A portfolio prediction model based on Bayesian game algorithm is proposed to improve the accuracy of correlation analysis model in portfolio application. Firstly, the portfolio problem is analyzed. The market value constraint and the upper bound constraint are combined considering the general portfolio model according to Markowitz theory, it improves the portfolio model and obtains the portfolio model with mixed constraints. Secondly, a probabilistic portfolio monitoring strategy based on incomplete information non-cooperative Bayesian game is proposed. The framework combines the enterprise investment behavior detection module with the specified rules to identify the portfolio pattern, the interaction between portfolio selection and monitored investment behavior is modeled as a two-person non-cooperative Bayesian game, which allows portfolio selection to adopt a probability monitoring strategy based on game Bayesian Nash equilibrium, thereby reducing the computational complexity of the portfolio approach introduced into the game algorithm. Finally, the effectiveness of the proposed algorithm is verified by simulation experiments.