A Social Community Sensor for Natural Disaster Monitoring in Indonesia Using Hybrid 2D CNN LSTM.

One of the most notable advantages of utilizing social media is its ability to aid communication during natural disasters, specifically in regards to disseminating information. Social media users serve as community monitors by collecting public messages from a multitude of social media platforms. The implementation of artificial intelligence can facilitate the automatic recognition of messages concerning natural disasters. Deep learning-based artificial intelligence models for text classification, such as Convolutional Neural Network (CNN) and Long Short-Term Memory networks (LSTM), each possess their own unique strengths and weaknesses. Nevertheless, the utilization of word padding techniques presents a further challenge to accurately classify texts, as its implementation may negatively impact classification performance. In this study, feature extraction based on word embedding and word padding techniques were employed, utilizing a maximum number of words to be processed by a hybrid 2D CNN LSTM model. The results indicate the highest level of accuracy in cases of floods, with an 81.27% accuracy rate, forest fires with an 86.14% accuracy rate, and earthquakes with an 80.16% accuracy rate. This outcome represents a significant advancement over the accuracy attained by the classification model constructed with feature extraction based on word embedding and word padding based on mean and using solely 2D CNN.