One Method To Detect Abnormal Prescription Based On Density

With the fast development of the information technique, many hospitals have been using software to record information which may contain messages originated from patients, medicine, medical cost, and so on. The quantity of records one large-scale hospital generates one moth could be millions.Try to think about it, when getting and analyzing these records, we could discover many interesting knowledge so that we could sufficiently find abnormal prescription. We have enough ways to find the problems ensconced in prescriptions and knows the relationship between medicines.In many medical institutions of China, some people may have encountered one situation that you have to pay more money to make you cured. The traditional way to detect abnormal prescription is to check information such as medical qualication, unefficient and costly. In order to detect abnormal prescriptions efficiently, I use the algorithm of combination of many data-mining techniques applied to the prescription detection.The detailed work is as follows.Step 1:Modify the algorithm of LOF. and propose one way to prune data based on DBSCAN algorithm, put forward one way to set weight of attribute based on prescriptions.Step 2:Pre-process data of prescriptions and solve the problem of high dimension and sparsity.Step 3:To test the difference between algorithm LOF and the proposed algorithm, we use the dateset KDD’99 and prescription from local hospital to do this experiment, and to detect the outlier data. The result of this experiment proves that the time complexity of our algorithm is better than LOF algorithm and the accuracy slightly better than LOF algorithm.