Introduction

Data analysis is becoming one of the most in-demand skills of the 21st century. The exponential growth of data, the increase of computation power, and the reduced costs for cloud and high-performance computing allow both companies and individuals to analyze large amounts of data that were intractable 20 years ago. This type of analysis unlocks new business opportunities for companies that have decided to adopt data analytics in their business.

Integrating data analytics into the core business of a company is not an easy task. A well-established team of software engineers, data engineers, and data scientists is required, in which the team members not only have a broad experience of algorithms, software architecture, and machine learning, but also a good understanding of the business of the company. While the first three skills are easily transferable from one type of business to another, understanding the business itself takes time.

In this book, we provide an introduction on how to apply analytical skills to various business problems, aiming, in this way, to reduce the gap between theoretical knowledge and practical experience.

We have decided to adopt Python as the only language for the analyses in this book. There are several reasons for this: first, Python is a general-purpose language, with a large community constantly working to improve its features and functionalities. Second, it is also one of the easiest languages to start with, even with no prior experience of programming. Third, there is a broad variety of excellent scientific computing and machine learning libraries written in Python, and, finally, all of these libraries are open source.

Since an introduction to Python is beyond the scope of this book, we assume that the reader has some prior knowledge of it. Furthermore, we assume some basic knowledge of the standard Python packages for data analysis (such as pandas, numpy, and scipy). Although we do not provide a rigorous introduction to these libraries, explanations of the used functions and modules are given where necessary.

We start our first chapter with a relatively easy problem, involving bike sharing data.

Bike sharing is a fundamental service, commonly used in the urban mobility sector. It is easily accessible (as no driving license is required to ride a bike), is cheaper than normal car sharing services (since bike maintenance and insurance are substantially cheaper than automobile ones), and, finally, is often a fast way to commute within the city. Therefore, understanding the driving factors of bike sharing requests is essential for both companies and users.

From a company's perspective, identifying the expected bike demand in a specific area, within a specific time frame, can significantly increase revenue and customer satisfaction. Moreover, bike relocation can be optimized to further reduce operational costs. From a user's perspective, probably the most important factor is bike availability in the shortest wait time, which we can easily see aligning with the company's interests.

In this chapter, we will analyze bike sharing data from Capital Bikeshare in Washington, D.C., USA, for the period between January 1, 2011, and December 31, 2012. The data is aggregated on an hourly basis. This means that no initial and final locations of the individual rides are available, but only the total number of rides per hour. Nevertheless, additional meteorological information is available in the data, which could serve as a driving factor for identifying the total number of requests for a specific time frame (bad weather conditions could have a substantial impact on bike sharing demand).

Note

The original dataset is available at https://archive.ics.uci.edu/ml/datasets/Bike+Sharing+Dataset#.

For further information on this topic, check out the following journal article: Fanaee-T, Hadi, and Gama, Joao, 'Event labeling combining ensemble detectors and background knowledge', Progress in Artificial Intelligence (2013): pp. 1-15, Springer Berlin Heidelberg.

Note that although the conducted analysis is related to bike sharing, the provided techniques could be easily transferred to other types of sharing business models, such as car or scooter sharing.