Earthquake Aftershock Prediction Based Solely on Seismograph Data

Important contributions of this work are: (1) Earthquake aftershock prediction, both location and time of occurrence, using seismographic data only; (2) Representing seismographic data as a directed weighted graph and analyzing it to extract relevant data for localized prediction; (3) Validation of the data and prediction algorithm. The key contribution is to make aftershock prediction using seismographic data alone. Seismographic data units are represented as nodes of a graph, and a special relation between them depicts the directed weighted edge between two nodes. This special relation facilitates connecting main earthquake with aftershocks caused by it. Converting vast amount of seismographic data to corresponding connected graph, and efficiently analyzing the graph to find cluster of nodes as hubs, is the first step. Though the link weights are real, in this work we consider whether two nodes are connected or not, depending on the connection weight is above some threshold or not. Extracting data of a hub, to use it for aftershock prediction of that individual localized area, is proposed. Data Driven approach is better because it is universally applicable and does not require domain knowledge (here, geophysical and geological knowledge of the plate structure). The idea could be extended for mining any big data.