Skip to content

Latest commit

 

History

History
28 lines (18 loc) · 1.71 KB

README.md

File metadata and controls

28 lines (18 loc) · 1.71 KB

MBTI-Personality-Classifier

This project uses Natural Language processing to classify a person's personality based on their social media posts.

For additional information please checkout the presentation pdf, or ppt file in the repository.

MBTI persnality classifies a person's personality into 4 broad categories, namely:

  • Extroversion and Introversion
  • Sensing and Intuition
  • Thinking and Feeling
  • Judging and Perceiving

The data was imported from Kaggle: https://www.kaggle.com/datasnaek/mbti-type

The tokenization was done using NLTK toolk, and the model was based on Bayesian Model in Natural Language processing.

For this particular problem, i trained 4 separte models each one predicted one of the attributes. The final output is based on percentage certainty the model has. Divy Bramhecha Personality Type by the classifier when feed in quora answers (Output using matplotlib) This worked quite well in classifying my personality type, It's INTJ, although you can see that the F:J is 55:45 which is quite close. Unfortunately, some of the personality types had a lot more data, and so the model often fails to classify a personality with the lower end.

Despite that, the model has an accuracy of 80% on train:test split of 80:20 for each of the attributes.

Hopefully, in future I will try with a bigger dataset, maybe try to extract data from Reddit posts.

Here is my solution to the dataset on Kaggle: https://www.kaggle.com/tgdivy/mbti-personality-classifier

Note: The files have been reogranised, so you may have to change the output directories when you run the code.