Extreme Climate Event Detection Through High Volume of Transactional Consumption Data

Abstract

Extreme weather events cause irreparable damage to society. At the beginning of 2017, the coast of Peru was hit by the phenomenon called “El Niño Costero”, characterized by heavy rains and floods. According to the United Nations International Strategy for Disasters ISDR, natural disasters comprise a 5-step process. In the last stage - recovery - strategies are aimed at bringing the situation back to normality. However, this step is difficult to achieve if one does not know how the economic sectors have been affected by the extreme event. In this paper, we use two well-known techniques, such as Autoregressive integrated moving average (ARIMA) and Kullback-Leibler divergence to capture a phenomenon and show how the key economic sectors are affected. To do this, we use a large real dataset from banking transactions stored in a Massively Parallel Processing (MPP). Our results show the interest of applying these techniques to better understand the impact of a natural disaster into economic activities in a specific geographical area.