A Bayesian Network Incremental Algorithm for Public Safety Data Analysis

Bayesian network learning is another research field of machine learning and data mining. In the research of Bayesian network learning algorithm, on the one hand we have to consider the issue of how to avoid the disclosure of data privacy. On the other hand, in the real world applications, data may be gradually available for Bayesian network, thus traditional Bayesian network learning algorithms can not be effectively applied. Therefore, this paper proposes a privacy-preserving Bayesian network incremental learning algorithm. We use sufficient statistical information in our incremental learning method, so we firstly proposed a formula for calculating the amount of sufficient statistics information. Then we improve the traditional K 2 algorithm, add the concept of sufficient statistical information in it, and then propose an incremental K 2 algorithm. At last, we proposed a privacy-preserving Bayesian network incremental learning algorithm. The algorithm only need to save the sufficient statistical information of each node and its possible parents in order to calculate score function value of each node and its parents, and then construct the Bayesian network structure. We can use the proposed approach to complete the public safety data analysis with a higher efficiency.