Intermediate Deep Learning: Predicting Closing Price for Kaggle Competition

Stock exchanges have very high volatility in which stock prices can change drastically in the final few minutes. At the end of every trading day the official closing prices are established, which serve as important indicators for investors in order to evaluate the performances in certain stocks.

In this paper, our goal is to predict closing stock prices for different stocks given data from the order book and the auction book. We will be using data from the Kaggle competition “Optiver Trading at the Close”, which consists of data for 200 different stocks for several different time points.

Furthermore, we shall be attempting to fit different deep learning models that we have learnt throughout the course in 10-617 to determine whether they achieve higher performances than linear regression models or regular machine learning algorithms. In the end, we discover that machine learning algorithms outperformed linear regression models, whereas linear regression models still performed better than all deep learning models.