Optimal budget control in real-time advertising

Thumbnail Image

Files

master_Baggetto_Stefano_2026.pdf (2 MB)

URL

Journal Title

Journal ISSN

Volume Title

School of Science | Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2025-12-23

Department

Major/Subject

Data Science

Mcode

Degree programme

Master's Programme in ICT Innovation
Master's Programme in ICT Innovation
Master's Programme in ICT Innovation

Language

en

Pages

73

Series

Abstract

This thesis focuses on bidding strategies in Real Time Bidding (RTB) for advertisers with constrained resources. The work shows how a limited budget increases the need to optimize it and, therefore, to leverage different bidding strategies to maximize profit while satisfying the constraint. In particular, this work evaluates the use of Reinforcement Learning (RL) to solve this problem, having the agent learn to optimize the budget within the RTB environment. The proposed approach employs a Double Deep Q-Network (DDQN) agent that learns to adjust bid multipliers dynamically, using app installs as the reward signal to align with real business objectives. RL is then compared with a simple baseline and a stronger one used in production, and highlights possibilities and limitations. This thesis demonstrates that RL has potential for solving online optimization problems with performance close to industry standard, although further improving it would be at the cost of exponentially growing complexity. In conclusion, this work expands on the practical considerations related to production grade algorithms, such as explainability and maintainability, and highlight the area of improvement that future work using RL for real life optimization problem should tackle.

Description

Supervisor

Uitto, Jara

Thesis advisor

Bragard, Quentin

Keywords

real-time bidding, reinforcement learning, budget optimization, deep q-network, online advertising, deep learning

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202601201677

Collections

[dipl] Perustieteiden korkeakoulu / SCI