Conventional Research Project submitted for the Computing Project (COMP90055 - 25 credit points) of Unimelb.
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
In this project, machine learning methods and neural networks are trained to classify and analyze sentiment of fake and real news on social networks.
/////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
All the codes are in the code folder;
The files in the relative path used in code are in the Desktop/dataset;
All the datasets are in Desktop/dataset and we also provide dataset.zip for downloading easily.
Harvester: harvester_fakenewsnet.py and harvester_credbank.py
Python machine learning: ML_ver.py, including data-preprocessing, featrue selection, Naive Bayes model training, Decision Tree model training, Random Forest model training and sentimental analysis.
Neural network model:BiLSTM.ipynb and Glove method:https://nlp.stanford.edu/projects/glove/
Sentimental analysis:Feature_Sentiment.ipynb which is used after all data being processed.
Test data set process: Testset_Preprocessing.ipynb which is only for data pre-processing used for test data.