Assessing and Improving Prediction and Classification

Copyright © 2018 by Timothy Masters This book discusses methods for accurately and appropriately assessing the performance of prediction and classification models. It also shows how the functioning of existing models can be improved by using sophisticated but easily implemented techniques. Researchers employing anything from primitive linear regression models to state-of-the-art neural networks will find ways to measure their model’s performance in terms that relate directly to the practical application at hand, as opposed to mo re abstract terms that relate to the nature of the model. Then, through creative resampling of the training data, multiple training sessions, and perhaps creation of new supplementary models, the quality of the prediction or classification can be improvedsignificantly. re abstract terms that relate to the nature of the model. Then, through creative resampling of the training data, multiple training sessions, and perhaps creation of new supplementary models, the quality of the prediction or classification can be improvedsignificantly.