基于Black-Litterman模型的基金投资组合策略研究

Research on Fund Portfolio Strategy Based on Black-Litterman Model

摘要: 基金投资交易规模随着我国金融市场的发展不断壮大，基金经理的资产配置和选股能力成为投资者购买基金产品时主要的关注因素。本文将基金经理个人特征指标纳入基金业绩评价指标中，构建基金综合评价指标，采用集成学习模型(Xgboost，Catboost和LightGBM)并选择预测误差(MSE)最小的模型，结合纳入主观因素的Black-Litterman模型进行资产配置，实证检验基金综合业绩指标是否能在B-L模型构建的资产配置策略中获得超过市场的收益率。实证结果表明，集成学习模型中，Xgboost对数据的预测效果较好，使用Xgboost进行基金收益率预测并使用四种资产配置模型(等权重模型、风险平价模型、最小方差模型、基于集成学习优化后的Black-Litterman模型)进行投资，发现Black-Litterman模型与风险平价模型具有较好的收益回报，而等权重模型、风险平价模型并不能带来超过市场指数的收益。

Abstract: With the development of China’s financial market, the scale of fund investment transactions has been growing rapidly, and the asset allocation and stock selection ability of fund managers have become the main factors for investors to pay attention to when buying fund products. In this paper, the personal characteristics of fund managers are included in the fund performance evaluation index, the comprehensive evaluation index of fund is constructed, and the model with the smallest prediction error (MSE) is selected by the integrated learning model (Xgboost, Catboost and LightGBM). Combined with the Black-Litterman model incorporating subjective factors for asset allocation, this paper empirically tests whether the comprehensive performance index of the fund can obtain the rate of return exceeding the market in the asset allocation strategy constructed by B-L model. The empirical results show that in the ensemble learning model, Xgboost has a better prediction effect on the data. Xgboost is used to predict the fund return rate and four asset allocation models (equal weight model, risk parity model, minimum variance model and Black-Litterman model optimized based on ensemble learning) are used for investment. It is found that Black- Litterman model and risk parity model have better returns, but the equal weight model and risk parity model cannot bring more returns than the market index.