Discovery of context-based knowledge patterns

Recall that a generic problem-solving process is specified by experts to solve a problem.

The specification includes the stages and their execution order. This work focuses on the execution of a sequence of stages. For example, the generic water supply problem-solving process is “Normal Management Stage (NM Stage) → Engineering Improvement Stage (EI Stage) → Exception Management Stage (EM Stage) → Quality Improvement Stage (QI

Stage) → Maintenance Management Stage (MM Stage). For any given problem, the situa-tions may vary; thus the follow-up acsitua-tions may also vary.

Discovery of context-based decision-making knowledge patterns. The decision-making knowledge patterns discovered from previous system framework indicate the frequent as-sociation of situations and actions in a problem-solving process. Different from deci-sion-making knowledge patterns, context-based decideci-sion-making knowledge patterns indi-cate the inferred associations of actions and situation features and actions in certain context of the problem-solving process. In specific stage, based on situation features and relevant context characteristics, the system continually infers situation features to form context-based decision-making knowledge pattern in certain context of specific problem-solving stage, as described in Section 5.2. Fig. 12 illustrates an example of the context-based decision-making knowledge pattern.

Fig. 12: An example of context-based decision-making knowledge pattern

The collected attributes: Air Condition (normal), Parameter (Incorrect), Power Status (12V), Doc(SOP) and inferred situation feature: Staff (Annie), Expert (PTC), Role (DG), System(DI), Power supply (normal) are used to discover the context-based inference rules that indicate taking modify action with CF value: 0.8 and taking reporting action with CF value: 0.6. The discovered inference rule forms the rule pattern considered as the con-text-based decision-making knowledge pattern.

Deriving the score of action. To recommend reasonable actions for specific situation in certain context, we use the weighted linear combination of acquired context-based knowl-edge, including the similarity of situation cases, the confidence of context-based deci-sion-making pattern, and the CF value of inferred action. Equation 18 defines the scoring method of context-based decision-making knowledge pattern of specific situation-action.

inf j

whose left-hand side match the current situation are suggested and ranked according to the scoring values of the context-based decision-making knowledge patterns.

)

where represents the action j; indicates the situation i; indicates the target situation t;

the represents the similarity value of situation and ; Conf( )indi-cates the confidence of decision-making knowledge pattern ; CF( ) indicates the certainty factor value of inferred action derived from context-based inference rules; w

Discovery of context-based dependency knowledge patterns. For a specific prob-lem-solving process, the dependency knowledge patterns discovered from previous system framework express the frequent relationships between situations and actions across different stages. Different from dependency knowledge pattern, context-based dependency knowl-edge patterns indicate the inferred relationships between situation/action features in current stage and situations/actions across different stages of the whole problem-solving process context. The discovered context-based inference rule may involve with several stages. The system uses its situation features as the seeds to infer situation features in relevant stages.

The inferred situation features of relevant stages form the context-based dependency knowledge pattern. Fig. 13 illustrates the example of the context-based dependency knowledge patterns.

Fig. 13: An example of context-based dependency knowledge pattern

In pipe abnormal situation of Normal Management stage, the collected situation fea-tures (e.g., Role: DG; System: DI; Water supply pressure: High; Power supply service:

normal) are used to infer the intra situation features (e.g., Staff: Annie; Pipe system pa-rameter: Incorrect) based on relevant context-based inference rule (e.g., [Role(DG)] → [Staff(Annie)] and [pipe pressure(High)] → [System parameter(Incorrect)]). The collected and inferred situation features are used to infer the inter situation features in different stages (e.g., Role: CR, System: DI in Engineering Improvement stage and Rule: Tuning, Service Testing: Correct in Exception Management stage). The inferred situation features are used to continually infer intra and inter situation features in current and different stages. The in-ferred situation features and relevant context-based inference rules form the context-based dependency knowledge patterns.

Context-based situation profiles and relevant documents. The context-based situation profiles are generated from the accessed documents in certain context, as described in sec-tion 5.2.2. For example, in the situasec-tion of abnormal water quantity, the accessed documents include: “DI analytical machine water quantity recording” and “DI GCHC machine water quantity recycling.” The relevant context information includes Location: 8C; System: DI;

System: DI; Service: Water Supply Service. The context-based situation profile is generated from the accessed documents in certain context. Once a worker encounters a problem situation or decides to take a particular action, the system provides relevant documents as knowledge support based on the context-based situation profiles. Fig. 14 illustrates a con-text-based situation profile and the relevant documents for the water supply problem-solving

process.

