Framework for Multi-Agent Simulation of User Behaviour in E-Commerce Sites (Intro)

Initial presentation of the preparation of my MSc dissertation

Transcript

1. FRAMEWORK FOR MULTI-AGENT SIMULATION OF USER BEHAVIOUR IN E-COMMERCE SITES Supervisors Hugo Sereno Ferreira João Azevedo Author Duarte Duarte 2015-12-08 1
2. Context • Customers interact with e-commerce websites in different ways • Companies want to optimize success metrics (CTR, CPC, …) for profit • Changing what, how and when content (ads, recommendations, …) is displayed influences customers’ interactions 2015-12-08 2 ? ?
3. 2015-12-08 3
4. Context • Summarizing and analysing this behaviour is expensive, hard, tricky, … • Data scientists need to resort to online techniques with a high operational cost 2015-12-08 4 A/B testing multi-armed bandit optimization others
5. Towards a solution 2015-12-08 5 usage + personas structure content evaluation of e-commerce success metrics multi-agent simulation (Bayesian networks, Markov chains, …) input
6. Why? • Evaluation of “what-if” scenarios by data scientists • Recommendation engines • Ad placement/relevance • Segmentation and targeting • Novelty 2015-12-08 6
7. Q & A Thank you! 2015-12-08 7