IVE: Enhanced Probabilistic Forecasting of Intraday Volume Ratio with Transformers
Hanwool Lee, Heehwan Park
TL;DR
The paper tackles minute-level intraday volume ratio forecasting to improve VWAP execution by introducing IVE, a Transformer-based model with a probabilistic forecast via a distribution head. It stabilizes volatile intraday volume with a log-normal transformation, fuses volume statistics, time, and stock features, and trains a four-layer encoder-decoder using a Student's t-distribution head. The approach achieves superior MAE/RMSE against recurrent baselines in both Korean and US markets and demonstrates practical value through live trading in Korea, outperforming VWAP benchmarks by 4.82 bp on average with a 59% beat rate. The work highlights the viability of probabilistic forecasts for handling intraday spikes and volatility, and points to future integration of additional indicators and real-time strategy optimization.
Abstract
This paper presents a new approach to volume ratio prediction in financial markets, specifically targeting the execution of Volume-Weighted Average Price (VWAP) strategies. Recognizing the importance of accurate volume profile forecasting, our research leverages the Transformer architecture to predict intraday volume ratio at a one-minute scale. We diverge from prior models that use log-transformed volume or turnover rates, instead opting for a prediction model that accounts for the intraday volume ratio's high variability, stabilized via log-normal transformation. Our input data incorporates not only the statistical properties of volume but also external volume-related features, absolute time information, and stock-specific characteristics to enhance prediction accuracy. The model structure includes an encoder-decoder Transformer architecture with a distribution head for greedy sampling, optimizing performance on high-liquidity stocks across both Korean and American markets. We extend the capabilities of our model beyond point prediction by introducing probabilistic forecasting that captures the mean and standard deviation of volume ratios, enabling the anticipation of significant intraday volume spikes. Furthermore, an agent with a simple trading logic demonstrates the practical application of our model through live trading tests in the Korean market, outperforming VWAP benchmarks over a period of two and a half months. Our findings underscore the potential of Transformer-based probabilistic models for volume ratio prediction and pave the way for future research advancements in this domain.