An approach to improve collusion set detection using MCL algorithm

Many malpractices in stock market trading e.g. price manipulation, circular trading, use the modus-operandi of collusion. Generally, a set of traders is a candidate collusion set when they are “trading heavily” among themselves in cross trading or circular trading. In real life not all colluders always trade with each other. In a perfectly circular collusion set of size 4, trader A will trade with B, B with C, C with D and D with A; there will be no cross trading among these traders. An existing method using shared, mutual nearest neighbor and collusion graph clustering algorithm fails to detect purely circular trading which is also a collusion set. In this paper, we have proposed a new approach to detect collusion sets using Markov Clustering Algorithm (MCL). Proposed method can detect purely circular collusions as well as cross trading collusions. We have used MCL at various strength of “residual value” to detect different cluster sets from the same stock flow graph. We have combined our collusion clusters with the existing method using Dempster Schafer theory of evidence. The experimental result shows that MCL algorithm provides better collusion clusters and the performance improved significantly.