Building an Effective Data Science Practice: A Framework to Bootstrap and Manage a Successful Data Science Practice

Description

Gain a deep understanding of data science and the thought process needed to solve problems in that field using the required techniques, technologies and skills that go into forming an interdisciplinary team. This book will enable you to set up an effective team of engineers, data scientists, analysts, and other stakeholders that can collaborate effectively on crucial aspects such as problem formulation, execution of experiments, and model performance evaluation. You’ll start by delving into the fundamentals of data science – classes of data science problems, data science techniques and their applications – and gradually build up to building a professional reference operating model for a data science function in an organization. This operating model covers the roles and skills required in a team, the techniques and technologies they use, and the best practices typically followed in executing data science projects. Building an Effective Data Science Practice provides a common base of reference knowledge and solutions, and addresses the kinds of challenges that arise to ensure your data science team is both productive and aligned with the business goals from the very start. Reinforced with real examples, this book allows you to confidently determine the strategic answers to effectively align your business goals with the operations of the data science practice. What You’ll Learn  Transform business objectives into concrete problems that can be solved using data science Evaluate how problems and the specifics of a business drive the techniques and model evaluation guidelines used in a project Build and operate an effective interdisciplinary data science team within an organization Evaluating the progress of the team towards the business RoI Understand the important regulatory aspects that are applicable to a data science practice  Who This Book Is For Technology leaders, data scientists, and project managers  Table of Contents About the Authors About the Technical Reviewer Acknowledgments Introduction Part I: Fundamentals Chapter 1: Introduction: The Data Science Process What We Mean by Data Science The Data Science Process Machine Learning Data Capture (from the World) Data Preparation Data Visualization Inference Data Engineering Terminology Chaos: AI, ML, Data Science, Deep Learning, Etc. Conclusion Further Reading References Chapter 2: Data Science and Your Business How Data Science Fits into a Business Operational Optimizations Product Enhancements Strategic Insights Is Your Business Ready for Data Science? A Cautionary Tale In the Beginning Was the Data And the Data Was with… Whom Exactly? The Model Said “Here Am I, Send Me” Conclusion Further Reading References Chapter 3: Monks vs. Cowboys: Data Science Cultures The Two Cultures of Data Science Hybrid Cultures Cultural Differences Data Science Culture and Your Business The Cultural Spectrum of Data Scientists Theory and Experimentation in Data Science Data Engineering Conclusion Summary of Part 1 Part II: Classes of Problems Chapter 4: Classification Data Capture Data Preparation Data Visualization Machine Learning Inference Data Engineering Conclusion Chapter 5: Regression Data Capture Data Preparation Data Visualization Machine Learning Inference Conclusion Chapter 6: Natural Language Processing Data Capture Data Preparation Machine Learning Inference Conclusion Chapter 7: Clustering Data Capture Data Preparation Handling Missing Values Normalization Data Visualization Machine Learning Similarity of Observations Data Visualization Iteration Inference Interpreting the Dendrogram Actionable Insights for Marketing Conclusion Further Reading Reference Chapter 8: Anomaly Detection Anomaly Detection Using Unlabeled Data Novelty Detection Using Pure Data Data Science Process for Anomaly Detection The World and Data Capture Data Preparation Data Visualization Box Plots Conditional Box Plots Scatter Plots Machine Learning Inference Anatomy of an Anomaly Complex Anomalies Collective Anomalies Contextual Anomalies Time Series Conclusion Further Reading References Chapter 9: Recommendations Data Capture Items and Interactions Quantifying an Interaction Example Data Data Preparation Normalization Handling Missing Values Data Visualization Machine Learning Clustering-Based Approach Inference End-to-End Automation Conclusion Further Reading References Chapter 10: Computer Vision Processing Images Image Classification/Regression Object Detection Datasets, Competitions, and Architectures Processing Videos Video