The Shark Research Institute (SRI) maintains a log of shark attacks worldwide, known as the Global Shark Attack File (GSAF). The raw data can be found on the SRI website here: https://www.sharks.org/global-shark-attack-file.

Each row of the dataset includes some of the following items:

Date of the incident

Location, including country, state, and/or location

Type of incident (unprovoked, provoked, sea disaster, watercraft, or questionable)

Activity of the human(s) involved

Type of shark involved, if known

Outcome, including fatal, non-fatal, or unknown

The goal of this data analysis project was to review historical shark attack records to look at trends in the number of attacks, whether those attacks were fatal, and which countries had the most number of shark attacks and human fatalities.

The GSAF data required a lot of cleaning since there does not appear to be a set standard on how to update the file. Some information contained misspellings and inconsistent names (countries/states, human activity, etc.), or data was entered in the wrong column.

To reduce the amount of data to be cleaned, data before 1950 was removed. The remaining data was reviewed for completeness and changes were made based on the available information. For example, a record with an "unknown" in the fatality column was changed to "Non-Fatal" because other information for that incident stated the victim survived. The data cleaning focused on the Country (multiple misspellings and/or extra spaces), incident type, and incident outcome.

The first look at the data was to determine how the different types of incidents have changed, if at all, since 1950. The types of attacks in the GSAF are below:

Unprovoked - a shark initiates an aggressive interaction with a human without any apparent provocation or stimulus from the human.

Provoked - a shark reacts aggressively towards a human due to a deliberate action or provocation initiated by the human.

Watercraft - bites or rams some type of boat or other watercraft.

Sea Disaster - includes any air or sea disaster, such as a plane crash landing in the water or a boat capsizing or sinking.

Questionable - there was insufficient information to determine if the incident was caused by a shark. This can include incidents where a person drowns and is scavenged by sharks.

Globally, the number of unprovoked shark attacks has been increasing since the 1980s. However, there does not appear to be an increase in the number of provoked attacks.

With the number of attacks has been increasing over time, is there an increase in the number of fatalities?

The data for shark incident outcomes show that the annual number of fatalities is not increasing. While the number of shark attacks has been increasing, the annual number of fatalities has remained relatively consistent (~10 per year).

The number of shark attacks has been increasing since 1950, but fatalities are holding relatively steady. How many shark attacks do individual countries experience and how often are the attacks fatal?

A table was created from the cleaned GSAF data with each country listed and the number of incidents was counted, along with the number of fatalities and non-fatalities. From this table, the results were shown for those countries with at least 10 fatal shark attacks since 1950.

The United States has the greatest number of attacks by far, with Australia and South Africa a distant second and third, respectively. However, the US had the lowest fatality percentage of these countries. A person is more likely to be attacked by a shark in the US but will most likely survive.

Globally, the number of shark attacks on humans has been increasing since 1950. Unprovoked attacks have increased while other types of attacks have remained steady. The good news is that the increase in shark attacks has not resulted in increased human fatalities. The data shows that the United States leads all other countries in the number of shark attacks while Australia leads in the number of fatalities.

There is a lot of information available in the GSAF dataset. One of my future personal projects will be to analyze the activity of the human during the attack. The other analysis I'm interested in is the shark species involved in these incidents. Both of those will require additional cleaning due to the non-standard responses in the data.