Transactions on Machine Learning and Data Mining (Print ISSN: 1865-6781) (ONLINE-ISSN: 2509-9337)
Volume 12 - Number 1 - July 2019 - Pages 3-22
Measuring Classifier Performance with Risk and Error Matrix Charts
H.K. Koesmarno
Data Science and Engineering, ATO, Canberra, Australia
Abstract
Measuring classifier performance is important in machine learning. Risk charts and error matrix charts have been developed for this purpose. The strengths and weak-nesses of using these charts are outlined. Challenges with using these charts are dis-cussed including how base rates and using prevalence data for building models and incidence data for evaluating models affect model performance. A number of solu-tions for overcoming these challenges are covered.
Keywords: Error Matrix, Confusion Matrix, Cumulative Gain Chart, Risk Chart
Download Paper (379 KB)