COM 328 PREDICTIVE DATA MODELING

An introduction to the computational modeling and critical assessment used for prediction problems. Topics in this course include data visualization, transformation, and imputation along with general linear models, randomized tree methods, maximum margin separators, and modern neural network forms (this includes deep learning). Unsupervised and supervised methods will be introduced along with semi-supervised methods. The course discusses normalization, regularization, validation, selection, and calibration (estimation of model performance on new unseen data). All examples and methods are demonstrated in the Python programming language and emphasis is placed on experimental Machine Learning application to analysis-ready heterogenous categorical, ordinal, and numerical multivariate data.