UTS site search

Li Liu

Contractor, A/DRsch Advanced Analytics Institute
+61 2 9514 4499


Luo, C., Zhao, Y., Luo, D., Ou, Y. & Liu, L. 2010, 'Recent Advances of Exception Mining in Stock Market' in Pedro Furtado (ed), Evolving Application Domains of Data Warehousing and Mining: Trends and Solutions, IGI Global, Washington, DC, USA, pp. 212-232.
View/Download from: UTS OPUS or Publisher's site
This chapter aims to provide a comprehensive survey of the current advanced technologies of exception mining in stock market. The stock market surveillance is to identify market anomalies so as to provide a fair and efficient trading platform. The technologies of market surveillance developed from simple statistical rules to more advanced technologies, such as data mining and artificial intelligent. This chapter provides the basic concepts of exception mining in stock market. Then the recent advances of exception mining in this domain are presented and the key issues are discussed. The advantages and disadvantages of the advanced technologies are analyzed. Furthermore, our model of OMM (Outlier Mining on Multiple time series) is introduced. Finally, this chapter points out the future research directions and related issues in reality.


Yang, Y., Cao, L. & Liu, L. 2010, 'Time-Sensitive Feature Mining for Temporal Sequence Classification', Lecture Notes in Artificial Intelligence 6230 - PRICAI 2010: Trends in Artificial Intelligence, Pacific Rim International Conference on Artificial Intelligence, Springer-Verlag Berlin Heidelberg, Daegu, Korea, pp. 315-326.
View/Download from: UTS OPUS or Publisher's site
Behavior analysis received much attention in recent year, such as customer-relationship management, social security surveillance and e-business. Discovering high impact-driven behavior patterns is important for detecting and preventing their occurrences and reducing resulting risks and losses to our society. In data mining community, researchers pay little attention to time-stamps in temporal behavior sequences (without explicitly considering inherent temporal information) during classification. In this paper, we propose a novel Temporal Feature Extraction Method - TFEM. It extracts sequential pattern features where each transition is annotated with a typical transition time (its duration or interval). Therefore it substantially enriches temporal characteristics derived from temporal sequences, yielding improvements in performances, as demonstrated by a set of experiments performed on synthetic and real-world datasets. In addition, TFEM has the merit of simplicity in implementation and its pattern-based architecture can generate human-readable results and supply clear interpretability to users. Meanwhile, it is adjustable and adaptive to userâs different configurations, allowing a tradeoff between classification accuracy and time cost.
Ou, Y., Cao, L., Luo, C. & Liu, L. 2008, 'Mining Exceptional Activity Patterns in Microstructure Data', 2008 IEEE/WIC/ACM international Conference on Web Intelligence and Intelligent Agent Technology, IEEE/WIC/ACM international Conference on Web Intelligence and Intelligent Agent Technology, IEEE Computer Society, University of Technology, Sydney, Australia, pp. 884-887.
View/Download from: UTS OPUS or Publisher's site
Market Surveillance plays an important role in maintaining market integrity, transparency and fairnesss. The existing trading pattern analysis only focuses on interday data which discloses explicit and high-level market dynamics. In the mean time, the existing market surveillance systems are facing challenges of misuse, mis-disclosure and misdealing of information, announcement and order in one market or crossing multiple markets. Therefore, there is a crucial need to develop workable methods for smart surveillance. To deal with such issues, we propose an innovative methodology -- microstructure activity pattern analysis. Based on this methodology, a case study in identifying exceptional microstructure activity patterns is carried out. The experiments on real-life stock data show that microstructure activity pattern analysis opens a new and effective means for crucially understanding and analysing market dynamics. The resulting findings such as exceptional microstructure activity patterns can greatly enhance the learning, detection, adaption and decision-making capability of market surveillance.
Qin, Z., Zhang, S., Liu, L. & Wang, T. 2008, 'Cost-sensitive Semi-supervised Classification using CS-EM', Proceedings of 2008 IEEE 8th International Conference on Computer and Information Technology, IEEE International Conference on Computer and Information Technology, IEEE Computer Society, Sydney, Australia, pp. 131-136.
