ABSTRACT
A model's performance is influenced by several components, including training data, input attributes, learner technique, and learner parameter settings to name just a few. Each component that in some way dictates model performance is a candidate for evaluation. This chapter presents the discussion of performance evaluation by focusing on formal evaluation methods for supervised learning and unsupervised clustering. It emphasizes the practical application of standard statistical and nonstatistical methods rather than the theory behind each technique. The chapter highlights the component parts of the data mining process that are responsive to an evaluation. It provides an overview of several foundational statistical concepts such as mean and variance scores, standard error computations, data distributions, populations and samples, and hypothesis testing. The chapter shows how to use RapidMiner's T-test and ANOVA operators to compare the performance vectors of competing models. RapidMiner provides several statistical analysis tools to help us formally evaluate the goodness of the data mining models.
