Real-time information sharing and propagation using social media such as Twitter has proven itself as a potential resource to improve situational awareness in a timely manner for disaster management. Traditional disaster management systems work well for analyzing static and historical information. However, they cannot process dynamic streams of data that are being generated in real-time. This paper presents STIMULATE - a System for Real-time Information Acquisition and Learning for Disaster Management that can (1) fetch and process tweets in real-time, (2) classify those tweets into FEMA defined categories for rescue priorities using pre-trained deep learning models and generate useful insights, (3) find FEMA defined stranded people for rescue missions of varying priorities, and (4) provide an interactive web interface for rescue management given the available resources. The STIMULATE prototype is primarily built using the Python Flask framework for web interaction. Additionally, it is deployed in the cloud environment using Hadoop and MongoDB for scalable storage, and on-demand computing for processing extensive social media data. The deep learning models in the STIMULATE prototype use Python Keras and the TensorFlow library. We use Bi-directional Long Short-Term Memory (BLSTM) and Convolutional Neural Network (CNN) for developing the tweet classifier. Further, we use the Python PyWSGI WebSocket server for rescue scheduling operations. We present a deep learning system trained on hurricane Harvey and Irma datasets only. The tweet classifier is evaluated using 15 different disaster datasets. Finally, we present the results of multiple simulations using synthetic data with different sizes to measure the performance and effectiveness of the tweets processor and rescue scheduling algorithm.

Every activity in disaster management demands accurate and up-to-date information to allow a quick, easy, and cost-effective response to reduce the possible loss of lives and properties. It is a challenging and complex task to acquire information from different regions of a disaster-affected area in a timely fashion. The extensive spread and reach of social media and networks such as Twitter allow people to share information in real-time. However, gathering of valuable information requires a series of operations such as (1) processing each tweet for the text classification, (2) possible location determination of people needing help based on tweets, and (3) priority calculations of rescue tasks based on the classification of tweets. These are three primary challenges in developing an effective rescue scheduling operation using social media data. In this paper, first, we propose a deep learning model combining attention based Bi-directional Long Short-Term Memory (BLSTM) and Convolutional Neural Network (CNN) to classify the tweets. Next, we perform feature engineering to create an auxiliary feature map which dramatically increases the model accuracy. In our experiments using data from Hurricanes Harvey and Irma, it is observed that our proposed approach performs better compared to other classification methods based on Precision, Recall, F1-score, and Accuracy, and is highly effective to determine the priority of a tweet. Furthermore, to evaluate the effectiveness and robustness of the proposed classification model a merged dataset comprises of 4 different datasets from CrisisNLP and another 15 different disasters data from CrisisLex are used. Finally, we develop an adaptive multi-task hybrid scheduling algorithm considering resource constraints to perform an effective rescue scheduling operation considering different rescue priorities.