This thesis focuses on the use of machine learning in financial event prediction. In the past, finance academics had to be content with mostly linear models that could only ingest a small number of variables of a particular type. Now we can use non-linear models with a larger number of variables...

I study the use of non-linear models and accounting inputs to predict the occurrence of litigated bankruptcies and their associated filing outcomes. The main purpose of this study is to identify the accounting patterns associated with bankruptcies. The filing outcomes include, among others, how...

Nonlinear classification models can predict future earnings surprises with a high accuracy by using pricing and earnings input data. Surprises of 15% or more can be predicted with 71% accuracy. These predictions can be used to form profitable trading strategies. Additional variables have been...

This paper predicts the likelihood that a restaurant will close within the next one to two years using a Yelp restaurant dataset and a high dimensional gradient boosting machine called LightGBM (hereafter GBM). This model, trained on more than 20,000 individual restaurants, has an accuracy just...

This short report deals with the recent rise of programmatic time series methods. This decade has witnessed the proliferation of commercial and open source time-series tooling, which calls for an exposition of what is publicly available. In tandem with this survey, AtsPy, an open source...

We build statistical models to describe how market participants choose the direction, price, and volume of orders. Our dataset, which spans sixteen weeks for four shares traded in Euronext Amsterdam, contains all messages sent to the exchange and includes algorithm identification and member...