Skip to content

Latest commit

 

History

History
20 lines (16 loc) · 1.04 KB

README.md

File metadata and controls

20 lines (16 loc) · 1.04 KB
Raw

Disaster Tweets Classifications by Machine Learning, which is currently Kaggle Competition.

  • train.csv and test.csv files can be found via https://www.kaggle.com/competitions/nlp-getting-started/data.
  • Columns in `train.csv' dataset are:
    • id
    • text
    • location
    • keyword
    • target
  • You will be predicting if tweet is a real disaster (1) or not (0).
  • Machine learning models such as LightGBM, XGBoost, RandomForest, and CatBoost Classifiers have been used to predict the disaster tweets.
  • RandomizedSearchCv is used to tune hyperparameters for models.
  • There is a commented out code in jupyter notebook in which you can combine other features with tf-idf matrix using hstack just in case of use if wanted.
Models LGBMClassifier CatBoostClassifier XGBClassifier RandomForestClassifier
Accuracy 0.7634 0.7706 0.7648 0.7873
  • RandomForestClassifier has demonstrated higher accuracy than rest of the models. Therefore, Test data is evaluated using RandomForestClassifier.