We live in a world in which massive amounts of data are produced and stored every day. This data comes from a plethora of information systems, devices, and sensors. Almost everything you do, and items you use to do it, produces data which can be, or is, captured.

This has been greatly enabled by the ubiquitous nature of services that are connected to networks, and by the great increases in data storage facilities; this, combined with the ever-decreasing cost of storage, has made capturing and storing even the most trivial of data effective.

This has led to massive amounts of data being piled up and ready for access. But this data is spread out all over cyber-space, and is cannot actually be referred to as information. It tends to be a collected collection of the recording of events, whether financial, of your interactions with social networks, or of your personal health monitor tracking your heartbeat throughout the day. This data is stored in all kinds of formats, is located in scattered places, and beyond its raw nature does give much insight.

Logically, the overall process can be broken into three major areas of discipline:

• Data manipulation
• Data analysis
• Data science

These three disciplines can and do have a lot of overlap. Where each ends and the others begin is open to interpretation. For the purposes of this book we will define each as in the following sections.

Data is distributed all over the planet. It is stored in different formats. It has widely varied levels of quality. Because of this there is a need for tools and processes for pulling data together and into a form that can be used for decision making. This requires many different tasks and capabilities from a tool that manipulates data in preparation for analysis. The features needed from such a tool include:

• Programmability for reuse and sharing
• Access to data from external sources
• Storing data locally
• Indexing data for efficient retrieval
• Alignment of data in different sets based upon attributes
• Combining data in different sets
• Transformation of data into other representations
• Cleaning data from cruft
• Effective handling of bad data
• Grouping data into common baskets
• Aggregation of data of like characteristics
• Application of functions to calculate meaning or perform transformations
• Query and slicing to explore pieces of the whole
• Restructuring into other forms
• Modeling distinct categories of data such as categorical, continuous, discrete, and time series
• Resampling data to different frequencies

There are many data manipulation tools in existence. Each differs in support for the items on this list, how they are deployed, and how they are utilized by their users. These tools include relational databases (SQL Server, Oracle), spreadsheets (Excel), event processing systems (such as Spark), and more generic tools such as R and pandas.

Data analysis is the process of creating meaning from data. Data with quantified meaning is often called information. Data analysis is the process of creating information from data through the creation of data models and mathematics to find patterns. It often overlaps data manipulation and the distinction between the two is not always clear. Many data manipulation tools also contain analyses functions, and data analysis tools often provide data manipulation capabilities.

Data science is the process of using statistics and data analysis processes to create an understanding of phenomena within data. Data science usually starts with information and applies a more complex domain-based analysis to the information. These domains span many fields such as mathematics, statistics, information science, computer science, machine learning, classification, cluster analysis, data mining, databases, and visualization. Data science is multidisciplinary. Its methods of domain analysis are often very different and specific to a specific domain.

pandas first and foremost excels in data manipulation. All of the needs itemized earlier will be covered in this book using pandas. This is the core of pandas and is most of what we will focus on in this book.

It is worth noting that that pandas has a specific design goal: emphasizing data

But pandas does provide several features for performing data analysis. These capabilities typically revolve around descriptive statistics and functions required for finance such as correlations.

Therefore, pandas itself is not a data science toolkit. It is more of a manipulation tool with some analysis capabilities. pandas explicitly leaves complex statistical, financial, and other types of analyses to other Python libraries, such as SciPy, NumPy, scikit-learn, and leans upon graphics libraries such as matplotlib and ggvis for data visualization.

This focus is actually a strength of pandas over other languages such as R as pandas applications are able to leverage an extensive network of robust Python frameworks already built and tested elsewhere by the Python community.