Researchers From Imperial College London Have Developed A New Machine Learning Model That Uses Social Media Data To Predict And Monitor Wildfires More Accurately In Real-Time

As social media use has grown over the last decade, people have developed a tendency to write about events occurring surrounding them on social media. So, there are vast amounts of data containing lots of information, for example, reporting disasters, spreading diseases, etc. In recent years, scientists and researchers are also leveraging social media to build systems to detect and predict natural disasters. A group of scientists at Imperial College, UK, has made the very first system for real-time detection of wildfires using physical sensor data and social media data.

The researchers have used Twitter data to create a wildfire dataset using advanced filtering and query optimization methods. Then they performed sentimental analysis on those Twitter data to extract human sentiments associated with wildfires burning. This results in a robust dataset containing not only regional wildfire data but also local wildfire data. Several ML algorithms further process this high-quality dataset to predict wildfire best, creating a robust disaster response system.


The researchers have used historical satellite wildfire data containing various physical attributes in this work. They have used some of these attributes, namely latitude, longitude, and start and end time of wildfire, to generate queries for Twitter data and to select relevant tweets. For maximum relevance of the tweet, the researchers have used several criteria to select tweets, like if the tweet contains relevant keywords and hashtags. However, that could also contain noisy tweets; for example, ‘fire’ ‘burn’ etc. can express several meanings, from political views to several other analogies. Hence, the tweets are again filtered to remove these noises.

The filtered tweets are passed through NLP (Natural Language Processing) models to generate social sentiment variables, namely Sentiment score (it describes the emotional attribute of the text, a value ranging between -1 to 1) and Sentiment Magnitude (it describes Magnitude of the emotional attribute, value ranges between 0.0 to +inf. For an individual post, the score and magnitude are computed by averaging all the scores and magnitudes of all words. Then the ML models are used to learn mapping to predict social sentiments towards a wildfire from the physical attributes of the fire and vice-versa. The researchers have experimented with Gradient-Boosted Random Forest, Neural Network Regressor, Support Vector Regressor, etc in the prediction module.

Thus, researchers have demonstrated a relationship between online social sentiments and wildfire activity in this work. This would help in robust wildfire prediction, localizing dangerous areas as well as emergency services. In the future, this approach can be further applied to other natural events and disasters.

This Article is written as a research summary article by Marktechpost Staff based on the research paper 'Sentimental wildfire: a social-physics machine learning model for wildfire nowcasting'. All Credit For This Research Goes To Researchers on This Project. Checkout the paper and reference article.

Please Don't Forget To Join Our ML Subreddit

I'm Arkaprava from Kolkata, India. I have completed my B.Tech. in Electronics and Communication Engineering in the year 2020 from Kalyani Government Engineering College, India. During my B.Tech. I've developed a keen interest in Signal Processing and its applications. Currently I'm pursuing MS degree from IIT Kanpur in Signal Processing, doing research on Audio Analysis using Deep Learning. Currently I'm working on unsupervised or semi-supervised learning frameworks for several tasks in audio.

[Announcing Gretel Navigator] Create, edit, and augment tabular data with the first compound AI system trusted by EY, Databricks, Google, and Microsoft