Abstract: The problem of model uncertainty is the model's basic characteristic in the system of society and economy, the portfolio choice often studies the characteristic of investment decision under a specific environment based on an asset return model that describs the process of asset price. Since the investor is often hard to imagine all states of nature relevant to his portfolio choice problem which causes an uncertainty concerning an asset return model, the uncertainty is characterized as general uncertainty where neither states of nature nor their probabilities are known. So, in this paper, the problem of optimal portfolio choice is studied under the mean-variance model with general uncertainty by introducing a stochastic variable to measure its general uncertainty, which is characterized by the memory and indifference parameter, and reflects the investor's model belief degree. Based on the theory of capital market line, it builds that the optimal portfolio choice is a linear function of the model belief degree and the conventional portfolio choice based on the mean-variance model; based on the different pairs of memory and indifference parameter, the method of case-based reasoning is applied to determine the optimal model belief degree of the investor with quadratic utility function, obtain the optimal portfolio choice under the mean-variance model with general uncertainty, and do an empirical study based on the sample of Shanghai Exchange Composite Index monthly returns from January 1997 to August 2014. Empirical results show, the investor with a large risk aversion, characterized by high memory parameter and small indifference parameter, quickly adjusts its model belief degree and the portfolio choice. Such an investor may exhibit availability and representativeness biases, and often take the active portfolio strategy; on the contrary, the model belief degree and the portfolio choice is gradually adjusted, the investor may exhibit anchoring and conservatism biases, and often take the passive portfolio strategy; the effect of the nearer stock market information on optimal portfolio choice is stronger than that of the farther stock market information on optimal portfolio choice, but the effect of model general uncertainty on optimal portfolio choice is strongest. This study reflects the investor's bounded rationality and extends the problem of portfolio choice to the field of behavioral finance.

Key words: portfolio choice, general uncertainty, case-based reasoning, model belief degree