Prediction Of Medical Imaging Examination Based On Data Mining Technology

To provide interesting information or knowledge to hospital administrators, data miningtechnology which is a new kind of data analysis technology to reveal data relationship is applied toanalyze the data of medical imaging information system. The prediction is one of the key content indata mining. Based on the understanding of the data mining technology background, the datamining application in the medical field and the current typical prediction algorithm, this paperanalyzed the existing problems in the amount of medical imaging examination prediction by themost representative prediction algorithm. Here summarized the main research work as follows:(1)Optimization and improvement of the traditional GM (1,1) model parameters. Thetraditional GM (1,1) model has the boundary value problem and the least squares parameterestimation problem, which are one of the causes leading the degree of fit and predictive accuracy ofthe traditional GM (1,1) prediction model to be inadequate. Under consideration those problems, anew improved model of GM (1,1) based on quantum genetic algorithm is put forward. Through thecomparison of prediction result, it affirmed that the improved GM(1,1) model improves theprediction accuracy.(2)The study on the association rule application in the medical imaging examinationinformation. This paper used Apriori algorithm to study and analyze the medical imaging examrecords of a large hospital from2005January to2012December, and found potential factors thataffected the amount of medical imaging examination. The study result showed: patients elder than70years old, inpatients type, patients with respiratory disease and patients suffering fromneurological diseases are the main factors affected the amount of CT examination.(3)A new prediction model of the amount of medical imaging examination prediction namedGray vector machine prediction model is presented. The model combines the advantages of the greyprediction model which only needs a small amount of data and support vector regression which canbe characteristics of nonlinear mapping. An example proves its feasibility in the amount of medicalimaging examination prediction, and has higher accuracy than the support vector machine or GM (1,N) model prediction.Medical imaging information system contains a large amount of hidden information andknowledge. This paper studied the amount of medical imaging examination prediction based on thedata mining techniques in order to assist the hospital scientific decision making.