Exploration of Document Classification Techniques: Classifying news articles with CNN, Dense and Logistic models
The code in the various models can be used as a basis for more complex custom document classification systems. These models were benchmarked using the BBC news dataset in the data folder, with the following results:
- Dense network: accuracy 95.96%, evaluation time: 0.0076s
- CNN: accuracy 96.63%, evaluation time: 2.7s
- Logistic: accuracy 96.18%, evaluation time: 0.0015s