fact_check_worthy_claim

Description

This is the code repository accompanying the mandatory project in the course DAT550 - Data Mining and Deep Learning, University of Stavanger, Spring 2023.

Requirements

Please note that if you're using the conda package manager, you may need to enable the conda-forge repo to install the required versions of the pandas and fastparquet packages.

conda config --add channels conda-forge
conda config --set channel_priority strict

Confirm that the conda-forge channel has been added and set to highest priority.

conda config --show channels

The output should look similar to this (notice that conda-forge is listed above the defaults channel):

channels:
  - conda-forge
  - defaults

Then, install the required packages from requirements.txt

conda install --file requirements.txt

pip install -r requirements.txt

Spacy
After installing the requirements you need to download the Spacy model:

python -m spacy download en_core_web_sm

Data Exploration

See the data_exploration.ipynb notebook. Here we do initial data exploration, visualisations and correlation.

Baseline Model

See the baseline_model.ipynb notebook. Here we establish a baseline model using the scikit-learn RandomForestClassifier with default settings.

Feature Generation

See the feature_generation.ipynb notebook. Here we generate the features used in the classifiers.

Random Forest Classifier

See the RandomForest.ipynb notebook. Here we attempt improve on the RandomForestClassifier results by adding features and tuning parameters.

SVM

See the SVM.ipynb notebook. Here we use the SVM model to see if we can improve the results. We use the same features as created for the RandomForestClassifier, and also try tuning parameters for improved results.

AdaBoost Classifier

See the AdaBoost.ipynb notebook. Here we use the AdaBoost model to see if we can improve the results. We use the same features as created for the RandomForestClassifier, and also try tuning parameters for improved results.

GradientBoosting Classifier

See the GradientBoosting.ipynb notebook. Here we use the GradientBoosting model to see if we can improve the results. We use the same features as created for the RandomForestClassifier, and also try tuning parameters for improved results.

NaiveBayes Classifier

See the NaiveBayes.ipynb notebook. Here we use the NaiveBayes model to see if we can improve the results. We use the same features as created for the RandomForestClassifier, and also try tuning parameters for improved results.

Results

See the Results.ipynb notebook. Here we have combined some of the results from the model-specific notebooks for comparisons.

SigneK42/fact_check_worthy_claim