Practical Machine Learning with H2O - Cook, Darren Pszów

Machine learning has finally come of age. With H2O software, you can perform machine learning and data analysis using a simple open source framework that...s easy to use, has a wide range of OS and language support, and scales for big data. This hands-on guide teaches you how to use H20 with only minimal math and theory behind the learning algorithms.If you...re familiar with R or Python, know a bit of statistics, and have some experience manipulating data, author Darren Cook will take you through H2O basics and help you conduct machine-learning experiments on different sample data sets. You...ll explore several modern machine-learning techniques such as deep learning, random forests, unsupervised learning, and ensemble learning.Learn how to import, manipulate, and export data with H2OExplore key machine-learning concepts, such as cross-validation and validation data setsWork with three diverse data sets, including a regression, a multinomial classification, and a binomial classificationUse H2O to analyze each sample data set with four supervised machine-learning algorithmsUnderstand how cluster analysis and other unsupervised machine-learning algorithms work Spis treści: Preface Who Uses It and Why? About You Conventions Used in This Book Using Code Examples OReilly Safari How to Contact Us Acknowledgments 1. Installation and Quick-Start Preparing to Install Installing R Installing Python Privacy Installing Java Install H2O with R (CRAN) Install H2O with Python (pip) Our First Learning Training and Predictions, with Python Training and Predictions, with R Performance Versus Predictions On Being Unlucky Flow Data Models Predictions Other Things in Flow Summary 2. Data Import, Data Export Memory Requirements Preparing the Data Getting Data into H2O Load CSV Files Load Other File Formats Load Directly from R Load Directly from Python Data Manipulation Laziness, Naming, Deleting Data Summaries Operations on Columns Aggregating Rows Indexing Split Data Already in H2O Rows and Columns Getting Data Out of H2O Exporting Data Frames POJOs Model Files Save All Models Summary 3. The Data Sets Data Set: Building Energy Efficiency Setup and Load The Data Columns Splitting the Data Lets Take a Look! About the Data Set Data Set: Handwritten Digits Setup and Load Taking a Look Helping the Models About the Data Set Data Set: Football Scores Correlations Missing Data And Yet More Columns How to Train and Test? Setup and Load The Other Third Missing Data (Again) Setup and Load (Again) About the Data Set Summary 4. Common Model Parameters Supported Metrics Regression Metrics Classification Metrics Binomial Classification The Essentials Effort Scoring and Validation Early Stopping Checkpoints Cross-Validation (aka k-folds) Data Weighting Sampling, Generalizing Regression Output Control Summary 5. Random Forest Decision Trees Random Forest Parameters Building Energy Efficiency: Default Random Forest Grid Search Cartesian RandomDiscrete High-Level Strategy Building Energy Efficiency: Tuned Random Forest MNIST: Default Random Forest MNIST: Tuned Random Forest Enhanced Data Football: Default Random Forest Football: Tuned Random Forest Summary 6. Gradient Boosting Machines Boosting The Good, the Bad, and the Mysterious Parameters Building Energy Efficiency: Default GBM Building Energy Efficiency: Tuned GBM MNIST: Default GBM MNIST: Tuned GBM Football: Default GBM Football: Tuned GBM Summary 7. Linear Models GLM Parameters Building Energy Efficiency: Default GLM Building Energy Efficiency: Tuned GLM MNIST: Default GLM MNIST: Tuned GLM Football: Default GLM Football: Tuned GLM Summary 8. Deep Learning (Neural Nets) What Are Neural Nets? Numbers Versus Categories Network Layers Activation Functions Parameters Deep Learning Regularization Deep Learning Scoring Building Energy Efficiency: Default Deep Learning Building Energy Efficiency: Tuned Deep Learning MNIST: Default Deep Learning MNIST: Tuned Deep Learning Football: Default Deep Learning Football: Tuned Deep Learning Summary Appendix: More Deep Learning Parameters 9. Unsupervised Learning K-Means Clustering Deep Learning Auto-Encoder Stacked Auto-Encoder Principal Component Analysis GLRM Missing Data GLRM Lose the R! Summary 10. Everything Else Staying on Top of and Poking into Things Installing the Latest Version Building from Source Running from the Command Line Clusters EC2 Other Cloud Providers Hadoop Spark / Sparkling Water Naive Bayes Ensembles Stacking: h2o.ensemble Categorical Ensembles Summary 11. Epilogue: Didnt They All Do Well! Building Energy Results MNIST Results Football Data How Low Can You Go? The More the Merrier Still Desperate for More Filtering for Hardness Auto-Encoder Convolute and Shrink Ensembles That Was as Low as I Go Summary Index