Using neuroevolution for predicting mobile marketing conversion
This paper addresses user Conversion Rate (CVR) prediction within the context of Mobile Performance Marketing. Specifically, we adapt two main neuroevolution methods: Neuroevolution of Augmenting Topologies (NEAT) and Hypercube-based NEAT (HyperNEAT). First, we discuss two mechanisms for increasing...
| Autor principal: | |
|---|---|
| Outros Autores: | , , , |
| Formato: | conferencePaper |
| Idioma: | eng |
| Publicado em: |
2019
|
| Assuntos: | |
| Texto completo: | http://hdl.handle.net/1822/62746 |
| País: | Portugal |
| Oai: | oai:repositorium.sdum.uminho.pt:1822/62746 |
| Resumo: | This paper addresses user Conversion Rate (CVR) prediction within the context of Mobile Performance Marketing. Specifically, we adapt two main neuroevolution methods: Neuroevolution of Augmenting Topologies (NEAT) and Hypercube-based NEAT (HyperNEAT). First, we discuss two mechanisms for increasing execution speed (parallelism and data sampling); a strategy for preventing excessive network complexity with NEAT; and a rolling window scheme for performing an online learning. Then, we present experimental results, using distinct datasets and testing both offline and online learning environments. |
|---|