Short-term market impact prediction using data-driven insights

Mohammad Fesanghary 

  Bloomberg LLP




In this work a non-parametric method is proposed to study the predictive power of automated data-driven insights. The insights are trying to identify market irregularities covering more than 200 different story types ranging from hedge fund positions to anomalies in individual equities and indices. The best predictors are identified and results are validated with ML methods.