Spirit diagnosing is an important theory in TCM (Traditional Chinese Medicine), by which a TCM doctor can diagnose a patient’s body state. But this theory is complicated and difficult to master simply learned from books. To further the theory and skill of spirit diagnosing, in this paper, the authors propose a remote education system that can accept videos from a user and give the user an auto-diagnosed spirit. The key technology in this system is eye feature computation in spirit diagnosing, for which rules describing “the spirit” (spirit in TCM refers to the human’s mental state which reflects the one’s general physical condition) state are mined by the quantitative features regarding the human eyes. With videos capturing eye condition during a short period, a set of eye features are extracted. On this basis, attribute intervals of the eye feature space is generated by CAIM (class-attribute interdependence maximization). Several of the candidate rules are then mined by the association rule based on the cloud model. Finally, three complementary rule-pruning methods are modified and combined to trim the candidate rules. The cross validation test for mined rules has an average accuracy of 93%, which shows the high performance of the proposed method.
TopIn the recent 20 years, TCM diagnostic information extraction and processing are mainly centered on the pulse and tongue(Weiwu, Zheli, & Qun, 2005; Haixia, Yiqin, & Fufen, 2005). For the last five years, tongue information processing technology has a great development and its practically clinical application has been started. However, for such complex information as eyes’ movement which is also important in TCM diagnosis, the research on the extraction and Spirit diagnosing by eye features is still at a blank stage. There is no research report in this field till now.
The concept and model of association rules were first raised by Agrawal et al. (1993). Since then, there have been a series of association rule mining algorithms. Agrawal and Srikant (1994) proposed Apriori algorithm, which is used for mining association rules. Although the algorithm can generate all the association rules, it is an inefficient mining algorithm. As a result, some of the improved Apriori algorithms have been emerged. There are: Hash algorithm (Park, Chen, & Yu, 1995), Block technique Algorithm (Savasere, Omiecinski, & Navathe, 1995), Sampling algorithm (Toivonen, 1996), Dynamic Item set technique Algorithm (Brin, 1997), Incremental mining algorithm (Cheung, 1996), Parallel and Distributed mining algorithms, Associated with Database System Integration mining algorithms.