Fig. 14: Relevant context-based situation profile of situation EI_S4

Based on the context-based situation profiles, the system gathers previous and new relevant documents, such as “DI analytical machine water quantity recording” and “DI GCHC machine water quantity recycling” and new documents “DI system waste water quantity evaluation” and “DI UF Flush water quantity recycling”.

5.2.4. Knowledge recommendation

The proposed system suggests relevant documents according to the context-based situation profile of the current situation or similar cases, as shown in Fig. 14. The system also recommends relevant action documents (e.g., operating procedures and guidelines) according to the action profile, as shown in Fig. 15. Note that the top-N relevant documents are recommended according to the cosine measure of the term vectors of the documents and the context-based situation profiles, as described in Section 5.2.2.

Fig. 15: Context-based situation profiles and relevant documents of the pattern Moreover, the system suggests possible actions for handling the current situation ac-cording to the context-based decision-making knowledge patterns. When a situation matches a specific context-based knowledge pattern, the inferred situation features and relevant context-based inference rules in certain stages will be suggested as knowledge support. Furthermore, the context-based dependency knowledge patterns also denote the chain reaction across different stages. This helps workers plan appropriate actions for dif-ferent problem-solving stages.

As the example of Fig. 16, the context-based knowledge patterns are in a chain reaction across different stages. In normal management stage, based on situation features collected by system, inferred situation feature, and scoring mechanism described in Section 5.2.3, the context-based decision-making knowledge patterns suggest the workers that pipe abnormal situation → Monitoring the output action with score 0.006 and pipe abnormal situation → Reporting the outcome action with score 0.0021 under the context consideration. The con-text-based dependency knowledge patterns also provide workers inferred situation features and relevant context-based inference rules as knowledge support to plan appropriate actions for different problem-solving stages

Fig. 16: Context-based knowledge patterns in a chain reaction across different stages

5.3. System implementation

In this section, we illustrated a system implementation to demonstrate the effectiveness of context-based rule inference. The implementation is conducted using several software tools, including the Eclipse Version 3.2 Software Development Kit (SDK) and Java(TM) 2 Platform Standard Edition Runtime Environment (J2SE) Version 5.0. The system function uses Drools of JBoss Rules Version 3.04 which is the plugin of Eclipse as the inference engine to get Certainty Factor (CF) value and infer situation feature. Microsoft SQL Server 2000 is used as the database system for storing data related to the problem-solving process and codified knowledge documents. The data mining tool Weka 3.4 is used to discover context-based inference rules in the historical problem-solving log.

The system function shows relevant problem-solving information collected in knowl-edge base, including problem-solving process, stage, situation/action, context-based infer-ence rule with confidinfer-ence and support value. Based on collected problem-solving informa-tion, the system function enforces the inference process and shows the inferred knowledge.

For example, the system function gathers the relevant problem-solving information, including current problem-solving process: Water Supply Problem-solving Process; current stage: Normal Management Stage; current situation: Controller Temperature Abnormal Situation; and context-based inference rule: Staff(Annie) → DI Water Supply System Ser-vice (Parameter: incorrect) with confidence 0.13 and support 0.02. The system interface and relevant problem-solving information are illustrated in Fig. 17.

Fig. 17: A prototype system of context-based problem-solving knowledge support After the inference process, the CF value of situation features are provided to worker, as shown in Fig. 18. According to the relevant information stored in enterprise knowledge base, the CF value of Staff(Annie) is “1”, support of “DI Water Supply System Ser-vice(Parameter: incorrect)” is 0.1, the system enforces inference process (the details men-tioned in Section 5.2.1) and gets the CF value of “DI Water Supply System Ser-vice(Parameter: incorrect)” is 0.033. The system function assists worker get the CF values of context-based inference rules and situation features in order to infer more situation fea-tures continually.

Fig. 18: Inference knowledge for the water supply problem

5.4. Discussions and comparisons

In this section, we discuss and compare the proposal knowledge support framework and context-based knowledge support.

5.4.1. Discussions

This work proposes a knowledge support system for problem-solving on a produc-tion-line. The descriptions of situation/action, attributes collected by system, and inferred situation features assist case-based reasoning in situation identification. Information Re-trieval (Automatic Indexing) techniques are applied to discover the key terms of a situation.

The terms form relevant situation profiles that model the information needs of workers to handle a problem. Association rule mining and sequential pattern mining techniques are used to discover decision-making and dependency knowledge patterns, and context-based inference rules. The context-based inference rules are used to infer more relevant situation features. This system discovers context-based decision-making and dependency knowledge based on context-based inference rules and inferred situation features. The situation profiles, discovered knowledge patterns, context-based inference rules, inferred features, and

con-text-based situation profiles forms the basis to support problem-solving on a production line.

