A Proposed Entropy based Recommender Framework for Disaster Management
Harita Ahuja, Sunita Narang, Veer Sain Dixit · 2021
Disaster management system employs an Early Warning System (EWS) to generate warnings. The limitations of existing early warning systems are that the quality of warning is missing, and usefulness of suggested actions has not been measured. A recommender system when integrated with an EWS has the potential to overcome it. This paper proposes an entropy-based recommender system which works offline on the warning web log obtained from an EWS. The warning web log essentially stores past warnings along with a set of pre-indicators and post-indicators of the warnings. By using novel pre-disaster similarity and post-disaster similarity between warning pairs, a warning knowledge base is generated. It consists of set of trustworthy warnings for each warning in the web log. The trustworthy warnings for the current warning issued by EWS collectively form the Nearest Neighbor Warning Set (NN-WSet) for the target user. The quality of the selection of NN-WSet governs the generation of priority of actions to be taken by the target user. Considering the fact that user has limited response time, this arrangement is capable of minimising pre-disaster risk. Here authors are focussing on floods only, however, the proposed framework is generic and can be applied to management of all kinds of disasters.