Parallel implementation in case-based reasoning bankruptcy prediction system

One of the strategies for enhancing the high performance computing in machine learning is by implementing the parallel computation, both multi-cores in Central Processing Unit (CPU) or multi-processor in Graphics Processing Unit (GPU). That evaluation is measured by how much the computation time is consumed. Case-based Reasoning (CBR) is a data historical based prediction algorithm which has satisfactory results but has low performance at the computation time. Due to the enormous financial historical data, the original CBR seems slow. This research shows how to speed up the computational time by parallelizing CBR calculations while maintaining accuracy in a bankruptcy prediction system. The computation time ratio of original sequential CBR, multi-cores CPU and multi-processor GPU is about 540:74:1. The experimental outputs prove that the strategy of parallelization success without reducing much of the classification performance. The parallel-CBR algorithm has 81% of accuracy.