Research On Data Sequential Pattern Mining Based Hospital Material Consumption Prediction Model

Medical materials are the base of hospitals for carring out medical activities.Related research shows that the procurement cost of medical materials accounts for40%-50% of the total cost of the hospital.Therefore,reducing the purchase cost of medical consumables is particularly important in the cost control of the entire hospital.Under the current background of new medical reform,hospitals must control procurement costs if they want to improve their competitiveness in the fierce competition of medical services.At present,the purchase of materials in the hospital mainly relies on the experience of the purchasing personnel to make the purchase plan,which has the problems of overpurchase,stock backlog,and even the expired material.These are serious occupying and even wasting hospital funds and so on.On the other hand,the development of information technology has greatly promoted the informatization process of medical institutions.The implementation of regional health information interconnection and interoperability policy has solved the current situation of scattered storage of large amounts of data in various medical units in the region and interconnected with each other,providing a data base for the application of large data technology.Based on this,this paper proposes a precision prediction scheme for the use of regional medical materials based on distributed data mining,which provides support for medical institutions to formulate procurement plans and control procurement costs.The main contents and results of this paper are as follows:1)Based on the distributed system,this paper constructed a precise prediction scheme for the use of hospital materials and built a prediction model of the hospital material usage based on Hadoop to ensure the prediction of the purchasing data in the large data environment.2)The identification of the parameters in the traditional ARIMA model depends on the trailing and truncating properties of the artificial recognition correlation function to estimate the model parameters.This method is easy to fall into the local optimum,and is not suitable for modeling a wide range of different materials in the hospital.In this paper,an improved model parameter identification method based on the akaike information criterion and bayesian information criterion is proposed.At the same time,the local optimal problem is solved.Finally,based on this method,a prediction model of material usage based on the improved ARIMA model is proposed.3)This paper proposed and implemented a forecast model of material consumption based on LSTM neural network prediction model.This paper studied the improvement method of LSTM network for the deficiency of RNN network,summarized the training process and training points of LSTM network model,and compared and evaluated the prediction effect of ARIMA model and LSTM neural network model in the prediction of hospital material usage.