【正文】 id casebased reasoning and neural work , have also been developed. Here, a data mining technique that integrates case based reasoning, neural work and rulebased reasoning is defined. These two are incorporated into the framework of the CBR cycle. Instead of using the nearest neighbor technique of traditional CBR systems, a neural work is used for indexing and retrieval of most appropriate service records based on 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。 then improve the design and reliability of those machine models in order to increase sales. Target the customers with mail campaigns of the machine models in which they are likely to be interested. _ Customer support: provide the best possible service to customers based on the machine model, the nature of the problem, and geographical location. _ Resource management: assign duties to service engineers based on their expertise and past experience. Promote service engineers based on their performance. 、 Selection of data This step identifies a subset of variables or data samples, on which mining can be performed. There are many tables in the database. However, not all are suitable for mining, since they are not sufficiently large. After an initial study, the structured data tables EMPLOYEE and CUSTOMER were found unsuitable for mining, while MACHINE and SERVICE_REPORT were considered suitable for mining. 、 Data preprocessing This step removes the noisy, erroneous, and inplete data. The presence of too many different categories of categorical data makes visualization of the displayed information very difficult. Hence, those categories with only a few records are eliminated. Moreover, all the records with missing values are deleted to avoid problems in visualization. Since the proportion of such records is quite small, their deletion will have little effect on the results. 、 Data transformation The data stored in the various tables are in a specified format (defined during the construction of the database). Sometimes, it is useful to transform the data into a new format in order to mine additional information. For example, a new column svc_repair_time (service repair time) is created by calculating the difference, measured in number of days, between svc_start_dt and svc_end_dt in the SERVICE_REPORT table. This new attribute is useful in analyzing the performance of the service engineers. 、 Data warehousing Data warehousing is the process of visioning, planning, building, using, managing, maintaining and enhancing databases. The data suitable for mining are collected from the various tables of the customer service database and stored in DB Miner39。 others may require more than one. In this research, Database Miner was chosen. This system was developed by the Database Miner Research Group from the Intelligent Database Systems Research Laboratory at Simon Fraser University in Canada. The system, which integrates data warehousing, online analytical processing (OLAP) and data mining techniques, supports the discovery of various kinds of knowledge at multiple conceptual levels from large relational databases. The Database Miner system supports most of the major functions. It was implemented using many advanced data mining techniques. In addition, it provides multidimensional data visualization support and interacts with standard data sources through open database connectivity (ODBC) interface. 、 Mining unstructured data Although Database Miner is an excellent data mining tool for large databases with structured data, it is unsuitable for extracting knowledge from the textual data of the customer service database. As the information or knowledge on mon faults and their suggested remedies are stored in textual format as fault conditions and checkpoints, new techniques are needed to extract knowledge from this database for machine fault diagnosis. This is known as text mining. Traditionally, casebased reasoning (CBR) has been successfully applied to fault diagnosis for customer service support . CBR systems rely on building a large repository of diagnostic cases (or past service reports) in order to circumvent the difficult task of extracting and encoding expert domain knowledge . It is one of the most appropriate techniques for machine fault diagnosis, as it learns by experience gained in solving problems and hence emulates humanlike intelligence. However, the performance of CBR systems critically depends on the adequacy as well as the anization of