Название: Modern Business Analytics: Practical Data Science for Decision-making
Автор: Matt Taddy, Leslie Hendrix, Matthew C. Harding
Издательство: McGraw Hill
Год: 2023
Страниц: 465
Язык: английский
Формат: pdf (true)
Размер: 34.6 MB

Written by Matt Taddy, successful author of the McGraw Hill Professional title, Business Data Science graduate of University of Chicago and Amazon Chief Economist. This new higher-ed text takes a practical, modern approach to Data Science and business analytics for the graduate-level business analytics student or professional. It takes a learn-by-doing approach, with real data analysis examples that explain the "why", rather than the "what" in the decision-making discussions. It uses R as the primary technology throughout the text and includes an end-of-chapter reference to the basic R recipes in each chapter. The text uses tools from economics and statistics in combination with Machine Learning Techniques to create a platform for using data to make decisions.

The practice of data analytics is changing and modernizing. Innovations in computation and Machine Learning are creating new opportunities for the data analyst: exposing previously unexplored data to scientific analysis, scaling tasks through automation, and allowing deeper and more accurate modeling. Spreadsheet models and pivot tables are being replaced by code scripts in languages like R and Python. There has been massive growth in digitized information, accompanied by development of systems for storage and analysis of this data. There has also been an intellectual convergence across fields - Machine Learning and Computer Science, statistics, and social sciences and economics - that has raised the breadth and quality of applied analysis everywhere. This is the Data Science approach to analytics, and it allows leaders to go deeper than ever to understand their operations, products, and customers.

The Connect product that supports the text includes Interactive Activities that have students explore content more deeply, Excel activities like Integrated Excel & Applying Excel, and a Prep Course that helps students refresh on fundamental pre-requisite knowledge they need to know prior to this course.

Deep Learning makes use of deep neural networks (DNNs) which can process complicated inputs, like text and images and video, in a largely automatic fashion. You don’t need to spend huge effort in feature construction and pre-processing with deep learning: you just set up a model architecture and get training. But the full framework of deep learning is much more than just DNN models. It is the combination of DNNs with efficient programming frameworks and scalable parallel “stochastic gradient descent” optimization algorithms.

The next key ingredient is software: deep learning frameworks. As you’ve learned from this book, the R software (or an equivalent, like Python) is crucial for good analysis because it allows you to abstract away from a lot of the math and build models by defining functions and supplying data to those functions. Deep Learning frameworks go further by making it easy to define DNN models as a combination of different types of layers, and then automatically construct the right optimization algorithm for training whatever model you’ve defined. Frameworks like Keras and TensorFlow, which are what we will work with in this chapter, take a defined DNN, do the math (i.e., automatically calculate derivatives) as necessary to define the optimization routine, and efficiently implement that optimization on whatever hardware is available (whether on your laptop CPUs or on large distributed farms of specialized processing units). These frameworks abstract away the details so that the coder, whether business analyst or software engineer, can concentrate on the big picture of what inputs to use and what predictions they care about.