

install python libs

pip3 install pandas 
pip3 install sklearn
pip3 install matplotlib

Run

% python3 classify.py

Result

Using TensorFlow backend.
Training Accuracy: 0.9973
Testing Accuracy:  0.8400