Volatility in the financial markets is commonplace and it comes with a cost. One of these costs is abrupt and huge drop in stock price that is known as stock price crash. To model this, we propose a new machine-learning based stock crash risk measure using minimum covariance determinant (MCD) to detect stock price crash. Using this proposed dependent variable, we try to predict stock price crash using cross-sectional regression. The findings confirm that the method properly capture the stock price crash and our proposed model performs well in terms of statistical significance and financial impact. Moreover, using newly introduced firm-specific investor sentiment index, it is identified that stock price crash and firm-specific investor sentiment are positively correlated. That is, higher sentiment leads to an increase with stock price crash risk, a relation that remains robust even when different firm sizes and detoned firm-specific investor sentiment index are considered.