Bayesian biclustering by dynamics: Algorithm testing, comparison against random agglomeration, and calculation of application specific prior information

Abstract

Bayesian Biclustering by Dynamics (BBCD) is a new clustering algorithm for Steam-Assisted Gravity Drainage (SAGD) oil recovery time series data [1]. In this companion paper the BBCD algorithm is tested on synthetic data, demonstrating use of the algorithm, as well as its robustness, and performance accuracy against Random Agglomeration. Supplementary information includes formulae to calculate analytical steam and oil volume data used as background knowledge for the SAGD application. Advantages of the BBCD algorithm are listed below. • It includes background knowledge directly into the clustering process. • It finds similarity between series and over time. • It allows a user-specified definition for behaviour of interest, which relaxes dependency on series shape. This is important when similar behavioural events do not necessarily occur in the same temporal order.

Keywords

• Bbcd
• Biclustering algorithm
• Bayesian statistics
• Steam-assisted gravity drainage (sagd) application