Mastering Predictive Analytics with Python

Description

Key FeaturesMaster open source Python tools to build sophisticated predictive modelsLearn to identify the right machine learning algorithm for your problem with this forward-thinking guideGrasp the major methods of predictive modeling and move beyond the basics to a deeper level of understandingBook Description The volume, diversity, and speed of data available has never been greater. Powerful machine learning methods can unlock the value in this information by finding complex relationships and unanticipated trends. Using the Python programming language, analysts can use these sophisticated methods to build scalable analytic applications to deliver insights that are of tremendous value to their organizations. In Mastering Predictive Analytics with Python, you will learn the process of turning raw data into powerful insights. Through case studies and code examples using popular open-source Python libraries, this book illustrates the complete development process for analytic applications and how to quickly apply these methods to your own data to create robust and scalable prediction services. Covering a wide range of algorithms for classification, regression, clustering, as well as cutting-edge techniques such as deep learning, this book illustrates not only how these methods work, but how to implement them in practice. You will learn to choose the right approach for your problem and how to develop engaging visualizations to bring the insights of predictive modeling to life What you will learnGain an insight into components and design decisions for an analytical applicationMaster the use Python notebooks for exploratory data analysis and rapid prototypingGet to grips with applying regression, classification, clustering, and deep learning algorithmsDiscover the advanced methods to analyze structured and unstructured dataFind out how to deploy a machine learning model in a production environmentVisualize the performance of models and the insights they produceScale your solutions as your data grows using PythonEnsure the robustness of your analytic applications by mastering the best practices of predictive analysisAbout the Author Joseph Babcock has spent almost a decade exploring complex datasets and combining predictive modeling with visualization to understand correlations and forecast anticipated outcomes. He received a PhD from the Solomon H. Snyder Department of Neuroscience at The Johns Hopkins University School of Medicine, where he used machine learning to predict adverse cardiac side effects of drugs. Outside the academy, he has tackled big data challenges in the healthcare and entertainment industries. Table of ContentsFrom Data to Decisions – Getting Started with Analytic ApplicationsExploratory Data Analysis and Visualization in PythonFinding Patterns in the Noise – Clustering and Unsupervised LearningConnecting the Dots with Models – Regression MethodsPutting Data in its Place – Classification Methods and AnalysisWords and Pixels – Working with Unstructured DataLearning from the Bottom Up – Deep Networks and Unsupervised FeaturesSharing Models with Prediction ServicesReporting and Testing – Iterating on Analytic Systems Cover......Page 1 Copyright......Page 3 Credits......Page 4 About the Author......Page 5 About the Reviewer......Page 6 www.PacktPub.com......Page 7 Table of Contents......Page 8 Preface......Page 14 Chapter 1: From Data to Decisions – Getting Started with Analytic Applications......Page 20 Designing an advanced analytic solution......Page 23 Data layer: warehouses, lakes, and streams......Page 25 Modeling layer......Page 27 Deployment layer......Page 33 Reporting layer......Page 34 Case study: sentiment analysis of social media feeds......Page 35 Data input and transformation......Page 36 Model development......Page 37 Case study: targeted e-mail campaigns......Page 38 Data input and transformation......Page 39 Visualization and reporting......Page 40 Summary......Page 42 Chapter 2: Exploratory Data Analysis and Visualization in Python......Page 44 Exploring categorical and numerical data in IPython......Page 45 The notebook interface......Page 46 Loading and inspecting data......Page 49 Basic manipulations – grouping, filtering, mapping, and pivoting......Page 52 Charting with Matplotlib......Page 57 Cleaning and converting......Page 65 Time series diagnostics......Page 67 Joining signals and correlation......Page 69 Loading geospatial data......Page 72 Working in the cloud......Page 74 Creating the SparkContext......Page 75 Creating an RDD......Page 77 Creating a Spark