Global Shark Attacks Data Analysis

Global Shark Attacks Data Analysis

An analysis of shark attacks reported worldwide from 1854-2015

Introduction:

A shark attack is an attack on a human by a shark. Every year around 80 unprovoked attacks are reported worldwide. Despite their relative rarity, the effect the media has on the population's view of shark attacks has generally been negative. Starting with the effects generated from news broadcasts, a shark attack is quickly broadcasted across the country, particularly if fatal, even though more people die from random occurrences such as lightning strikes than from a shark attack

The International Shark Attack File, internationally recognized as the definitive source of scientifically accurate information on shark attacks, is a compilation of investigations of all known shark attacks.  Established in 1958, it is administered by the Florida Museum of Natural History at the University of Florida under the auspices of the American Elasmobranch Society, the world's foremost international organization of scientists studying sharks, skates, and rays. The database contains information on more than 6,000 individual investigations from the mid-1500's to present.  Many of the data in the ISAF originate from the voluntary submissions of numerous cooperating scientists who serve worldwide as regional observers.  Data submitted to the ISAF is screened, coded, and computerized.

The International shark attack file reported 5750 incidents of alleged shark-human interaction occurring worldwide and dating back from 1854 to 2016, 2,785 were confirmed unprovoked shark attacks, of which 439 were fatal. The aim of this project is to investigate the frequency of these attacks and to produce a comprehensive exploratory data analysis of the global shark attack datasets with an emphasis on the incidents happening at the united states of America and in later years.

The analysis conducted in this project stresses out where shark attacks are most common, which country has the most shark attacks and what activities are associated with each of these incidents. In addition to the fatality rate of shark attacks and when exactly have they most occurred.  Our analysis also briefly highlights the type of injuries resulting of each attack, number of survivors over the years and the specific shark species involved in each incident.

The probability of a recorded shark attack can be modeled as the joint probability of multiple processes: the probability that a person and a shark encounter one another p(E), that such an encounter results in a bite p(B), and that the attack is communicated p(C). p(E) depends on the abundance of humans (H) and sharks (S) in the water, and on the spatial overlap between people and sharks (O). If we assume that for any given encounter p(B) and p(C) remain constant, then the probability of a recorded shark attack depends only on H, S, and O. We are analyzing the data using this probability framework.

The conclusions drawn from this study are the responses to the number of questions that can be asked about the global shark attacks data. Our goal is that the comprehensive analysis helps us introduce a perception of the risk of a shark attack.

Understanding the data:

The Global shark attacks data is in csv format and consists initially of 5750 observations and 16 attributes. Most of these attributes are categorical variables. Table 1 is describing each one of the attribute and its domain:

                        Table1: Attributes domain and specifications of the Global Shark Attacks Dataset