Classification Object Tracking Data Science Process for Computer Vision The World and Data Capture Data Preparation Data Visualization Machine Learning Model Performance Evaluation Inference Data Engineering Conclusion Further Reading References Chapter 11: Sequential Decision-Making The RL Setting Basic Knowledge and Rules Training Nestor Episode Training Phases Past Cases Ongoing New Cases, with Imitation Supervised Exploration Supervised Exploitation Data in the RL Setting Data of Experts’ Decisions Simulated Data Challenges in RL Availability of Data Information in Observations Exploration vs. Exploitation Data Science Process for RL Conclusion Further Reading References Part III: Techniques and Technologies Chapter 12: Techniques and Technologies: An Overview Chapter 13: Data Capture Data Sources (1) Ingestion (2) Data Storage Data Lake (3) Data Warehouse (4) Shared File Systems (5) Read Data (6) Programmatic Access SQL Query Engine Open Source vs. Paid Data Engineering Conclusion Chapter 14: Data Preparation Handling Missing Values Feature Scaling Text Preprocessing Stemming TF-IDF Converting Categorical Variables into Numeric Variables Transforming Images Libraries and Tools Libraries Tools Data Engineering Conclusion Chapter 15: Data Visualization Graphs/Charts/Plots Legends Layouts Options Interactive Visualizations Deriving Insights from Visualizations Histogram Kernel Density Estimate Plot Libraries and Tools Libraries Tools Data Engineering Conclusion Chapter 16: Machine Learning Categories of Machine Learning Algorithms Supervised Learning Unsupervised Learning Reinforcement Learning Popular Machine Learning Algorithms Linear Regression Logistic Regression Support Vector Machine Decision Tree Random Forest Gradient Boosted Trees Artificial Neural Network Convolutional Neural Network Evaluating and Tuning Models Evaluating Models Tuning models Cross-Validation Libraries and Tools Data Engineering Conclusion Further Reading References Chapter 17: Inference Model Release Process (1) Model Registry Model Converter Interexchange Format Target System Model Packaging Production Inference Server (2) Inference/Prediction Service Model Monitoring Mobile and Web Applications (3) ML Ops Open Source vs. Paid Data Engineering Conclusion Chapter 18: Other Tools and Services Development Environment Experiment Registry Compute Infrastructure AutoML Purpose of AutoML AutoML Cautions Tools and Services Multimodal Predictive Analytics and Machine Learning Data Science Apps/Workflows Off-the-Shelf AI Services and Libraries When to Use Open Source vs. Paid Conclusion Chapter 19: Reference Architecture Experimentation Dev Environment (1) Data Sources (2) Ingestion (3) Core Infra (4) Analytics (5) Data Science Apps/Workflows (6) AutoML (7) From Experimentation to Production AI Services Conclusion Chapter 20: Monks vs. Cowboys: Praxis Goals of Modeling Estimating Truth: Simplicity of Representation Estimating Truth: Attribution Prediction: Interpretability Prediction: Accuracy Grading ML Techniques Cultural Differences Conclusion Summary of Part 3 References Part IV: Building Teams and Executing Projects Chapter 21: The Skills Framework The Three Dimensions of Skills Data Analysis Skills Software Engineering Skills Domain Expertise The Roles in a Data Science Team Citizen Data Scientist Data Analyst Data Science Technician ML Ops Data Engineer Data Architect ML Engineer Data Scientist Chief Data Scientist Deviations in Skills Conclusion Chapter 22: Building and Structuring the Team Typical Team Structures Small Incubation Team Mature Operational team Team Evolution The Key Hire: Chief Data Scientist Evaluating the Culture Hiring vs. Getting a Consultant Data Engineering: Requirements and Staffing Notes on Upskilling Conclusion Chapter 23: Data Science Projects Types of Data Science Projects Knowledge Discovery from Data/Data Mining Data Science Infusion in Processes Data Science Infusion in Products Data Science-Based Product Typical Traits of Data Science Projects KPIs Model Performance Experimentation Cycle Time Effort-Cost Trade-Offs Data Quality Importance of Data Quality Issues Arising from Poor-Quality Data Severity of Impact Dimensions of Data Quality Measuring Data Quality Ensuring Data Quality Resistance to Data Quality Efforts Data Protection and Privacy Encryption Access Controls Identifiable/Protected/Sensitive Information Federated Learning Legal and Regulatory Aspects When Are These Relevant? Nondiscrimination Explainability and Accountability Explainable AI: What Is an “Explanation”? Cognitive Bias Cognitive Bias and Data Science Projects Conclusion and Further Reading References Index