An Improved Approach for Generation of a Basic Probability Assignment in the Evidence Theory Based on Gaussian Distribution

Dempster–Shafer evidence theory (D-S theory) is a commonly used reasoning method for uncertain information. Generating the basic probability assignment (BPA) functions is the first step of applying D-S theory in practical engineering. The quality of generated BPA will have a direct impact on the result of evidence fusion and final decision-making. However, the generation of BPA still owns no fixed model. In response to this situation, a new BPA generation method based on the Gaussian distribution is proposed in this paper. First, the Gaussian distribution is constructed based on the mean and variance value of training sample in the data set. Second, calculate the function value of the test sample on the Gaussian distribution to generate BPA function. Third, data fusion based on Dempster’s combination rule. Finally, decision-making based on information fusion. The feasibility and effectiveness of the proposed method are verified in classification problem by using the UCI data sets.