A Research Of Patient Flow Anomaly Detection Method Based On Data Mining

The current status of difficult and high cost to access medical service due to unbalanced distribution of high quality medical resources.China makes efforts to promote the hierarchical medical system.Patients flow monitoring is the key of the hierarchical medical supervision.But there are many disadvantages,such as low efficiency,lack of pertinence and so on,in the supervision that relies on traditional statistical indicators.This thesis presents a view of by using the patient data and data mining technology to assist in the supervision.The following problems are studied in depth:(1)Data cleaning and feature selection.This thesis firstly analyzes the population data and the patient’s medical record,and then uses the data preprocessing technology for data cleaning,conversion and feature selection to solve the problem that the original data set exists such as low degree of integration,poor availability and so on.(2)How to find anomalous cross-hospital level flow.This thesis proposes a new anomaly detection method based on the characteristics of the current anomaly data and optimizes traditional clustering algorithm.First,the thesis is based on conceptual clustering and completes the characteristic description of the normal flow,and then identifies anomaly flow by comparing the differences between the normal flow and cross-hospital level flow.(3)How to find anomalous referral flow.First of all,this thesis chooses a common chronic disease,through the BP neural network to build the hospital days prediction model,and then identifies the patient flow due to the lack of timely referral led to the actual hospital days far more than normal hospital days.Based on data mining anomaly detection method can timely,comprehensively and effectively assess whether the flow of patients is reasonable.This method not only overcomes the disadvantages of current monitoring method,but also can more easier to find the cause of abnormalities with the help of multi-dimensional information,and then provides government with some decision support to reverse the unreasonable flow.