Requirements: Basic knowledge of math, statistics and databases.
The world does not behave in a linear fashion. Boundaries are not straight lines, and this applies also with data, behaviors and boundaries in classifications. Linear models allow us to run agile analyses but with lower accuracy than non-linear models. In this course, we will review non-linear models, applying them to real examples, so that we can compare the results against linear models. The results obtained will be discussed in groups, as well as the performance of the selected models. The results obtained will be compared graphically to clarify any doubts that may exist regarding the superiority in accuracy of the models, as well as the limitations in terms of processing, remembering the balance between performance and accuracy. Validation schemes for predictive models will be presented, including examples in RapidMiner, to understand the confusion matrix and the validity of the different methods. Complex analytical projects require the combination of several advanced analytical and statistical models, called ensembles. Examples of these will be reviewed.