A Hybrid Business Outlier Detection Algorithm Basing on Creative Computing Methods

Business outliers can influence the development of a company negatively. Detecting the outliers is always an essential topic. Traditionally, the outlier detection can be completed by using statistical analysis software or just manual analysis. Low efficiency is one of the disadvantages of such an approach. In this research, an outlier detection algorithm is established through combining three different types of approaches with a creative computing method. The hybrid business outlier detection algorithm uses the statistical approach to pre-process the data, PCA algorithm to discover the main components and BP neural network to complete the outliers separating. Each part can reach an efficient working condition in an appropriate place in this algorithm. Meanwhile, BP neural network and PCA can improve the accuracy of the outlier detecting results.