# Quintet Volume Projection

## TL;DR

This paper introduces an ensemble of econometric models for intra-day trading volume prediction, utilizing Bayesian methods for adaptive, unified, and uncertainty-aware estimations across different volume aspects in equities.

## Contribution

It presents a comprehensive, unified modeling framework for intra-day volume prediction, integrating multiple sub-models with Bayesian methods and introducing ALE for better calibration.

## Key findings

- Bayesian methods enable adaptive volume estimation.
- Unified models improve consistency across volume predictions.
- ALE effectively addresses overestimation risk.

## Abstract

We present a set of models relevant for predicting various aspects of intra-day trading volume for equities and showcase them as an ensemble that projects volume in unison. We introduce econometric methods for predicting total and remaining daily volume, intra-day volume profile (u-curve), close auction volume and special day seasonalities and emphasize a need for a unified approach where all sub-models work consistently with one another. Historical and current inputs are combined using Bayesian methods, which have the advantage of providing adaptive and parameterless estimations of volume for a broad range of equities while automatically taking into account uncertainty of the model input components. The shortcomings of traditional statistical error metrics for calibrating volume prediction are also discussed and we introduce Asymmetrical Logarithmic Error (ALE) to overweight an overestimation risk.