An improved B&B technique applying to telecom industry

The data mining technology is more and more widely used in the telecom industry. When we construct Bayesian Belief Network, the branch and bound technique based on the minimum description length principle (B&B technique) is one of the classical algorithm. To telecom data, high dimensionality and huge volume set obstacle when constructing Bayesian Belief Network. But we utilize the inconspicuous correlation between telecom attributes, improve the process of B&B technique and simplify the structure to solve complexity rooting from telecom data´s feature. The algorithm first construct a dependence ordering, then a simplified B&B technique suitable for telecom data is applied. We compare the result and complexity with the original B&B technique. This paper uses real datasets from the telecom industry. The result shows that the new algorithm can construct the network almost the same as the original one, but with good performance.