传统的MNL(Multinomial Logit)模型是在假设消费者是完全理性的前提下预测消费者的选择行为,但在实际生活中,消费者往往是有限理性的。在考虑套餐的多个属性及其参考点的前提下,借鉴参考依赖模型改进传统的MNL模型,预测有限理性的消费者选择套餐的行为。另外,在各个属性参考点的设定上分别采用了两种方法。在实例分析中,对改进前后的两个模型利用某电信运营商的商务消费者的实际数据,通过SPSS软件分别进行回归分析,所得结果进一步验证了改进新模型的有效性。 The traditional MNL (Multinomial Logit) model predicts the consumer’s choice behavior under the assumption that consumers are completely rational, but in real life, consumers tend to be bounded rationally. Considering the multiple attributes of packages and their reference points, the reference MNPs are refered to improve the traditional MNL model and predict the behavior of bounded rationality consumers choosing packages. In addition, two methods are used for setting the respective attribute reference points. In the case study, the two models before and after the improvement use the actual data of the business consumers of a telecom operator and conduct regression analysis respectively by SPSS software. The results further validate the effectiveness of the new model.