Abstract: To resolve the problem of the limitation of low accuracy and reliability in Case-based Reasoning(CBR), Gaussian Case-based Reasoning(GCBR)is proposed. In this research, based on the research on similarity measure on the basis of Euclidean metric, a new means of dealing with similarity measure by employing Gaussian transformations in place of the Euclidean metric is proposed. The concept of distance proportion is introduced to make a transformation that can transfer distances between a pair of cases on each feature into Gaussian indicators, and similarity between two cases will be derived. On that basis, the nearest-neighbor is applied to retrieve case similarity, and the predictive value will be generated by modifying cases. Taking the prediction for the risks of logistics outsourcing enterprise for example, compared with classical CBR model on the basis of Euclidean metric and the Grey CBR, GCBR model statistically and significantly performs better on predictive accuracy, which illustrates the feasibility of applying GCBR model to predict the risks of logistics outsourcing enterprise.

Key words: artificial intelligence, gaussian case-based reasoning, similarity, gaussian index, prediction model

摘要： 针对目前案例推理方法中存在的预测准确性低,可靠性差等方面的局限,提出了高斯—案例推理方法。在传统基于欧氏距离相似度计算研究的基础上提出了一种新的相似度计算方法,即运用高斯转换代替欧式距离度量来计算相似度的方法。引入距离比例的概念,将案例间的特征转化为高斯指标,进而计算案例间的相似度。在此基础上使用最近邻法检索出相似案例,通过对检索出的案例进行修改获得预测值。以物流外包企业风险预测为例,通过与两个案例推理模型(基于欧式距离的经典案例推理模型和灰色理论案例推理模型)的预测结果比较,得出在预测的准确性上,高斯案例推理预测模型优于其它两种模型。验证了利用高斯—案例推理模型预测企业物流外包风险的可行性。

关键词: 人工智能, 高斯案例推理, 相似度, 高斯指标, 预测模型