View/Download from: UTS OPUS or Publisher's site
In many real world data mining and classification tasks, we face with the problem of high cost in making training data sets. In addition, in many domains, different misclassification errors involve different costs. These two issues are often addressed by semi-supervised learning and costsensitive learning separately. Sometimes the two issues can happen at the same time in real world applications. However, existing semi-supervised learning algorithms never consider the misclassification costs. In this paper, we propose a simple and novel method, CS-EM for learning cost-sensitive classifier using both labeled and unlabeled training data. CS-EM modifies EM, a popular semi-supervised learning algorithm by incorporating misclassification costs into the probability estimation process. Our experiments show that CS-EM outperforms other two competing methods on three bench mark text data sets across different cost ratios.
Luo, C., Zhao, Y., Cao, L., Ou, Y. & Liu, L. 2008, 'Outlier Mining on Multiple Time Series Data in Stock Market', PRICAI 2008: Trends in Artificial Intelligence, Pacific Rim International Conference on Artificial Intelligence, Springer, Hanoi, Vietnam, pp. 1010-1015.
View/Download from: UTS OPUS or Publisher's site
In stock market, the key surveillance function is identifying market anomalies, such as insider trading and market manipulation, to provide a fair and efficient trading platform [2,6]. Insider trading refers to the trades on privileged information unavailable to the public [8]. Market manipulation refers to the trade or action which aims to interfere with the demand or supply of a given stock to make the price increase or decrease in a particular way [3]. Recently, new intelligent technologies are required to deal with the challenges of the rapid increase of stock data. Outlier mining technologies have been used to detect market manipulation and insider trading . The objective of outlier mining is to find the data objects which are grossly different from or inconsistent with the majority of data. However, in stock market data, outliers are highly intermixed with normal data [4] and it is difficult to judge whether an object is an outlier or not. Therefore, a more effective and more efficient approach is in demand. This paper presents a new technique for outlier detection on multiple time series data in stock market. At first, principal curve algorithm is used to detect the outliers from individual measurements of stock market. Then, the generated outliers are measured with the probability of being real alerts. To improve the accuracy and precision, these outliers are combined by some rules associated with the domain knowledge. The experimental results on real stock market data show that the proposed model is feasible in practice and achieves a higher accuracy and precision than traditional methods
Zhang, S., Liu, L., Zhu, X. & Shan, C. 2008, 'A Strategy for Attributes Selection in Cost-Sensitive Decision Trees Induction', CIT Workshops 2008, IEEE International Conference on Computer and Information Technology, IEEE Computer Society, Sydney, Australia, pp. 8-13.
View/Download from: UTS OPUS or Publisher's site
Decision tree learning is one of the most widely used and practical methods for inductive inference. A fundamental issue in decision tree inductive learning is the attribute selection measure at each non-terminal node of the tree. However, existing literatures have not taken both classification ability and cost-sensitive into account well. In this paper, we present a new strategy for attributes selection, which is a trade-off method between attributesâ information and cost-sensitive learning including misclassification costs and test costs with different units, for selecting splitting attributes in cost-sensitive decision trees induction. The experimental results show our method outperform than the existing methods, such as, information gain method, total costs methods, in terms of the decrease of misclassification costs with different missing rate and various costs in UCI datasets.
Luo, D., Cao, L., Ni, J. & Liu, L. 2007, 'Building Agent Service Oriented Multi-Agent Systems', Agent and Multi-Agent Systems: Technologies and Applications, International KES Symposium on Agents and Multiagent systems - Technologies and Applications, Springer, Wroclaw, Poland, pp. 11-20.
View/Download from: UTS OPUS or Publisher's site
An effective agent-based design approach is significant in engineering agent-based systems. Existing design approaches meet with challenges in designing Internet-based open agent systems. The emergence of service-oriented computing (SOC) brings in intrinsic mechanisms for complementing agent-based computing (ABS). In this paper, we investigate the dialogue between agent and service, and between ABS and SOC. As a consequence, we synthesize them and develop a design approach called agent service-oriented design (ASOD). The ASOD consists of agent service-based architectural design and detailed design. ASOD expands the content and range of agent and ABS, and synthesizes the qualities of SOC such as interoperability and openness, and the performances of ABC like flexibility and autonomy. The above techniques have been deployed in developing an online trading and mining support infrastructure F-Trade.
