kaliabarkai / kaggledisastertweets

Interpretability implementations for the Kaggle Disaster Tweets challenge.

Kaggle Disaster Tweets Model Interpretability

This GitHub repository includes a logistic regression implementation on the Kaggle Disaster Tweets challenge with a focus on model interpretability tools such as LIME and SHAP.

The Disaster Tweets challenge is a natural language processing (NLP) data science problem where the model aims to predict whether or not a Tweet is referring to a real disaster based on the Tweet text body, the keywords of the Tweet and the location it was sent from.

In this repository, you will find the dataset used for this implementation in the Data folder, while the model notebook (model.ipynb) can be found in the Model folder, along with a sub-folder of the visualisations produced.

Additionally, for a more extensive overview of model interpretability, and how LIME and SHAP work, read the tutorial blogpost about this code.