The Study Of Performance And Effectiveness Prediction Of Medical Equipment Based On Data Mining

Firstly, the related knowledge about data mining (DM) is discussed in this paper, and finally it is focused on the construction of data warehouse and the mining algorithm of decision tree and their application in performance and effectiveness prediction of medical equipment.Now in the medical organization, the most part of the income is accumulated for medical equipment, and about 80% of the available fund is invested in medical equipment every year. So, the performance and effectiveness prediction analysis of the medical equipment, especially the large scale medical equipment is an important project for the medical organization to serve the patients and keep stable and sustaining development. The performance and effectiveness prediction analysis for the considered equipment of the related department is a key step in the feasibility study of medical equipment purchase and also an important reference to the leaders' decision. Data mining technology now becomes one of the front subjects on the fields of database and information decision. It is widely focused by the scholars and widely applied in commerce, industry and medicine with remarkable social and economic benefits. Thus it is very important to extend the social and economic benefit of the medical equipment through the deep study of performance and effectiveness prediction for the medical equipment with the application of the theory and methods of DM.In the latest years, with the rapid development of computer and network technology, Hospital Information System (HIS) is widely applied in the hospitals. Although the patients' information management module and the medical equipment module are included in HIS, both of them are separated from each other, and the related software module is limited to record and revise the patients' information and medical equipment information and run query and statistics separately. They are still on the level of supporting the transaction processing based on database technology. The decision support of the feasibility study before the purchase of the medical equipment and the deep processing of the equipment information of operation and service after purchase are lacked. It is now focused that how to present right and reliable supporting to the manager and decision-maker of the hospital about the high effective management and scientific decision with these valuable information resource to improve the social benefit and economic benefit and the rapid and healthy development of the hospital.The term of Data Mining (DM) with another name of Knowledge Discovery in Database (KDD) was present in 1989. It is an uncommon process that a lot of effective, novel, potentially useful and finally understandable modes can be recognized from the database. The technology of DM has developed rapidly since 1990s, and it has inherited the achievement of knowledge discovery in theory and technology, at the same time it has imbibed many theory and algorithm from other fields such as database system, machine learning, pattern recognition, artificial intelligence, data visualization, information retrieval and statistics. In essential, DM is a kind of data analysis. Data analysis has a history of many years, but limited by the computing capability, the application of data analysis to mass and difficult data analysis is restricted. In the latest years, with the development and prevalence of electronic information technology, a lot of commercial or other data is produced. It will be very important to analyze the data and provide valuable information for commercial decision to gain more profit.The result of DM is usually described as conception, rule, law, mode, restriction, visualization etc. The knowledge returned by DM can be applied directly to assist decision or revise the existent knowledge system; it also can be saved to the application system as new knowledge. The original data of DM may be structurized data or half structurized data such as text, graph or image, even be heterogeneous data distributing on the network. The discovery knowledge may be used for information management, query optimization, decision supporting and process control as well as data self maintenance. As a cross subject, DM extends the data application from low level application such as simple statistics and query to high level as mining knowledge and providing decision support.For DM, the nice data management and pure data are needed and the data quality will affect the effect of DM directly, while the feature of data warehouse is right coincident with the need. The data from all kinds of data source is extracted and washed, integrated, chosen and transformed, then high quality data is provided. Data warehouse is the high stage of database. It is the data set which faces with main theme, integration, stability and change with time and can be applied in supporting the custom process of management decision. Data warehouse system can support the integrations of many kinds of application systems and databases and provide the robust platform for united history data analysis. Its main objective is to provide the support for decision and the platform for the deep data analysis such as OLAP, DM, etc. It is to say that DM can provide effective analysis and processing technology for data warehouse, and data warehouse can provide good base for DM.SQL Server 2005 Integration Services (SSIS) is a platform for creating high quality data integration project with the functions of extraction, transformation and load (ETL), serving as the data warehouse. SSIS presents the environment of workflow for constructing the package of data transformation, extracting the data from different data source and operating with the data.Medical database is a very complex database. Now the technology of DM is applied only in the relational database with structured data, transaction database and data warehouse, and is just the first step of DM application for complex database. If the particularity and complexity of the medical information are considered and the key technology of the mining process is managed, the technology of medical data mining will be extended in a wider field. Classification and prediction are two data analysis modalities of DM, the general algorithm of which includes rough set theory, decision tree, artificial neural networks and evolutionary computation etc. Most of the algorithms are not specific solution for any specialty and do not exclude each others. Generally say, there isn't an optimal algorithm. All the algorithms should be tried before reasonable one is adopted. Some of the algorithms can be revised, extended and optimized properly in practice for classifying and predicting in all kinds of special medical databases. In this paper, the theory and methods of decision tree algorithm are studied.The basic theory of decision tree algorithm is to split the data into subsets recursively, each subset has the similar states of objective variables and all the objective variables are predictable. At every split of the tree, the affection of all input attributes to the predictable attributes will be evaluated. While the recursiveness is finished, the creation of decision tree will succeed. The famous decision tree algorithms include ID3, ID4, ID5R and C4.5.Nowadays because one algorithm cannot be competent for any different kinds of data mining, the advanced data mining tools may carry many kinds of data mining algorithms. Here the mining task comprises two parts: classification and regression.Classification: To classify the medical equipment usage information under different condition. Suppose the relationship between the number of illness and the equipment usage times be linearity, let y = ax + b , here x is the number of illness, yis the equipment usage times, so the equipment usage can be predicted according the formula. Simply say, the main objective of regression algorithm is how to deduce the formula.At the same time it is supposed that different equipment and different kinds of illness have different linear relations (or functions), then the different relations should be classified. For example, the relationship between 'CT examination' and 'pneumonia' can be described as y = a₁x + b₁, and the relationship between 'CT examination' and 'lung cancer' is y = a₂x + b₂.For this mining, decision tree algorithm may better than other data mining algorithm. For example, a mining model can be defined quickly and explained easily with decision tree algorithm. Every path from root node to leaf node is called one rule, and the prediction based on decision tree is high quality. In this paper, the main objective is to discuss the prediction of the performance and effectiveness of medical equipment, so, it may be a better choice to run mining with decision tree algorithm.In SQL server 2005, a solution for regression is attached to Microsoft decision tree algorithm. It needn't to split data with Microsoft decision tree algorithm and the regressive formula is based on all data set. Microsoft decision tree algorithm of SQL Server Analysis Services 2005 is a composite algorithm which supports classification and regression. Every leaf node of Microsoft regressive tree has a linear regressive formula and there is at least one regressor in a regressive model. Regressor is a successive input attribute which can modeling the successive predictable attribute with linear model. For example, if the number of usage of ultrasonic device is a successive input attribute, the illness number is a regreesor. The classic linear regressive formula is number of usage of ultrasonic device = a + b * illness number + e, where e is noise with 0 mean value, and coefficient a (intercept) and b (slope) is defined by the summation of err equation, both of the coefficients should be small as possible.In this paper, the theory and methods of data mining are discussed and applied in the study of performance and effectiveness prediction of medical equipment. Compared with the data presented by our data mining model, the real data of seven medical equipments in our hospital as CT, MRI, CR, ultrasonic, Color Doppler Flow Imaging (CDFI), electronic gastroscope, electronic intestinal endoscope is coincident. With the study, the scientific and exact prediction of the medical equipment before purchasing is realized and factual reference is provided for clinic path study, and decision support of whole life of the medical equipment from feasibility study to being unusable is presented for management of the leaders. Through our study, the difficulty of the mutual restraint among performance and effectiveness, cost and risk of the medical equipment can be solved and the social and economic benefit of the medical equipment is fully exerted.