Time filter

Source Type

Beijing, China

Lin Y.,Beijing Jiaotong University | Wan H.,Tsinghua University | Jiang R.,Beijing Jiaotong University | Wu Z.,Beijing Jiaotong University | Jia X.,TravelSky Technology Ltd
IEEE Transactions on Intelligent Transportation Systems | Year: 2014

People usually travel to the same destinations and for the same purposes together with other people in groups. Inferring the travel purposes of passenger groups is a very interesting research problem in the field of passenger transport, because it can help us to better understand passengers and should bring about meaningful changes for personalized travel service and decision making of passenger carriers, organizations, and even governments. In this paper, we attempt to solve this problem by utilizing the historical travel records of passengers. To overcome the constraint of the independent and identical distribution assumption of traditional classifiers, we propose a novel iterative classification approach based on the idea of collective inference. First, we construct cotravel networks by extracting social relations between passengers from their historical travel records that are available in carriers' passenger information systems. Then, we generate a series of sophisticated features for each passenger group in the context of cotravel networks to capture the link structure information between passengers and use the overlapping relations between passenger groups to capture the probabilistic dependence relations between their labels. Finally, we collectively infer the labels of all the groups in an iterative way. Experimental results on a real data set of passenger travel records in the field of civil aviation demonstrate that our proposed iterative classification approach can efficiently infer the travel purposes of passenger groups. © 2014 IEEE. Source

Gao M.,Beijing University of Technology | Chen F.,TravelSky Technology Ltd
2013 25th Chinese Control and Decision Conference, CCDC 2013 | Year: 2013

This paper proposed an interactive query mode for an OWL-oriented question-answering system, named Agile, to refine answers. The key problem for interactive query modes in QAs is producing natural language sub-questions. This paper used a two-phases semantic mapping method to produce sub-questions. The first phase builds RDF triples based on existing OWL knowledge and given original questions. The second phase maps RDF triples into natural language sub-questions through some lingual common sense and predefined rules. To validate effect of interactive queries, we have conducted some experiments involving two different domain of knowledge bases and accepted some promising results. © 2013 IEEE. Source

Yang Y.,East China Normal University | Guan H.,TravelSky Technology Ltd | Xu W.,East China Normal University
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2014

Most previous researches on out-of-vocabulary words recognition are concentrating on traditional areas, while in electronic commerce area this recognition is rarely involved. In this paper, we focus on the out-of-vocabulary words recognition in the field of electronic commerce. Based on the unique characteristics of electronic commerce data, we introduce the Conditional Random Fields (CRFs) into the out-ofvocabulary words recognition, and use the electronic commerce text corpus for the experimental verification. The experimental results show that CRFs in the recognition of out-of-vocabulary words in the field of electronic commerce is effective. © Springer International Publishing Switzerland 2014. Source

Ding J.,Civil Aviation University of China | Zhang C.,Civil Aviation University of China | Wang J.,Civil Aviation University of China | Wang Y.,TravelSky Technology Ltd
Journal of Software Engineering | Year: 2014

Alarm association rules mining is an important task in system fault hagnosis and localization. Once the system fails, it will produce a large number of alarm information. By analyzing the characteristics of the booking system alarm data, this study puts forward alarm association rules mining algorithm based on sliding time window model to find the fault source and the correlation between fault factors in a large number of alarm information. The experiments show that the valuable alarm association rules can be acquired from the alarm data accurately and rapidly. These rules can provide support decision for the system maintenance personnel. © 2014 Academic Journals Inc. Source

Huang W.,Beijing Technology and Business University | Jia X.,TravelSky Technology Ltd | Tian F.,Beijing Technology and Business University | Zhang Y.,Beijing Technology and Business University | Zhou Z.,Beijing Technology and Business University
International Journal of Multimedia and Ubiquitous Engineering | Year: 2015

Customer churn analysis has become an important focus of corporate marketing. It will be a great help to profitability if there is a method can find losing customers in time. In the paper, a method based on RFM and Cross-correlation model is proposed. Firstly, the customer’s value is calculated by RFM. Secondly, the typical losing curves of customer value are matched via cross-correlation. And finally, integrated with social network analysis (SNA) and community detection, the group of potential losing customers are revealed. The effectiveness of the presented method has been proven in a dataset of retail sales records. © 2015 SERSC. Source

Discover hidden collaborations