Publicação
Identifying clients’ bad experiences with their internet service
| Resumo: | Identifying clients who had experienced a bad internet service is important for network providers, as bad service experiences may lead to less client satisfaction. It is possible to measure quality of service by looking at objective network quality measures. However, a decrease in the quality of service will not translate into a bad quality of experience for all clients at all times. This is because a) if the client does not try to use the internet, he or she would not notice the deterioration of the service; and b) different clients have different needs in terms of service quality; a slight decrease in network quality maybe be noticed by an intensive user but not by a light user, even if the latter is using the internet. In the present report, we describe the work we have done to develop: a) a segmentation the clients according to their typical internet usage; b) a probability that a given client would use the internet at a given time. These two features were then fed to a classifier, along with the objective network quality measures. This classifier, a gradient boosted model, was able to classify clients who filled a service request due to lack of access to the internet with an accuracy of 0.98, sensitivity of 0.87 and specificity of 0.98. The results of the classifier and the role of the special features we developed is discussed, along with future directions for this work. |
|---|---|
| Autores principais: | Lavado, Susana Margarida Silva Ferreira |
| Assunto: | Quality of experience Internet service Gradient boosted models Clustering Internet usage Serviço de internet Qualidade de experiência Análise de clusters Utilização da internet |
| Ano: | 2019 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Identifying clients who had experienced a bad internet service is important for network providers, as bad service experiences may lead to less client satisfaction. It is possible to measure quality of service by looking at objective network quality measures. However, a decrease in the quality of service will not translate into a bad quality of experience for all clients at all times. This is because a) if the client does not try to use the internet, he or she would not notice the deterioration of the service; and b) different clients have different needs in terms of service quality; a slight decrease in network quality maybe be noticed by an intensive user but not by a light user, even if the latter is using the internet. In the present report, we describe the work we have done to develop: a) a segmentation the clients according to their typical internet usage; b) a probability that a given client would use the internet at a given time. These two features were then fed to a classifier, along with the objective network quality measures. This classifier, a gradient boosted model, was able to classify clients who filled a service request due to lack of access to the internet with an accuracy of 0.98, sensitivity of 0.87 and specificity of 0.98. The results of the classifier and the role of the special features we developed is discussed, along with future directions for this work. |
|---|