The Stochastic Prediction DEA Models With Undesirable Outputs And Its Application

Data Envelopment Analysis(DEA) depends on mathematical programming to evaluate the relative efficiency of DMUs with multiple inputs and multiple outputs. It's an intersectional field of operation research, management science and mathematical ecnomics. It has developed rapidly since established by A. Charnes and others in 1978, not only in theory field but also in application. By now, the study on stochastic DEA becomes the foreland of DEA.Extended to the prediction field, DEA method can solve the prediction problem with multiple inputs and outputs which can not be solved easily by the Regression Analysis method. But the traditional DEA models can not solve the problem with undesirable outputs. Based on the theory of multiple goal programming, this dissertation will explore the inherent relationship between multiple goal programming and DEA method, to solve the prediction problem.In pratical situations the data of inputs and the outputs are stochastic. Therefore this dissertation is to extend the existing DEA prediction model into stochastic field and construct stochastic prediction DEA models.Finally, this dissertation applys the new stochastic prediction model into the process of HIV immunology treatment analysis. It's used to measure the efficiency of drug treatments, and predict the trend of the treatments for every phase. Stochastic errors and side effects are considered in this model, which is an advantage over other deterministic efficiency models.