Techniques d'analyse dynamique des média sociaux pour la relation client
Duc Kinh Le Tran · HAL (Le Centre pour la Communication Scientifique Directe) · 2015
This thesis is in the field of data mining and in the context of Customer Relationship Management (CRM). With the emergence of social media, companies today have seen the need for an interchannel (or cross-channel) strategy in which they keep track of their clients' histories through a consistent combination of multiple channels. The goal of this thesis is to develop new data mining methods which allow predicting customer behaviors using data collected from multiple channels such as social media, call center¿ We are interested in all types of customer behaviors that characterized their engagement with respect to the company. First of all, we perform a needs analysis in terms of data mining for interchannel CRM strategy. Next, we propose a new method of prediction of customer behaviors in the context of interchannel CRM. In our method, we use a social attributed network to represent the data from multiple channels and perform incremental learning based on latent factor models. We then carry out experiments on both synthetic and real data. We show that our method based on the latent factor models is capable of leveraging informative latent factors from interchannel data. In future works, we consider some ways to improve the performance of our method, especially latent factor models that are able to leverage different types of relational correlation between individuals in the social graph.