Tweet-classification

This project experiments different models perfomance in predicting whether a given tweet is about a real disaster or not.

The data used for this tutorial can be downloaded from Kaggle by following this link. The dataset contains text, keywords (may be blank), location (may be blank), a unique identifier for each tweet and the corresponding label.

The training data can be found in train.csv while the test data used for prediction is in test.csv; the python code is in the notebook and the model is saved in the cache folder. Necessary packages are stored in requirements.txt. All code is written in Python 3.

Models tried out were logistic regression, gaussian classifier and a Recurrent Neural Network built using Keras. Hyperparameters were tuned using GridSearchCV model selection and GradientBoostingClassifier to improve performance.

To run the model

Clone the repo

git clone https://github.com/AniekanInyang/tweet-classification

Install packages

pip install -r requirements.txt

Load the model

from keras.models import load_model
model = load_model(os.path.join(os.path.join("cache", "sentiment_analysis"), "rnn_model.h5"))

I wrote about the work I did for this project in Part I and Part II on my blog.