Aiming at covering the weaknesses of prospect models in investment sentiment and preference decision-making, we propose the investment decision model with sentiment attribute based on the definition of investor sentiment in perspective of preference got from the BW(Baker and Wurgler)theory. We firstly define a probability distribution of emotional transition based on the influence of herd mentality and income tracking on investors. Then, we construct a value function of the additional binary nonlinear parameters to sort the preference of the optional investment items. Finally, the maximum information entropy under the income constraint is formalized as the objective function of the risk aversion and the maximum mean rate of return under the entropy constraint is formalized as the objective function of the risk appetite respectively according to the preference decision and the heterogeneous risk, acquiring the investment ratio through iteratively calculate. The simulation results show that the two-stage decision model can better fit the local optimal decision-making scheme chosen by irrational investors.The result of simulation and illustrative example show that: The proportion of pessimistic investors will increase when returns fluctuate beyond the critical value; The initial proportion of optimistic investors will change individual preference decisions to some extent; The risk seekers prefer to assign greater investment weight on high-risk projects; The well-organized economic market is helpful for investors to make investment decisions.