Forecasting large avalanches in Bak, Tang and Wiesenfeld sandpile model
Leow, Jackson Yu Sheng
Date of Issue2017
School of Physical and Mathematical Sciences
This project attempts to forecast large relaxation events in complex systems, and in particular, the Bak-Tang-Wiesenfeld (BTW) Sandpile Model. While this unique Self- Organised Criticality (SOC) model has been intensively studied, due to the random nature of its external perturbation, few believe in the feasibility of predicting large relaxation events in such complex systems. That being said, with recent findings on the predictability of stock market crashes by Cheong et al., we are motivated to propose otherwise. Given that a large avalanche event requires time to build up, history of past avalanche events and interactions among sites can provide crucial information for further studies. The information can be extracted out from sliding-window correlation matrices, leading up to a large relaxation event for analysis. The methods for analysis include the Fusion-Fission Model by Johnson et al., Hierarchical Clustering by Mantegna et al. and Interaction Hierarchical Clustering by Cheong et al. Results computed from these methods have shown unique cluster dynamic behaviour and provides insight to properties that are crucial before the occurrence of a large relaxation event. We believe that this serves as a springboard for prediction approaches in other Self- Organised Criticality (SOC) systems.
Final Year Project (FYP)