
Certificate of Completion
THIS ACKNOWLEDGES THAT
HAS COMPLETED THE FALL 2026 DATA SCIENCE BOOT CAMP
Swarnendu BANERJEE
Roman Holowinsky, PhD
November 15, 2026
DIRECTOR
DATE

TEAM
Short Horizon Forecasting from trade activity data
Nhi Dinh,Swarnendu BANERJEE,Francis White

Stocks trade through limit order books, where buyers and sellers post orders at different prices. At any moment the best bid is the most a buyer will currently pay, and the best ask is the least a seller will currently take. The gap between them is the bid–ask spread, which is a rough measure of liquidity and of what it costs to trade. The mid-price sits halfway between the two and is the usual reference for short-term price moves. In this project we study how predictable these two quantities are inside a ten-minute trading window, using the first five minutes of order-book and trade activity data. Forecasting the mid-price indicates the price direction, and predicting how the spread changes can reveal whether liquidity is likely to get better or worse. For this project we use the dataset from the well-known Optiver realized volatility prediction competition on Kaggle.
