Novos indicadores de QoE para redes de serviços móveis por meio da extração de opiniões das redes sociais

Abstract

Mobile networks have played an important role in society, allowing users to perform a wide variety of activities, such as video and voice calls, internet connection, banking operations, social media communication, and others. This has boosted the telecommunications market, which is constantly seeking to meet customers who are increasingly accustomed to the facilities offered by these networks. In this reality, it becomes essential that mobile network service providers seek to understand the quality of experiences experienced by their users. In this context, studies on quality of experience (QoE) have contributed to the assessment based on the user experience in relation to certain products or services, being applied especially in the field of telecommunications. In this research, our objective was to identify new QoE indicators for the main mobile network operators in Brazil (Oi, Tim, Vivo and Claro). For this, user opinions on social media were used as a basis for analysis. They were collected opinions on Twitter and applied machine learning techniques with the BERT model (Bidirectional Encoder Representations from Transformers), in the BERT multilingual and BERTimbau versions, because both models have linguistic knowledge for Brazilian Portuguese. The BERTimbau model obtained a superior accuracy, reaching 0.80. In the automatic prediction phase, the results showed parity in QoE indicators among operators on social media. They were compared these results with information from Anatel, the country’s mobile network regulatory agency, revealing that social media are mainly used by users to express dissatisfaction with services, while Anatel surveys revealed differences in user satisfaction. In addition, a correlation was observed between the number of complaints registered with Anatel and the demonstrations on social media.