DataFrame......Page 78 Summary......Page 80 Chapter 3: Finding Patterns in the Noise – Clustering and Unsupervised Learning......Page 82 Numerical distance metrics......Page 83 Correlation similarity metrics and time series......Page 89 Similarity metrics for categorical data......Page 97 K-means clustering......Page 102 Affinity propagation – automatically choosing cluster numbers......Page 108 k-medoids......Page 112 Agglomerative clustering......Page 113 Where agglomerative clustering fails......Page 115 Streaming clustering in Spark......Page 119 Summary......Page 123 Chapter 4: Connecting the Dots with Models – Regression Methods......Page 124 Linear regression......Page 125 Data preparation......Page 128 Model fitting and evaluation......Page 133 Statistical significance of regression outputs......Page 138 Generalize estimating equations......Page 143 Mixed effects models......Page 145 Time series data......Page 146 Generalized linear models......Page 147 Applying regularization to linear models......Page 148 Decision trees......Page 151 Random forest......Page 157 Scaling out with PySpark – predicting year of song release......Page 160 Summary......Page 162 Chapter 5: Putting Data in its Place – Classification Methods and Analysis......Page 164 Logistic regression......Page 165 Multiclass logistic classifiers: multinomial regression......Page 169 Formatting a dataset for classification problems......Page 170 Learning pointwise updates with stochastic gradient descent......Page 174 Jointly optimizing all parameters with second-order methods......Page 177 Fitting the model......Page 181 Evaluating classification models......Page 184 Strategies for improving classification models......Page 188 Separating Nonlinear boundaries with Support vector machines......Page 191 Fitting and SVM to the census data......Page 193 Gradient boosted decision trees......Page 196 Comparing classification methods......Page 199 Case study: fitting classifier models in pyspark......Page 201 Summary......Page 203 Chapter 6: Words and Pixels – Working with Unstructured Data......Page 204 Cleaning textual data......Page 205 Extracting features from textual data......Page 208 Using dimensionality reduction to simplify datasets......Page 211 Principal component analysis......Page 212 Latent Dirichlet Allocation......Page 224 Images......Page 228 Cleaning image data......Page 229 Thresholding images to highlight objects......Page 232 Dimensionality reduction for image analysis......Page 235 Case Study: Training a Recommender System in PySpark......Page 239 Summary......Page 241 Chapter 7: Learning from the Bottom Up – Deep Networks and Unsupervised Features......Page 242 A network of one – the perceptron......Page 243 Combining perceptrons – a single-layer neural network......Page 245 Parameter fitting with back-propagation......Page 248 Discriminative versus generative models......Page 253 Vanishing gradients and explaining away......Page 254 Pretraining belief networks......Page 257 Using dropout to regularize networks......Page 260 Convolutional networks and rectified units......Page 261 Compressing Data with autoencoder networks......Page 265 Optimizing the learning rate......Page 266 The TensorFlow library and digit recognition......Page 268 The MNIST data......Page 269 Constructing the network......Page 271 Summary......Page 275 Chapter 8: Sharing Models with Prediction Services......Page 276 The architecture of a prediction service......Page 277 The GET requests......Page 279 The PUT request......Page 281 Server – the web traffic controller......Page 282 Application – the engine of the predictive services......Page 284 Persisting information with database systems......Page 285 Case study – logistic regression service......Page 286 Setting up the database......Page 287 The web server......Page 290 The web application......Page 292 The flow of a prediction service – training a model......Page 293 On-demand and bulk prediction......Page 302 Summary......Page 306 Chapter 9: Reporting and Testing – Iterating on Analytic Systems......Page 308 Evaluating changes in model performance......Page 309 Changes in feature importance......Page 313 Changes in unsupervised model performance......Page 314 Iterating on models through A/B testing......Page 316 Experimental allocation – assigning customers to experiments......Page 317 Deciding a sample size......Page 318 Guidelines for communication......Page 321 Visualizing results......Page 322 The report server......Page 323 The report application......Page 324 The visualization layer......Page 325 Summary......Page 329 Index......Page 330