Abstract:     The Black-Litterman(B-L) model can combine traditional finance and behavioral finance to quantify investors′ behavioral decisions. Based on the B-L model, this paper studies the strategy of industrial asset allocation in China′s stock market, obtaining quantitative views of investors from historical data through the GARCH family model, building investor confidence level through investor sentiment index, to improve the traditional strategy of industry asset allocation.
    The main features of this paper are as follows: First, the GARCH model family is used to quantify the view vector of the investor′s view to allocate the industry assets, and supplement the reference literature that used only one parameter GARCH model to characterize the return rate of various industry assets; Secondly, this paper combines the investor sentiment index and quantifies the investor confidence level by means of the distribution function in probability science, and provides a new literature supplement to the setting method of confidence level in the BL model. And we discussed the cumulative effect of the industry′s asset allocation, when the investor′s emotional level is higher or lower than the actual market sentiment level. Third, there is an inverse relationship between the investor′s sentiment level and the future market′s medium and long-term profit rate, when the investor′s emotional overreaction. This paper simulates the effect of this behavioral finance theory in actual asset allocation and proposes an asset allocation strategy, which provides a new literature supplement for the relevant empirical research of behavioral finance.
    This paper takes the data of the CSI 300 industry index from January 2006 to December 2017 as sample data. During the period, China′s stock market experienced two baptisms of bull market and bear market, and the sample is more representative. And the investor sentiment index uses the investor sentiment index ISI from CSMAR database.
    The research is based on the B-L model to empirically study the optimal allocation of industry assets. The most important parameters in the B-L model are the investor′s opinion vector and the investor′s confidence level. These two parameters are equivalent to the "strength" and "angle" of the investor′s view about asset allocation. In the research, the GARCH model family was used to fit the return on assets of 10 industries and to forecast the investor′s view vector of return on the industry assets. Meanwhile, this paper uses the probability theory method to "map" the investor sentiment index to the investor′s confidence level, so that the investor′s confidence level could change with the current market situation.
     The results show that the asset allocation performance of the B-L model is better than that of the MV model and the actual market weight. At the same time, as the level of investor confidence increase, the cumulative rate of return continues to increase and the risk is increasing slightly. As the level of investor confidence has declined, the cumulative rate of return has remained essentially the same or even increased slightly. In addition, when the investor′s emotional level is overreacted, the emotional level is used as a reverse indicator in the B-L model to get the purpose of avoiding risk and optimizing asset allocation in the medium and long term.
    The research uses the B-L model combined with the investor sentiment index to study the allocation of industry assets empirically, and draws the following suggestions:
   First, using the B-L model can optimize asset allocation, which means that the B-L model has good application value in actual asset allocation.
   Second, investors who use the BL model for asset allocation can set their investor confidence level far below the actual confidence level of the market if they are not convinced of their own opinions. At least they can guarantee the bottom line and obtain normal returns as the market; if investors are very convinced of their views, the level of investor confidence can be set much higher than the actual level of confidence in the market. However, there are certain risks. If the judgment is wrong, it will face losses; if the investors judge accurately, they can obtain the excess returns.
    Third, when the market sentiment is in the normal level, asset allocation can be carried out according to the normal asset allocation strategy. When the market sentiment is overreacted(too high or too pessimistic), the market sentiment indicator can be used as a reverse indicator for asset allocation. 

Key words: industry asset allocation, Black-Litterman model, investor sentiment index, confidence level