Study on Feature Extraction in China Railway Ticketing and Reservation System
Xiaoyan Lui, Chunhuang Liu, Weiwei Wang · 2007
Because that data analysis methods for train ticket data are mostly designed for building predictive analysis models, which are always good at describing the characteristics of major classes but are lack of reflecting the minorities, this paper presents a new method FEBIR that is based on the set-partition for data feature extraction. The presented method can distill the pointed favorite class features without the limitations of current analysis methods in characterizing the minors. The characteristic rules which are extracted by this method include quantitative information, and the order of attributes in the rules reflects how importantly they contribute to sculpture the class, so it provides enough information for decision-makers to analyze the special class and will be an effective tool for railway managers to get useful information about their focuses.