Natural Disaster Detection, Response, and Feedback Management Using Machine Learning Techniques

Description

The machine learning techniques can be used with different formats of data i.e. text, image, and videos. Additionally, all these formats are very popular in social media data publishing. The social media data may have essential social information, which can be used for social welfare. In this work, social media text and image analysis techniques have been involved in designing a natural disaster detection system. This system is a machine learning application to assist the disaster responder’s team. This work includes the following three objectives:

1. A review of existing techniques has been proposed to perform. These techniques are based on machine learning and social media data, which are recently contributed by researchers.
2. There are some techniques are also available that utilize deep learning techniques. These techniques can work for both formats i.e. text and images. These techniques are known as multi-model fusion techniques. In this phase, the multi-model fusion technique has been explored for social media data analysis.
3. In this objective, a multi-model fusion approach has been developed to deal with text and image data. This approach is utilized to detect natural disaster events precisely. This model contains the system architecture to collect and analyze data, during different phases of natural disasters. Additionally, tries to generate notifications for the responders team to minimize the losses due to natural disasters.

The project includes one review paper writing work and two research articles (result papers). In addition, it includes two Python notebooks with the experimental consequences and fruitful results. There are two datasets have been utilized for demonstrating social media data processing and natural disaster event detection. Therefore, two different datasets have been used. The first is obtained from Kaggle, which consists of images to recognize the different disaster events. the second dataset is also taken from Kaggle to simulate how the image and text can be processed using the multi-model fusion technique.