A Post-Disaster Demand Forecasting System Using Principal Component Regression Analysis and Case-Based Reasoning Over Smartphone-Based DTN

The most dominant challenges in post-disaster emergency resource planning are forecasting the exact demand for emergency resources, communicating such demands to the control station, and validating these demands before using them for resource planning. Forecasting the exact demand for resources in a relief shelter becomes tricky, because situational parameters influencing these demands keep on changing. Moreover, the collection and transmission of demands of far-flung shelters are challenging, owing to the post-disaster disruption of communication infrastructure. All these lead to ad hoc allocation of emergency resources to the shelters. In this paper, we first derive a principal component regression model to forecast demand for emergency resources based on situational parameters at the shelters. Subsequently, we propose an opportunistic demand sharing scheme for gathering and disseminating resource demands to the control station using a smartphone-based delay-tolerant network (DTN). Finally, we suggest a case-based reasoning driven demand validation technique to ratify these demands and also to project the demands that do not get transmitted. Experimental results show that our system forecasts dynamically changing resource demands at the shelters with high precision, transmits near accurate demands to the control station, and perfectly validates the received resource demands.