Recommendation Systems using holistic customer insights
Big Data Spain
Commercialization of smartphones is a highly changing market where customers has the hard task to find a good quality-price relation in a ocean of possibilities. In this sense, we propose an innovate product able to help our customers select the best options based on their needs. From a technical perspective, we have been working in a system able to work with huge, real time data, with an online self-learning and with the possibility of not having historical information to learn about, i.e., with cold start. Algorithm is based on a Reinforcement Learning model called Contextual Multi-armed Bandit, used by Netflix in its frame’s content selection.
Otherwise, we wanted to add more value to the recommendation system adding the customer’s opinion based on Market Research techniques. We ask to our customers about smartphone preferences. The research was done online, using the Conjoint Analysis technique. Conjoint analysis allows the researcher to obtain the utility that customers give to different aspects in the purchase process, in this case, the smartphone selection. It also allows to explore the complete universe of possible mobile phones, not only those available in stock but any combination of device features. Then we use these utilities to enrich the output of the Reinforcement Learning algorithm.
Finally, we humanize the recommendation through a cognitive sentence generated by using NLP. As a result we offer our customers an awesome AI tool which can help them when they choose their next smartphones.