26th International Symposium on Automation and Robotics in Construction (ISARC 2009) 451 Mining knowledge management strategies from the performance data of cop Wen-der Yu 1 , Shih-ting Lin 2 , Shen-jung Liu 3 and Pei-lun Chang 4 1 Professor, Institute of Const. Mgmt., Chung Hua Univ., Taiwan, wenderyu@chu.edu.tw 2 Master Student, Institute of Const. Mgmt., Chung Hua Univ., Taiwan, m09616015@cc.chu.edu.tw 3 Assistant Vice President, CECI Engineering Consultants, Inc., Taiwan, sjliu@ceci.com.tw 4 Engineer, Depart. of Business & Research, CECI Engineering Consultants, Inc., Taiwan, peilun@ceci.com.tw Abstract Knowledge community of practice (CoP) is a popular approach for knowledge management implementation in construction organizations including contractors and A/E firms. In order to evaluate and improve the performance of the knowledge CoPs, quantification methods for performance measurement were proposed in previous researches. Profound implications may be inferred from the performance data recorded from daily knowledge management activities. Such implications provide directions of valuable strategies for administration schemes and system modifications. To achieve such goals, the performance improvement patterns and rules should be identified. In this paper MS SQL Server® was adopted to performed Data Ming (DM) tasks that dig out the abovementioned patterns and rules from the CoP performance data. Three DM techniques (Decision trees, Clusters, and Association Rules) were employed to mine the rules and patterns existing in the 4,892 historic performance data recorded from the CoPs of a leading A/E consulting firm in Taiwan. Performance improvement strategies are then inferred and planned based on the rules and patterns discovered. Keywords: Knowledge Management; Data Mining; Consulting firm; Performance Measurement, Strategy planning. 1. Introduction Knowledge Management System (KMS) is a popular approach for knowledge management implementation in construction organizations including contraction and A/E firms. A KMS does not only provide a platform for knowledge generation, storing, retrieval, and sharing, but also enable an organization a tool to measure and monitor its intellectual property. In order to evaluate and improve the performance of the KMS, quantification methods for KMS performance were proposed in several previous works [1][2][3]. From those works, it was found that profound implications may be inferred from the performance data recorded in daily knowledge management activities. Such implications may indicate valuable strategies for increasing benefits resulted from the KMS both in terms of administration and system modification schemes. The key to achieve such objectives is finding out the performance improvement knowledge. The Data Mining (DM) and Knowledge Discovery in Databases (KDD) are proven to be very effective in mining patterns and rules residing in large databases [4][5][6][7]. In this paper, a case study is conducted on mining knowledge of improvement strategy from the performance data of a generic CoP in a leading A/E consulting firm in Taiwan, the CECI Engineering Consultants, Inc. (CECI). The proposed methodology combines two major elements: (1) a quantitative model for measurement of the performance of CoP; and (2) commercial DM software—Microsoft SQL Server®— for performing DM tasks. Totally 4,892 historic performance data were collected from nine selected CoPs of the case A/E firm for case study. Questionnaire surveys were conducted with the participants of CoP knowledge management (KM) activities via a web-based internet questionnaire surveying system. The survey results are then converted into data in the form acceptable for DM by the Microsoft SQL Server®. Three DM techniques (Decision trees, Clusters, and Association Rules) are employed to mine the rules and patterns existing in the performance data. Meaningful rules, useful patterns, and important association rules are found with DM. Performance improvement strategies are then inferred and planned.