As an innovative financial derivatives instruments ,Private equity investment funds can optimize the allocation of resources、promote the rational flow of funds and bring a positive effect on the social and economic development. But because of the irregularity of financial market volatility、economic downward pressure and the exit of many uncertain factors of improving investment environment，Making private equity fund lurking in the process of investment risks，Therefore，an effective prediction for private equity investment risk is of great significance. Based on this, Similarity Weighted Voting Combination is proposed in this paper to build private equity investment risk prediction model，On the basis of knowledge to guide the decision combining the advantages of nearest neighbor mixed case retrieval of decision rules，Putting forward the maximum similarity weighted voting probabilistic decision rules and setting the model parameters in the range of experience value，which makes model effective classification for predicting the risk of our investment case，The empirical simulation results also prove the reliability and effectiveness of the proposed method.