Pakistan Journal of Engineering Technology and Science (PJETS) Volume 11. Issue. 1, PP. 45-73, November 2023 E-ISSN 2224-2333 https://doi.org/10.22555/pjets.v11i1.1004 P-ISSN 2222-9930 45 EXPLORING MACHINE LEARNING AND DEEP LEARNING APPROACHES FOR DISASTER PREDICTION AND MANAGEMENT: A SURVEY OF DIFFERENT APPROACHES Hira Farman 1 , Noman Islam 2 , Affan Alim 3 1,2 Faculty of Computing, Iqra University, Karachi Institute of Economics and Technology, Karachi, Pakistan 3 Department of Computer Science, Mohammad Ali Jinnah University, Karachi, Pakistan * Corresponding author: hira.farman@iqra.edu.pk Abstract: Natural disasters affect both human and animal existence everywhere in the world. Not to mention seriously damaging land, wildlife, etc. Thousands of people worldwide lose their lives to natural catastrophes every year, including landslides, cloudbursts, heat waves, storms, tsunamis, floods, earthquakes, and wildfires. On the social networking site Twitter, individuals can exchange thoughts, news, and personal narratives. Real-time data is widely available, and many service agencies routinely use it to identify emergencies, reduce risk, and save lives. However, it has been suggested in numerous studies to provide words in forms that computers can understand and, on the basis of word representations, use machine learning techniques to accurately determine the meaning behind a post. This is because humans are unable to manually filter through the vast number of records and spot hazards in real-time. By posting risks associated with disaster events on social media, the community can keep an eye out for disasters, which has been crucial for emergency preparedness. This study looks at the potential use of the social media site Twitter for disaster-related research. It focuses on the most recent methods for disaster prediction, deep learning, and machine learning. Another objective of the effort is to have a comprehensive understanding of the various data types and their sources in connection to a range of jobs and crisis management scenarios. Furthermore, the study intends to provide a thorough analysis of the different data mining techniques applied to address diverse problems associated with natural disasters in addition to thorough instructions on how to classify tweets as "Related to Catastrophe" or "Not related with Catastrophe" using natural processing techniques. Keywords: Tweets, machine earning, deep learning, big data, natural disasters, NLP I. INTRODUCTION On Earth, natural catastrophes are a significant, harmful, and unbearable phenomenon that frequently occurs [1,2,3]. Disasters, both natural (the power of nature) and man-made (accidental, intentional), have an impact on the lives of millions of individuals, including both humans and other animals. Besides, human lives are lost, with significant implications on infrastructures in terms of property and political stability [4,5]. Exposure to a natural disaster rises in recent months in children under the age of five by 9-18%, increasing their likelihood of getting serious illnesses such as diarrhea, temperature, and extreme respiratory disease. This is in addition to the immediate effect seen. The scope and pattern of these effects are directly related to the socioeconomic condition of the households. The disasters also have significant effects on business establishments as already discussed. Disaster management and monitoring effectively is a global challenge. Both natural and man-made tragedies can affect any community. A disaster is described as an unplanned and frequently sudden event that results in 1 significant damage, destruction, death, and suffering and calls for external assistance at the national or worldwide level. Flood, wildfires, hurricanes, storms, chemical and oil spills, terrorist attacks, nuclear accidents, and other This is an open access article published by CCSIS, IoBM, Karachi Pakistan under CC BY 4.0 International License