BIG DATA EMOTION CLASSIFICATION

Affecting our mind in terms of decision-making, influencing our moods and behaviors, emotion makes up a major part of our daily lives. People's physical and mental health and work status can be adversely affected by persistent negative emotions, but positive emotions can enhance subjective well-being and promote physical and mental health. I then formulated a question that can positive/negative emotion be automatically classified using a model, i.e, does the sentence contains enough information for the computer to make a sentiment judgment. With a massive amount of text data, I here build up an automatic emotion classification model that could read and distinguish sentences with negative emotions from sentences with positive emotions. Specifically, I studied the penalized logistic regression model with Stanford movie review data as the input. The AUC metric is used for model evaluation and outputted a promising out of sample score of 0.96.