Kong, X., Liu, L. & Lowe, D.B. 2005, 'Web system trace model using a web application architecture framework', Proceedings of 2005 IEEE International Conference On E-Technology, E-Commerce And E-Service, IEEE International Conference on e-Technology, e-Commerce and e-Service, IEEE Computer Society, Conference Publishing Services, Hong Kong, China, pp. 508-513.
View/Download from: UTS OPUS
Lin, L., Luo, D. & Liu, L. 2005, 'Mining Domain-Driven Correlations in Stock Markets', AI 2005: Advances in Artificial Intelligence, 18th Australian Joint Conference on Artificial Intelligence Proceedings, Australasian Joint Conference on Artificial Intelligence, Springer, Sydney, Australia, pp. 979-982.
View/Download from: UTS OPUS or Publisher's site
There have been many technical trading rules in stock market since the first stock exchange founded. Along with the developing of computer technology, the technical trading rules are playing more and more important roles in the stock market trading system. However, there are many problems also occurred, such as the huge database, inefficiency, etc. So, the in-depth data mining technology is becoming a powerful tool to overcome the shortage of the current technologies. In this paper, we give some applications of in-depth data mining method: to find the optimal range, to find the stock-rule pair and find the relationship between the number of pair and investment. This method can improve both efficiency and effectiveness.
Li, C. & Liu, L. 2005, 'MAHIS: An Agent-Oriented Methodology for Constructing Dynamic Platform-Based HIS', AI 2005: Advances in Artificial Intelligence, Australasian Joint Conference on Artificial Intelligence, Springer, Sydney, Australia, pp. 705-714.
View/Download from: UTS OPUS or Publisher's site
Hierarchical structure, reusable and dynamic components, and predictable interactions are distinct characteristics of hybrid intelligent systems (HIS). The existing agent-oriented methodologies are deficient in HIS construction because they did not take into account the characteristics of HIS. In this paper, we propose a Methodology for constructing Agent-based HIS (MAHIS). MAHIS consists of eight models: Hybrid Strategy Identification Model, Organization Model, Task Model, Agent Model, Expertise Model, Coordination Model, Reorganization Model, and Design Model. The Reorganization Model is the key model to support dynamic platform-based HIS. It consists of category role, group roles, virtual organization role, and dynamics rules. This model describes the characteristics of HIS with virtual organization, category, and group perspectives. Some previously developed agents can be reused by means of involving them in a new virtual organization dynamically. The output of the Reorganization Model is the specification of the dynamic platform which comprises middle agents and makes all agents and agent groups hierarchical and dynamic.
Zhang, S., Liu, L., Lu, J. & Ou, Y. 2004, 'Is minimum-support appropriate to identifying large itemsets?', Pricai 2004: Trends In Artificial Intelligence, Proceedings, Pacific Rim International Conference on Artificial Intelligence, Springer-Verlag Berlin, Auckland, New Zealand, pp. 474-484.
View/Download from: UTS OPUS
Cao, L., Luo, C., Luo, D. & Liu, L. 2004, 'Ontology services-based information integration in mining telecom business intelligence', Pricai 2004: Trends In Artificial Intelligence, Proceedings, Pacific Rim International Conference on Artificial Intelligence, Springer-Verlag Berlin, Auckland, New Zealand, pp. 85-94.
View/Download from: UTS OPUS
Qin, Z., Liu, L. & Zhang, S. 2004, 'Mining Term Association Rules for Heuristic Query Construction', Advances in Knowledge Discovery and Data Mining, 8th Pacific-Asia Conference, PAKDD 2004, Pacific-Asia Conference on Knowledge Discovery and Data Mining, Springer, Sydney, Australia, pp. 145-154.
View/Download from: UTS OPUS or Publisher's site
As the Web has become an important channel of information floods, users have had difficulty on identifying what they really want from huge amounts of rubbish-like information provided by the Web search engines when utilizing the Webrsquos low-cost information. This is because most users can only give inadequate (or incomplete) expressions for representing their requirements when querying the Web. In this paper, a heuristic model is proposed for tackling the inadequate query problem. Our approach is based on the potentially useful relationships among terms, called term association rules, in text corpus. For identifying quality information, a constraint is designed for capturing the goodness of queries. The heuristic information in our model assists users in expressing their queries desired.
