Predicting the Abnormal Market Movement from Annual Earnings Announcements

Earning statements are vital in understanding the financial condition of a company. These statements provide investors the ability to make informed decisions about their involvement with a company. Event Studies were created to measure the impact of an event on the price of a security. This thesis uses the event study methodology in conjunction with many machines learning methods to predict if there will be abnormal movement to a securities price in the stock market, following their yearly earnings announcement. Machine learning methods include logistic regression with regularization, support vector machines, random forests, and neural networks. Final analysis supports use of the Lasso logistic regression for feature selection to be converted into a random forest. The feature selection allowed the model to predict abnormal price movement with 84% accuracy