Some issues or shortcomings of this framework are discussed as follows.

(1) The knowledge support system discovers relevant knowledge rules in order to provide knowledge support in a problem-solving process. However, the items of knowledge rules may involve with various types of data. For example, an attribute value may be nominal, binary, or numeric; the numeric attributes, a data discretization process is conducted to transform their values into value ranges or user-defined concept terms (such as low, middle or high). Therefore, in rule processing, the rule matching is an important issue that needs to be addressed.

(2) Based on CBR, data mining, and rule inference techniques, the context-based knowl-edge support system enforces context modeling to formalize the relevant situation features and context characteristics of a problem-solving process. The situation features and context characteristics are considers as the items of transaction in order to discover context-based inference rules. However, situation features and context characteristics in different context may have different importance. Therefore, the importance of various situation features and context characteristics should be considered in different levels of context modeling.

5.4.2. Comparisons

Comparison to related work. We compare the proposed knowledge support system with related work, the details are illustrated as follows.

(1) Liao (2002) investigates the types of knowledge used for problem-solving and suggests the circulation of knowledge to avoid knowledge inertia. Although a knowledge-based architecture that incorporates case-based, rule-based, and heuristic-based approaches is proposed for managing problem-solving knowledge and dealing with knowledge iner-tia, the details of the system are not presented. In this work, the proposed system framework presents the procedures of knowledge discovery and recommendation processes. Moreover, the details of system implementation and real scenario are also illustrated clearly.

(2) Existing studies focus on using case-based reasoning to identify similar previous cases and derive a solution for a new case from previous problem solutions (Chang et al.,

process, problem-solving is usually knowledge intensive and requires effective knowledge support to provide workers with the necessary information to identify the causes of situations and taking appropriate action to solve them. However, identifying similar cases among previous problem cases is not sufficient to satisfy workers’ in-formation needs for solving a new problem. The required knowledge is usually hidden in various codified knowledge documents that must be proactively delivered to workers.

The CBR approach does not provide such problem-relevant documents for knowl-edge-intensive problem solving. In this work, we adopt text mining (Automatic In-dexing) techniques to compensate for the shortcomings of CBR technique.

(3) Problem-solving is the thought process that resolves various difficulties and obstacles spread in the gap between the current problem and its desired solution. (Heh, 1999).

Problem-solving process includes a series of uncertain situations and operational ac-tions. Moreover, situation features are usually occurred according to the context characteristics of problem. Due to the uncertain features of situations, several causes and possible solutions may exist for a specific situation. The causes and possible solu-tions are usually hidden in relevant data resources and difficult to extract. In such un-certain environments, situation features collected by system are usually partial or in-complete. Workers need to use knowledge gathered and inferred from relevant context information and previous problem-solving experience to clarify the causes and take appropriate action effectively. This work adopts rule inference techniques to consoli-date the knowledge support for problem-solving.

Comparison of two kinds of knowledge supports. The knowledge supports for prob-lem-solving are separated into two parts. One uses the CBR and data mining techniques to provide knowledge support for problem-solving, which does not consider context charac-teristics; the other uses the CBR, data mining, and rule inference technique to provide knowledge support for problem-solving considering context characteristics. There are some comparisons illustrated as follows.

(1) The problem-solving process is a complex process that involves with wide scope of enterprise information and knowledge. In the knowledge support framework without the context consideration, using CBR and data mining techniques to process user de-scriptions and collected attributes may not be enough to support the problem-solving process. Based on the consideration of context, the context-based knowledge support

enforces CBR, data mining, and rule inference techniques to discover and infer more relevant knowledge, thus can help worker identify the certain situation and obtain relevant knowledge support effectively.

(2) The setting of minimum support and minimum confidence criteria may filter out some non-frequent but important clues of problem-solving process. Without context con-sideration, the system may derive very few decision-making and dependency knowl-edge rule patterns. Accordingly, worker may not obtain relevant knowlknowl-edge support for certain situation/action. Based on collected context characteristics and inferred knowledge, the context-based knowledge support can infer more context-based knowledge to compensate the shortcoming of incomplete information and sparsity of rule patterns.

Chapter 6. Conclusions and Future works

6.1. Summary

In this work, a novel knowledge support system has developed for problem-solving on a production-line. The description of situation/action and collected attributes assist case-based reasoning to identify similar situations/actions. Information Retrieval (Automatic Indexing) techniques are then applied to discover the key terms of a situation/action. The terms form situation/action profiles that model the information needed of workers to handle a problem. Association rule mining and sequential pattern mining are used to discover de-cision-making and dependency knowledge patterns, respectively. The situation/action pro-files and discovered knowledge patterns are used to construct a knowledge support network, which forms the basis of support for solving problems on a production line. The proposed system provides integrated browsing and suggestions about problem-solving knowledge.

