Multi-Dimensional Business Data Fusion Modeling Based on Dynamic Bayesian Network

Business data fusion from various data sources in enterprises has been one of the hot issues in current research. While previous work has made great efforts to achieve data fusion, existing methods still suffer from three key limitations: (i) Most methods do not guarantee accurate connections to the data sources; (ii) Most approaches disregard the time-varying of business data; (iii) These approaches ignore the fact that the dimensionality of business data is high, resulting in query instances failing. To solve the above problems, we propose a novel multi-dimensional business data fusion method based on Dynamic Bayesian Networks, namely DBN. Specifically, we design a DBN for all the attributes in the data and generate a random variable for each attribute, which can be learned based on domain dependence and correlation or traditional structure. Then, we compute the query using the underlying joint distribution of the Bayesian Network. Moreover, we use Structured Query Language (SQL) engine in an Electric Motor Temperature multi-sensor business data source for query and analysis. The experimental results show that our method can achieve state-of-the-art performance and exhibit competitive accuracy of data source connections.