The DMWay Analytic Engine is the core of DMWay technology. Based on innovative expert system, the Analytic Engine mimics the way an experienced data scientist would go about building large scale predictive models to create fully automatic suite of models. Built in a modular way, the suite of models will eventually span a whole range of models to answer the multitude of prediction issues encountered in practice, ranging from regression-based models, to decision trees, clustering and collaborative filtering, time series analysis, and others. The toughest problem in building large scale Predictive Analytic models is the feature selection problem – choosing the handful of predictors “explaining” the phenomenon under study from a much larger set of potential predictors that may consists of hundreds, if not thousands, of predictors. This is a huge combinatorial problem that we solve in a fully automatic manner, by means of a well-structured process, using the best practice in each step of the way, and resolving modeling and data issues (nonlinearity, outliers, missing values, numerical issues, convergence problems, statistical issues, decision issues,…) internally and “on the fly”. The resulting model is validated for accuracy, over fitting and profitability (ROI) to make sure we get accurate and precise models that are stable, consistent and free of any subjective bias. By and large, the models created by the Analytic Engine are much better than models build by expert data scientists.