Cao, L., Luo, D., Luo, C. & Liu, L. 2004, 'Ontology Transformation in Multiple Domains', AI 2004: Advances in Artificial Intelligence, 17th Australian Joint Conference on Artificial Intelligence Cairns, Australia, December 2004 Proceedings, Australasian Joint Conference on Artificial Intelligence, Springer, Cairns, Australia, pp. 985-990.
View/Download from: UTS OPUS or Publisher's site
We have proposed a new approach called ontology services-driven integration of business intelligence (BI) to designing an integrated BI platform. In such a BI platform, multiple ontological domains may get involved, such as domains for business, reporting, data warehouse, and multiple underlying enterprise information systems. In general, ontologies in the above multiple domains are heterogeneous. So, a key issue emerges in the process of building an integrated BI platform, that is, how to support ontology transformation and mapping between multiple ontological domains. In this paper, we present semantic aggregations of semantic relationships and ontologies in one or multiple domains, and the ontological transformation from one domain to another. Rules for the above semantic aggregation and transformation are described. This work is the foundation for supporting BI analyses crossing multiple domains.

Journal articles

Sun, D., Liu, L., Zhang, P., Zhu, X. & Shi, Y. 2011, 'Decision Rule Extraction For Regularized Multiple Criteria Linear Programming Model', International Journal of Data Warehousing and Mining, vol. 7, no. 3, pp. 88-101.
View/Download from: UTS OPUS or Publisher's site
Due to the flexibility of multi-criteria optimization, Regularized Multiple Criteria Linear Programming (RMCLP) has received attention in decision support systems. Numerous theoretical and empirical studies have demonstrated that RMCLP is effective and efficient in classifying large scale data sets. However, a possible limitation of RMCLP is poor interpretability and low comprehensibility for end users and experts. This deficiency has limited RMCLP's use in many real-world applications where both accuracy and transparency of decision making are required, such as in Customer Relationship Management (CRM) and Credit Card Portfolio Management. In this paper, the authors present a clustering based rule extraction method to extract explainable and understandable rules from the RMCLP model. Experiments on both synthetic and real world data sets demonstrate that this rule extraction method can effectively extract explicit decision rules from RMCLP with only a small compromise in performance.
Li, C., Liu, L. & Song, Q. 2004, 'A Practical framework for agent-based hybrid intelligent systems', Asian Journal of Information Technology, vol. 3, no. 2, pp. 107-114.
View/Download from: UTS OPUS
The design and development of hybrid intelligent systems (HIS) are difficult because there are many interactions between various components. Existing software development techniques cannot manage those complex interactions efficiently as those interactions may occur at unpredictable times, for unpredictable reasons, between unpredictable components. In this study, we contribute a multi-agent framework and a ring based architectural model to organize those components and interactions. The framework consists of user interface, decision making, knowledge discovering, information management facilitators, and distributed heterogeneous data resources IDHDR). We employ middle agent concept to match task requesters with specific intelligent agents. A ring-based architectural model to organize the middle agents in HIS is developed. We demonstrate the potentials of the framework by case study and present theoretical and empirical evidence that our framework is available and robust.
Zhang, S. & Liu, L. 2003, 'Mining Dynamic Databases by Weighting', Acta Cybernetica, vol. 16, no. 1, pp. 179-205.
View/Download from: UTS OPUS
A dynamic database is a set of transactions, in which the content and the size can change over time. There is an essential difference between dynamic database mining and traditional database mining. This is because recently added transactions can be more 'interesting' than those inserted long ago in a dynamic database. This paper presents a method for mining dynamic databases. This approach uses weighting techniques to increase efficiency, enabling us to reuse frequent itemsets mined previously. This model also considers the novelty of itemsets when assigning weights. In particular, this method can find a kind of new patterns from dynamic databases, referred to trend patterns. To evaluate the effectiveness and efficiency of the proposed method, we implemented our approach and compare it with existing methods.
Yan, X. & Liu, L. 2003, 'Indexing by Conditional Association Semantics', Asian Journal of Information Technology, vol. 2, no. 2, pp. 50-55.
Yan, X. & Liu, L. 2003, 'A Database-Independent Approach of Mining Association Rules with Genetic Algorithm', Asian Journal of Information Technology, vol. 2, no. 2, pp. 45-49.