【導讀】Abstract. 1、Introduction. r

　　

【正文】 ing a bar chart. This summarization is very useful for understanding the expertise of service engineers. This information can be used to assign appropriate engineers to servicing particular machine models. From the figure, it can be seen that the machine model AVK_2020S has been serviced only by service engineer KL006. However, this also shows that service engineer KL006 has not worked on any other machine model. Another example is given in Fig. 6 to illustrate the use of association rules mining. The association rules determine how the various attributes are related. Here, there is a strong association among the attributes in a textual format. The rules have a minimum support of 8% and a minimum confidence of 98%. High confidence rules represent distinctive patterns within a database: the first two association rules state that the customer TAIT has reported all the faults during the year 1996 and it was serviced by the service engineer KL006 only. This shows that the service engineer KL006 would be the most suitable person to be assigned to serve TAIT in future. The next two rules show that the machine model AVK_2020S is serviced by the service engineer KL006 only and the faults were reported in the year 1996. In addition, Rule 8 indicates that the service engineer 10530 had resolved all the fault problems within a day. 、 Evaluating the mining results Different data mining functions have been exercised, providing data. The information obtained is next analyzed. The results are: marketing: OLAP analysis and summarization have been applied to identify machine models with poor sales and high frequency faults. Clustering is used to identify customers suitable for cross sales. Customer support: association rules, classification, and clustering are used to identify those who have recently reported many faults. Better quality of service can then be provided based on customer geographical location and machine model purchased. Fig. 8 shows the knowledge extraction process for retrieving information from the unstructured textual data of the faultconditions and checkpoints in the customer service database. There are two major generation steps: neural work model and rule base. The neural work model generation phase extracts the knowledge from the faultconditions to train the neural work to build neural work models for classification and clustering. The fault conditions in the customer service database are first preprocessed to extract keywords. The preprocessing uses a wordlist, stoplist, and algorithms from WorldNet. The extracted keywords are used to form weight vectors to initialize the neural works. Then, the neural works are trained to generate the neural work models. Two types of neural works were investigated: the supervised learning vector quantization neural work and the unsupervised Cohunes self anizing map neural work. Resource management: summarization can be used to identify the expertise of different service engineers. Association rules provide useful information on efficient engineers who can repair machine faults within a day. Prediction analysis can be used to pare different service engineers who have repaired machine faults under the same conditions. With this, the pany can assign job duties to service engineers based on their expertise and efficiency. Data mining for machine fault diagnosis The unstructured textual data of faultcondition and checkpoint information of the customer service database provides useful machine service information. A data mining technique based on the integration of neural work, casebased reasoning, and rulebased reasoning has been applied to the customer service database to support intelligent machine fault diagnosis. It shows the framework of the data mining process. It consists of two major processes: the offline knowledge extraction process and the online fault diagnosis process. The first extracts knowledge from the customer service database to form a knowledge base that contains the neural work models and a rulebase. The neural work models and the rule base work within the CBR cycle to support the second, which uses the four stages of CBR cycle (retrieve, reuse, revise, and retain) to diagnose customer reported problems. It accepts user39。s problem description as input, maps the description into the closest faultconditions of the faults previously stored from the knowledge base, and retrieves the corresponding checkpoint solutions for the user. The user39。s feedback on the fault diagnosis process is used to revise the problem and its solution. The new result is ultimately retained as knowledge for enhancing performance of future problems.