Conjoint analysis, as one of main analytical tools in helping the marketing decision, has been used in marketing since 1970s due to methodological easiness and the variety of applications. The main reason why conjoint analysis has broadly been used in marketing field is in the high predictability of consumer choice behavior. Existing studies have mainly focused on the comparison of alternative conjoint models such as conventional, choice-based and hybrid conjoint model in terms of the relative predictability of consumer choice behavior or market share. But few studies consider consumer knowledge and the level of product involvement on the choice predictability of alternative conjoint models. The purpose of this study is to analyze how consumer characteristics influence the product choice prediction of alternative conjoint models. Specifically this study investigates the differential effect of consumer characteristics such as consumer knowledge and product involvement on the choice prediction of conventional and choice-based conjoint model. We choose the mobile phone for the high involvement product and instant noodle for the low involvement product. In order to take into consideration the level of information load, in the case of the high level of information load seven product attributes are given to the respondents and in the case of low information load, four product attributes are given. The product involvement of consumers is measured based on the scales developed by Zinkhan and Locander. And consumer knowledge is measured based on the scales developed by Brucks. The empirical data is collected by on-line survey research, where one thousand respondents answer the two set of conjoint questionnaire for mobile phone and instant noodle respectively. The level of consumer knowledge is found to give the positive impact on predictive power of conjoint model only when information load is not high. High information load leads to lowered prediction of consumer choice behavior regardless of the level of product involvement and the level of consumer knowledge. And the higher the level of product involvement, the better the predictive power of conjoint model for brand choice behavior. On the other hand, if the level of product involvement is low, the predictability of conjoint models significantly decreases compared to high level of product involvement. In the case of the low level of product involvement, predictive power of choice-based conjoint model and conventional conjoint model is different. When the level of product involvement is low, predictive power of choice-based conjoint model is higher than that of conventional conjoint model. This study shows that the high level of consumer knowledge and product involvement increase the choice prediction of conjoint model within the limitation of moderate or low level of information load. That is, the excessive information load leads to lowed choice prediction of both choice-based conjoint model and conventional conjoint model even when the level of product involvement or the consumer knowledge is high. The strategic implication of this study is that marketers should take into consideration the psychological characteristics such as the consumer knowledge and product involvement in choosing the appropriate conjoint model among many alternatives. It is recommended to utilize conventional conjoint model at the later stage of product life cycle, when consumer knowledge is sufficiently accumulated and at the introduction stage of new product, choice-based conjoint model would be more effective in predicting consumer choice behavior than conventional conjoint model. When the level of product involvement is high, both choice-based conjoint model and conventional conjoint model give the good prediction of brand choice. In the case of low level of product involvement, choice-based conjoint model is better in predicting the brand choice. The future study can investigate the differential effect of consumer characteristics on the choice predictability of alternative conjoint models including hybrid conjoint models which allow many product attributes.