关键词: 三支决策, 最优阈值, 网格搜索, 风险损失, 数据集

Abstract: The traditional three-decision-making model relies on the experience of experts to set the threshold, thus impeding the wide application of three-decision-making model in many fields. To minimize the decision-making risk, a computational model of the risk-loss was built, and a new classification algorithm which needs no priori knowledge was given. The algorithm used model conditions to determine the range of parameters value which minimized the risk-loss, then divided the range into several equal grids, got the smallest range of parameters through searching these grids, and the smallest range was the optimal threshold. At last, a three-decision-making classifier was built by using the threshold, and then this classifier and Bayesian classifier were applied to part of UCI data sets. The comparison shows that the performance of three-decision-making classifier is superior, which shows the effectiveness of the algorithm.

Key words: three-decision-making, optimal threshold, gid searching, risk-loss, data set