The purpose of this thesis is to examine the predictability of hedge fund performance by using survival risk and liquidity risk analyses. Institutional investors are interested in long-run investments in the hedge fund industry and the high liquidation rate in the hedge fund industry brings significant risk to their investors. This research not only estimates the relationship between hedge fund characteristics and failure risk, but also examines the relationship between hedge fund survival risk, liquidity risk and their relative performance. This thesis is relevant to both researchers and practitioners in exploring a tangible analysis of hedge fund performance.The sample of this study derives from the TASS database from January 1984 to July 2014. The sampling time period covers the Asian crisis in 1997, the Russian crisis in 1998, the collapse of the sub-prime mortgage crisis in the US in 2007 and the subsequent credit crunch. The original database contains 14,031 hedge funds for this period, of which 6,505 are live funds and 7,526 are liquidated funds. The first empirical chapter estimates the predictability of hedge fund performance by use of a semi-parametric procedure.The results suggest that hedge fund monthly returns are predictable with proper identification of fund failure. The identification of fund failure can extract funds that are liquidated because of poor performance. The empirical evidence suggests that fund failure risk has strong explanatory power regarding hedge fund performanceThe second empirical chapter estimates the predictability of hedge fund performance by using investor-induced liquidity. It suggests that hedge fund liquidity risk derived from investors is an important factor of hedge fund performance analysis. The result also confirms that investor-induced liquidity in the more recent past has more explanatory power regarding its post-performance. Moreover, incubation bias could influence the predictability of hedge fund performance significantly. The result from fund performance shows that the fire sale problem was more significant in the recent financial crisis period but not significant in a normal period.The last empirical chapter investigates the predictability of hedge fund performance by using a combined prediction model. The result indicates that a model combining survival risk and liquidity risk exhibits more detail and performs better than using a prediction model with a single dimension. The result also indicates that incubation bias influences the predictability of hedge fund performance. Moreover, more recent data influences the predictability of hedge fund performance more significantly.On the other hand, long distance past data can provide a more significant result in estimation of covariates by using the Cox proportional hazard model. It is helpful to investigate the interactions between the risk of hedge fund characteristics and their performance practically.