Using WEKA for Classification without feature selection Copyright

Using WEKA for Classification (without feature selection) Copyright 2004 limsoon wong

Getting started Copyright 2004 limsoon wong

Load training data Þ Preprocess Þ Open file Þ Select training data file Copyright 2004 limsoon wong

Load training data • After file is loaded, its attributes are displayed Copyright 2004 limsoon wong

Load testing data Þ Classify Þ supplied test set Þ open file Þ select testing data file Copyright 2004 limsoon wong

Select classifier Þ Classifier Þ Select classifier – e. g. , j 48 or SMO Copyright 2004 limsoon wong

Perform classification Þ Start • After completion, accuracy, decision tree, etc. are displayed Copyright 2004 limsoon wong

Using WEKA for Classification (with feature selection) Copyright 2004 limsoon wong

Load training data Þ Preprocess Þ Open file Þ Select training data file Copyright 2004 limsoon wong

Load training data • After file is loaded, its attributes are displayed Copyright 2004 limsoon wong

Select feature selection method Þ Select attribute Þ Attribute evaluator Þ Select feature selection method – e. g. , chi square Copyright 2004 limsoon wong

Select feature selection method • Some feature selection method requires preprocessing of attributes before evaluation Þ Search method Þ Select attribute preprocessing method Copyright 2004 limsoon wong

Select feature from training data Þ Use full training set Þ Start • Scores of selected features are displayed • Selected features are listed Þ Highlight and copy desired features Copyright 2004 limsoon wong

Extract features from training data Þ Preprocess Þ Add filter Þ Select “Attribute Filter” • Paste selected attributes • Add “class” attribute Copyright 2004 limsoon wong

Extract features from training data Þ Apply filters • selected attributes are extracted from training data Copyright 2004 limsoon wong

Extract features from training data ÞSave • Save the extracted features into a new training data file Copyright 2004 limsoon wong

Extract features from testing data ÞPreprocess ÞOpen file ÞSelect testing data file Copyright 2004 limsoon wong

Extract features from testing data • Apply filter Copyright 2004 limsoon wong

Extract features from testing data Þ Save • Save extracted features into a new testing data file Copyright 2004 limsoon wong

Load extracted training data Þ Preprocess Þ Open file Þ Select training data file Copyright 2004 limsoon wong

Load extracted training data • After file is loaded, its attributes are displayed Copyright 2004 limsoon wong

Load extracted testing data Þ Classify Þ Supplied test set Þ Select new testing data Copyright 2004 limsoon wong

Perform classification Þ Classifier Þ Select classification method – e. g. , j 48 Copyright 2004 limsoon wong

Perform classification Þ Start • classification accuracy, decision tree, etc. are displayed Copyright 2004 limsoon wong

Enjoy! Copyright 2004 limsoon wong