Cost-sensitive classification problem (Poster)
Tgk Toon Calders, Mykola Pechenizkiy · TU/e Research Portal · 2012
In practical situations almost all classification problems are cost-sensitive or utility based one way or another. This exercise mimics a real situation in which students first have to translate a description into a datamining workflow, learn a prediction model, apply it to new data, and set up a testing strategy to estimate what will be the performance. The exercise is suitable for students following an introductory data mining course; it has been used in my introductory data mining class (3ECTS; 3rd BSc Computer Science students) for two years now. Students work on it in class for approximately 1 hour and finish the exercise at home. Solutions are to be sent to the lecturer and discussion the solutions the next lecture takes approximately 30 minutes.