St Edward's Academic Review 2025

ACADEMIC REVIEW 2025

Literature Review

of extracting patterns could evolve, from finding missing values to making predictions.

Natural disasters have been responsible for the deaths of millions of people around the world. They are known to have caused a considerable amount of damage to numerous countries, permanently damaging property and the environment. It is estimated that in the last ten years, the cause of death for one in a hundred people around the world was natural disasters (Ritchie, Rosado, & Roser, 2022). According to France-Presse (2021) the ten most costly natural disasters in 2021, including Hurricane Ida and the winter storm in Texas, caused more than $170 billion in damage – this was an increase of about $20 billion from 2020. These ten natural disasters alone resulted in the deaths of at least 1,075 people and the displacement of about 1.3 million. Tremendous amounts of casualties and damage have left some countries devastated in the aftermath. With the aid of advanced technology and the emerging field of data science, the impacts of these natural disasters may be able to be minimised to potentially save lives and reduce the financial burden on disaster-prone countries. This essay explores the role played by predicting natural disasters, determining if there are ways to make predictions more accurate and ascertaining if natural disasters can be forecasted to give ample notice and thereby reduce the impacts. The first resource that I considered is a book by Kelleher & Tierney titled Data Science (2018). According to the authors, data science is a collection of principles, problem definitions, algorithms, and processes for discovering and extracting non-obvious and useful patterns from large data sets. They write that many aspects of data science have emerged in other fields related to it, such as machine learning and data mining, and that the terms data science, machine learning and data mining are often used interchangeably. Each of these fields is concerned with improving decision-making through data analysis. This book gives a deeper insight into the world of data science and explores the ways in which the field can help to extract different types of patterns, including clustering, prediction, and association-rule mining. The information retrieved from this book can be held to be reliable because it was published by the Massachusetts Institute of Technology, which is reputed to be one of the best schools in the world for Computer Science. The source is extremely relevant because understanding data science is necessary in understanding its role in predicting natural disasters. This understanding could also prompt further research into how its methods

Another useful resource that is directly linked to the previous one is ‘What is machine learning and why is it important?’ (Burns, 2021). This article explains that machine learning is a type of artificial intelligence that improves the accuracy of making predictions in software applications without being programmed to do so. It predicts output values by using data from the past as input. This source also raises fundamental questions about the accuracy of using data. It may be especially problematic using data related to natural disasters that have already occurred to predict future ones because natural disasters come in different scales and have many causes. The article ‘Why forecasts fail. What to do instead’ published by the Massachusetts Institute of Technology (2010) is another relevant resource to consider. It explains that forecasts tend to fail because statistical models help to extrapolate patterns from past data but don’t predict nearly as well. For example, it is impossible to predict the timing and location of large-scale earthquakes, but the intensity and frequency of earthquakes show consistent patterns. The article goes on to say that on average in each year, there are about 134 earthquakes measuring 6.0 to 6.9 on the Richter scale, approximately 17 with a measurement of 7.0 to 7.9 and one of 8.0 or above. However, it concludes by stating that statistical regularity does not mean they are predictable. There are places more prone to earthquakes, but seismologists do not know what the locations and timings of the earthquakes are going to be. It can be argued that the information included in this MIT article is not up to date because it was written 14 years ago and technology has evolved in the last decade. Although ‘ It is estimated that in the last ten years, the cause of death for one in a hundred people around the world was natural disasters ’

51