That is an open project of noise pollutions classification.
make_features.py
script calculates features of the 4-second snippets which represent noise pollution for one of the categories:
- Highways;
- Railway;
- Human flow;
- Construction.
After calculation, all the features are recorded to a dataframe with .csv
extension. The following pipeline makes classification basing on classical ML methods (noises_classification.py
). It includes Random Forest, Decision Tree classifiers, KNN as well as regularization for it.
The highest correctness of classification is 80% approximately.