NoisyDataCleaner

Python classes that identify and correct/remove noise in datasets

These models leverage on monte carlo simulation to approximate the correctness of a given label. The correction of the label builds on from the noise detection model.

Install:

pip install noisydatacleaner

Models:

1. NoiseRemover Identifies and then removes the noise from the dataset. Random Forest is used for smaller datasets as it yields better results. Whereas for larger datasets, k-Nearest Neighbors is much more efficient.

2. LabelClassificationCorrector Corrects the labels for classification datasets. Instead of only using 1 model like NoiseRemover, this model uses 5 different models:

models = {
   'Random Forest': RandomForestClassifier(n_estimators=128),
   'Extra Trees': ExtraTreesClassifier(n_estimators=128),
   'Linear Discriminant': LinearDiscriminantAnalysis(),
   'Logistic Regression': LogisticRegression(max_iter=128),
   'Neural Network': MLPClassifier(hidden_layer_sizes=(128,64,32))
}

All of which comes from the sklearn library