Relevant documents are recommended to help users identify the root cause of a problem situation and the appropriate action to take. Workers can also use the knowledge support network to navigate the knowledge patterns and obtain decision-making and dependency knowledge. The proposed knowledge support network, enhanced with suggestions about problem-solving knowledge, provides workers with the necessary knowledge to effectively solve problems. A prototype system is implemented using a data set from a company’s intranet portal, in which the log file contains a log of information for handling problems on the company’s production line.

Moreover, based on context modeling, context-based inference rules are discovered to infer more relevant situation features. The description of situation, collected attributes, and inferred situation features assist case-based reasoning in situation identification. Information Retrieval (Automatic Indexing) techniques are then applied to discover the key terms of a situation. The terms form context-based situation profiles that model the information needs of workers to handle a problem. The system uses the context-based situation profile to gather existing and new relevant knowledge documents for situation in certain context.

Furthermore, the system continually infers situation features to form context-based knowledge patterns which provide workers with relevant inferred knowledge, as well as decision-making and dependency knowledge.

6.2. Future works

In our future work, we will apply our proposed method to different data resources or other application domains. This work has focused on solving problems in stages in different situations with different actions. The stages need to be predefined by experts, which is the case with the company’s production line. For other application domains, the stages may not be easy to define. Moreover, the stages investigated in this work are limited to a sequential order, rather than a combination of AND/OR parallelisms and sequences, as in a workflow system. Accordingly, a more flexible approach to address these issues would be worthy of further study.

References

[1] Abecker, A., Bernardi, A., Maus, H., Sintek, M., & Wenzel, C. (2000). Information supply for business processes: coupling workflow with document analysis and infor-mation retrieval. Knowledge-Based Systems, 13(1), 271-284.

[2] Agostini, A., Albolino, S., De Michelis, G., De Paoli, F., & Dondi, R. (2003). Stimu-lating knowledge discovery and sharing. Proceedings of the International ACM Con-ference on Supporting Group Work, Sanibel Island, Florida, USA, November, 248-257.

[3] Agrawal, R., Imielinski, T., & Swami, A. N. (1993). Mining Association Rules between Sets of Items in Large Databases. Proceedings of the 1993 ACM SIGMOD Interna-tional Conference on Management of Data, Washington, D.C., USA, May, 207-216.

[4] Agrawal, R., & Srikant, R. (1994). Fast algorithms for mining association rules. Pro-ceedings of the VLDB conference, 407-419.

[5] Agrawal, R., & Srikant, R. (1995). Mining Sequential Patterns. Proceedings of the 11th International Conference on Data Engineering, Taipei, Taiwan, March, 3-14.

[6] Allen, J., Blaylock, N., & Ferguson, G. (2002). A Problem Solving Model for Col-laborative Agents. Proceedings of the first international joint conference on Autono-mous agents and multiagent systems, Bologna, Italy, July, 774-781.

[7] Baeza-Yates, R., & Ribeiro-Neto, B. (1999). Modern Information Retrieval, New York:

The ACM Press.

[8] Chakrabarti, S., Dom, B., Agrawal, R. & Raghavan, P. (1997). Using taxonomy, dis-criminants, and signatures for navigating in text databases. Proceedings of the 23rd VLDB conference Athens, Greece.

[9] Chang, K. H., Raman, P., Carlisle, W. H., & Cross, J. H. (1996). A Self-improving helpdesk service system using case-based reasoning techniques. Computer in Industry, 30, 113-125.

[10] Chen, G. & Kotz, D. (2000). A Survey of Context-Aware Mobile Computing Research, tech. report TR2000-381, Dept. of Computer Science, Dartmouth College, Hanover, N.H.

[11] Chen, M.S., Park J. S., & Yu, P. S. (1996). Data Mining: An Overview from Database Perspective, IEEE Transactions on Knowledge and Data Engineering, 8(6), December.

[12] Davenport, T.H., & Prusak, L. (1998). Working knowledge: How organizations manage what they know, Boston MA: Harvard Business School Press.

[13] Davies, N.J., Duke, A., & Stonkus, A. (2003). OntoShare: Evolving ontologies in a knowledge sharing system, In Davies, NJ, Fensel, D. van Harmelen, F. (Eds.): Towards

[13] Davies, N.J., Duke, A., & Stonkus, A. (2003). OntoShare: Evolving ontologies in a knowledge sharing system, In Davies, NJ, Fensel, D. van Harmelen, F. (Eds.): Towards