Research And Application Of Clinical Behavior Anomaly Detection Based On Association Rules

Social medical insurance system as the national welfare and the people's livelihood system have been played major roles in the aspect of ensuring workers healthiness and preventing from disease. As the lack of effective medical procedures and standardized of behavior, irregularities in the process of medical diagnosis are frequent, these are not only harmful to people's health, causing tension between doctors and patients, but also resulting in waste of medical resources and hindered the development of medical and health undertakings. Therefore, discovering unreasonable clinical phenomenon from the clinical diagnosis and treatments can be preventing irregularities and frauds, and can regulate behavior of clinical treatment and reduce the loss of medical resources. In short, this work has a great significance.Combination of research background and research at domestic and abroad, and based on the concept of "Clinical Path", Articles using association rule mining method for mining legitimate pattern of behavior from clinical data, and then building anomaly detection model, this model applied to find abnormal behavior in clinical diagnosis and treatment.The main works of the dissertation are listed as follows:1. Paper discusses the main methods of anomaly detection, and analyzed the association rules in medical clinical field. Research of the characteristics of medical behavior, and then present a method to preprocessing clinical data.2. The data for the feature of the time constraint of clinical behavior and the characteristics of single-disease treatment behavior based on clinical path. By the defects of existing algorithms, and then improving the GSP algorithm, present a CBS_GSPA algorithm to find frequent sequences. Described the timing of subsequence by the introduction of the concept of legal subsequences, whether the subsequences is legal judged by the time constrained, the correct candidate items areassured to becontinually generated and the legal subsequence support which conforms to clinical behavior standard is calculated out, achieving pruning the sequence patterns. and then find the frequent patterns of behavior sequence in clinical data.3. Due to the time constraints for the clinical behavior, we analyzing the problem of the legitimacy by generated rules from frequent sequences, and proposed ARCBS (Association Rule of Clinical Behavior Sequence) algorithm to mining the association rules of clinical behavior sequence, as the basis for building the model of clinical behavior anomaly detection.4. Design of the prototype system frame structure. The section of extraction the association rules, anomaly detection and others are implementation. By evaluating the effectiveness of anomaly detection to verifying the model to present their legitimacy and usability...