Title: Analysis of community engagement on social media during disasters: a machine learning approach

Authors: Arun Thomas; Sumit Kumar; Vinay V. Panicker

Addresses: Supply Chain and Systems Simulation Laboratory, Department of Mechanical Engineering, National Institute of Technology Calicut, Kerala, India ' Supply Chain and Systems Simulation Laboratory, Department of Mechanical Engineering, National Institute of Technology Calicut, Kerala, India ' Supply Chain and Systems Simulation Laboratory, Department of Mechanical Engineering, National Institute of Technology Calicut, Kerala, India; Centre of Excellence in Logistics and Supply Chain Management, National Institute of Technology Calicut, Kerala, India

Abstract: Recently social networking sites have emerged as an imperative source for providing information to identify and track disasters. The utilisation of social networking data streams to extract meaningful data is possible. A proper data gathering and an evaluation protocol is required for a predictive model. With this purpose, the features of tweets during the Kerala floods of 2018 were examined in this research. About 154,524 tweets were collected using the API during the period between July 2018 and February 2019. Data visualisation has been done with the help of Tableau, and the most commonly used words were identified using a dataset of 10,704 tweets. A topic modelling approach was adopted for exploring and recognising the unlabelled topics in the dataset. A typical machine learning technique is designed for the sentiment analysis of tweets. The long short-term memory (LSTM) model has been developed to find the accuracy of the data.

Keywords: Twitter API; machine learning; topic modelling; data analysis; sentiment analysis; tweets; binary classification; long short-term memory; LSTM.

DOI: 10.1504/IJSOM.2025.148832

International Journal of Services and Operations Management, 2025 Vol.52 No.1, pp.1 - 24

Received: 21 Feb 2023
Accepted: 11 Jul 2023
Published online: 29 Sep 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article