دانلود کتاب علوم داده و یادگیری ماشین- ابزارها و تکنیک‌های ضروری

Data science is inherently an interdisciplinary activity that has evolved from a synergy between computer science and statistics. To do data science, we need data! So we have a data set; it could be a structured database of employee records with all their details, an unstructured collection of textual documents (say emails), a large collection of images of animals, a time series of financial data from the stock market, epidemiological data giving the number of infected individuals per day for a given region over a period of time, or geographical data pertaining to businesses in central London. Having data is not enough; we need to have a problem we would like to solve or some questions we wish to answer using the data. For example, in an employee data set, we may wish to know the employee characteristics that determine their salary band or, in an epidemiological data set, we may wish to determine how fast a virus is spreading in the population. Now, in a broad sense, once the data is available and we have a well-defined problem to work on, several steps determine the tasks a data scientist should perform to tackle the problem at hand and find out what the data is telling us. It is always sensible to start with an exploratory data analysis phase, which is carried out with the aid of visualisation tools. Exploring data will help us form some hypotheses about the data, which in turn allows us to build a statistical model of the data. Inevitably, we will use an algorithmic method, based on our model, whose output will assist us in verifying or refuting the hypotheses we have formed. In a nutshell, this is what data science is about. The algorithmic method essentially enables the discovery of patterns in the data, which may be large in size and/or complex, according to the statistical model we have formed. This is often referred to as machine learning; however, the general process of pattern or knowledge discovery is known as data mining. In our exposition of data science, we prefer to use the term machine learning as the subfield of computer science responsible for the algorithmic part of data science. Therefore, in a very broad sense, data science comprises the methods and algorithms used to analyse the data and present the findings from the ensuing analysis. Taking this a step further, who are the stakeholders in this discipline and activity called data science? Computer scientists, such as the authors, are responsible for designing and implementing the algorithms in such a way that they scale to very large and potentially complex data sets. Then statisticians are responsible for model building, which is an essential part of data science. However, one could argue that the data scientist combines skills from these two disciplines of computer science and statistics, leaning toward one side or another depending on their background. Still, we have a third group of stakeholders who bring the data and the problems to the table: they may be social scientists, economists, epidemiologists, or any other professionals from any other discipline that would like to use data science to aid them in answering questions they have about the data they possess. For successful data science to take place, more often than not, an interdisciplinary team needs to be working on the problem at hand. There is also a breed of data scientists who, from the start, build their expertise in this field rather than in the field of computer science or statistics. Moreover, others, such as the authors, started off in computer science or statistics and have moved their expertise to the middle ground of data science. Ultimately, the question of what exactly is the relationship between data science, statistics, and computer science/machine learning will remain an ongoing debate. It is important from our perspective to appreciate that data science demands the application of expertise from both these disciplines to solve real-world problems emanating from data. Furthermore, our goal in this book is to provide a relatively brief technical introduction to this exciting field that can be understood by practitioners and researchers alike, coming from diverse backgrounds. In Chapter 2 we introduce the basic statistical notions needed to become a data scientist. In Chapter 3 we introduce the fundamental data types that data scientists need to understand when going about their daily job. Chapter 4 is a machine learning crash course for budding data scientists. In Chapter 5 we examine several topics of the authors’ choice in data science that will enhance data scientists’ knowledge and give them insight into typical applications they may come across during their work. Finally, in Chapter 6 we summarise the material we have covered in this introduction.