Politecnico di Torino (logo)

Proactive customer care solution for telecommunication companies by exploiting Amazon Web Services.

Valerio Volpe

Proactive customer care solution for telecommunication companies by exploiting Amazon Web Services.

Rel. Tania Cerquitelli, Marco Brambilla. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Matematica, 2019

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (3MB) | Preview

This work combines machine learning techniques and Amazon Web Services, the cloud computing platform provided by Amazon, to tackle the proactive customer care problem for telecommunication companies. Current solutions consist in a simple KPIs monitoring to identify and address near real time devices issues with algorithms able to trigger specific alarms when a fixed threshold is exceeded. The aim of this work is to carry out activities in background, using analytic tools and machine learning techniques to identify factors which are causing current problems or which are likely to cause future problems. Thanks to very powerful classification methods, the idea is to create algorithms able to drive insights, understand correlations and recognize symptoms potentially leading to customer calls. Solution could bring benefits to Telco companies such as prevent trouble tickets opening, reduce the churn rate and improve the overall customer experience. The cloud computing platform guarantees flexibility and scalability and allows companies to save on operating costs, run the infrastructure more efficiently and resize resources based on changing business needs. In a first phase, data are extracted, cleaned and explored in order to extract only meaningful information. In a second phase, a feature selection is performed in order to improve learning performance and to guarantee a better model interpretability. Then a binary classification model is considered to predict if a problem will occur or not and then a multiclass classification algorithm is trained to identify the specific issue. The developed pipeline is able to detect issue event with a precision of more than 80% and to satisfactorily classify the different types of problem with a weighted precision of 80%. The model is kept up-to-date through a retraining phase.

Relators: Tania Cerquitelli, Marco Brambilla
Academic year: 2019/20
Publication type: Electronic
Number of Pages: 121
Corso di laurea: Corso di laurea magistrale in Ingegneria Matematica
Classe di laurea: New organization > Master science > LM-44 - MATHEMATICAL MODELLING FOR ENGINEERING
Aziende collaboratrici: UNSPECIFIED
URI: http://webthesis.biblio.polito.it/id/eprint/11997
Modify record (reserved for operators) Modify record (reserved for operators)