Practitioner’s Guide to Data Science

This book aims to increase the visibility of data science in real-world, which differs from what you learn from a typical textbook. Many aspects of day-to-day data science work are almost absent from conventional statistics, machine learning, and data science curriculum. Yet these activities account for a considerable share of the time and effort for data professionals in the industry. Based on industry experience, this book outlines real-world scenarios and discusses pitfalls that data science practitioners should avoid. It also covers the big data cloud platform and the art of data science, such as soft skills. The authors use R as the primary tool and provide code for both R and Python. This book is for readers who want to explore possible career paths and eventually become data scientists. This book comprehensively introduces various data science fields, soft and programming skills in data science projects, and potential career paths. Traditional data-related practitioners such as statisticians, business analysts, and data analysts will find this book helpful in expanding their skills for future data science careers. Undergraduate and graduate students from analytics-related areas will find this book beneficial to learn real-world data science applications. Non-mathematical readers will appreciate the reproducibility of the companion R and python codes. Key Features: It covers both technical and soft skills. It has a chapter dedicated to the big data cloud environment. For industry applications, the practice of data science is often in such an environment. It is hands-on. We provide the data and repeatable R and Python code in notebooks. Readers can repeat the analysis in the book using the data and code provided. We also suggest that readers modify the notebook to perform analyses with their data and problems, if possible. The best way to learn data science is to do it! List of Figures Preface About the Authors Acknowledgment 1 Introduction 1.1 A Brief History of Data Science 1.2 Data Science Role and Skill Tracks 1.2.1 Engineering 1.2.2 Analysis 1.2.3 Modeling/Inference 1.3 What Kind of Questions Can Data Science Solve? 1.3.1 Prerequisites 1.3.2 Problem Type 1.4 Structure of Data Science Team 1.5 Data Science Roles 2 Soft Skills for Data Scientists 2.1 Comparison between Statistician and Data Scientist 2.2 Beyond Data and Analytics 2.3 Three Pillars of Knowledge 2.4 Data Science Project Cycle 2.4.1 Types of Data Science Projects 2.4.2 Problem Formulation and Project Planning Stage 2.4.3 Project Modeling Stage 2.4.4 Model Implementation and Post Production Stage 2.4.5 Project Cycle Summary 2.5 Common Mistakes in Data Science 2.5.1 Problem Formulation Stage 2.5.2 Project Planning Stage 2.5.3 Project Modeling Stage 2.5.4 Model Implementation and Post Production Stage 2.5.5 Summary of Common Mistakes 3 Introduction to the Data 3.1 Customer Data for a Clothing Company 3.2 Swine Disease Breakout Data 3.3 MNIST Dataset 3.4 IMDB Dataset 4 Big Data Cloud Platform 4.1 Power of Cluster of Computers 4.2 Evolution of Cluster Computing 4.2.1 Hadoop 4.2.2 Spark 4.3 Introduction of Cloud Environment 4.3.1 Open Account and Create a Cluster 4.3.2 R Notebook 4.3.3 Markdown cells 4.4 Leverage Spark Using R Notebook 4.5 Databases and SQL 4.5.1 History 4.5.2 Database, Table, and View 4.5.3 Basic SQL Statement 4.5.4 Advanced Topics in Database 5 Data Pre-processing 5.1 Data Cleaning 5.2 Missing Values 5.2.1 Impute Missing Values with Median/Mode 5.2.2 K-nearest Neighbors 5.2.3 Bagging Tree 5.3 Centering and Scaling 5.4 Resolve Skewness 5.5 Resolve Outliers 5.6 Collinearity 5.7 Sparse Variables 5.8 Re-encode Dummy Variables 6 Data Wrangling 6.1 Summarize Data 6.1.1 dplyr Package 6.1.2 apply(), lapply() and sapply() in base R 6.2 Tidy and Reshape Data 7 Model Tuning Strategy 7.1 Variance-Bias Trade-Off 7.2 Data Splitting and Resampling 7.2.1 Data Splitting 7.2.2 Resampling 8 Measuring Performance 8.1 Regression Model Performance 8.2 Classification Model Performance 8.2.1 Confusion Matrix 8.2.2 Kappa Statistic 8.2.3 ROC 8.2.4 Gain and Lift Charts 9 Regression Models 9.1 Ordinary Least Square 9.1.1 The Magic P-value 9.1.2 Diagnostics for Linear Regression 9.2 Principal Component Regression and Partial Least Square 10 Regularization Methods 10.1 Ridge Regression 10.2 LASSO 10.3 Elastic Net 10.4 Penalized Generalized Linear Model 10.4.1 Introduction to glmnet Package 10.4.2 Penalized Logistic Regression 11 Tree-Based Methods 11.1 Tree Basics 11.2 Splitting Criteria 11.2.1 Gini Impurity 11.2.2 Information Gain (IG) 11.2.3 Information Gain Ratio (IGR) 11.2.4 Sum of Squared Error (SSE) 11.3 Tree Pruning 11.4 Regression and Decision Tree Basic 11.4.1 Regression Tree 11.4.2 Decision Tree 11.5 Bagging Tree 11.6 Random Forest 11.7 Gradient Boosted Machine 11.7.1 Adaptive Boosting 11.7.2 Stochastic Gradient Boosting 12 Deep Learning 12.1 Feedforward Neural Network 12.1.1 Logistic Regression as Neural Network 12.1.2 Stochastic Gradient Descent 12.1.3 Deep Neural Network 12.1.4 Activation Function 12.1.5 Optimization 12.1.6 Deal with Overfitting 12.1.7 Image Recognition Using FFNN 12.2 Convolutional Neural Network 12.2.1 Convolution Layer 12.2.2 Padding layer 12.2.3 Pooling Layer 12.2.4 Convolution Over Volume 12.2.5 Image Recognition Using CNN 12.3 Recurrent Neural Network 12.3.1 RNN Model 12.3.2 Long Short Term Memory 12.3.3 Word Embedding 12.3.4 Sentiment Analysis Using RNN A Handling Large Local Data A.1 readr A.2 data.table— Enhanced data.frame B R Code for Data Simulation B.1 Customer Data for Clothing Company B.2 Swine Disease Breakout Data Bibliography Index