Prediction Intervals for Class Probabilities
Citation
Export citationYu, X. (2007). Prediction Intervals for Class Probabilities (Thesis, Master of Science (MSc)). The University of Waikato, Hamilton, New Zealand. Retrieved from https://hdl.handle.net/10289/2436
Permanent Research Commons link: https://hdl.handle.net/10289/2436
Abstract
Prediction intervals for class probabilities are of interest in machine learning because theycan quantify the uncertainty about the class probability estimate for a test instance. Theidea is that all likely class probability values of the test instance are included, with apre-specified confidence level, in the calculated prediction interval. This thesis proposes aprobabilistic model for calculating such prediction intervals. Given the unobservability ofclass probabilities, a Bayesian approach is employed to derive a complete distribution of theclass probability of a test instance based on a set of class observations of training instancesin the neighbourhood of the test instance. A random decision tree ensemble learningalgorithm is also proposed, whose prediction output constitutes the neighbourhood thatis used by the Bayesian model to produce a PI for the test instance. The Bayesianmodel, which is used in conjunction with the ensemble learning algorithm and the standardnearest-neighbour classifier, is evaluated on artificial datasets and modified real datasets.
Date
2007Type
Degree Name
Publisher
The University of Waikato
Rights
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.
Collections
- Masters Degree Theses [2496]