Keywords: Machine learning, data mining, predictive modeling, analytics, artificial intelligence, AI, GPU, H2O, AutoML, automation, EDA, genetic algorithms, interpretability, visualization
Many industries tend toward automation. Why would analytics be any different? H2O Driverless AI uses bleeding-edge advances in statistics, machine learning, and computer science to efficiently and automatically visualize input data, engineer highly representative features, train accurate regressors and classifiers, and generate explanations of its own decisions. The Driverless AI system uses an automated implementation of the Grammar of Graphics to perform exploratory data analysis (EDA), an evolutionary learning scheme honed by Kaggle grandmasters to extract new features from input data, GPU-enable enabled XGBoost and stacked generalization to train complex gradient boosting machine (GBM) ensembles, and new techniques such as leave-one-covariate-out (LOCO) local feature importance and local interpretable model explanations (LIME) to generate reason codes for every system prediction. If you're interested in standardizing your data science practices, faster times to data-driven insights, and decreasing the cost of practicing data science, come learn about this InfoWorld Editor's Choice winning system for performing automated machine learning.