Connectivity has been introduced to the car industry and currently Volvo, amongother automobile companies, currently has cars which are connected to the internetand can share data with external devices or services. However, these connected carsoften face issues with connectivity which is a concern for user quality of experience(QoE). One such issue is the difficulty of knowing how the connection changes over timeand if there are issues with said connectivity. In this work, use of different machinelearning techniques on charged data record (CDR) data is described to forecast thedefined key performance indicators (KPIs) derived from the CDR data. Additionally,use of unsupervised machine learning techniques to detect anomalies in the KPIs isinvestigated. The results show that in case of forecasting models, performance of Longshort term memory (LSTM) model surpasses other models.In case of unsupervisedmachine learning techniques like clustering methods, the performance of K-Means++model is found to be mediocre when evaluated using confusion matrix. / <p>Online</p>
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:mau-57872 |
| Date | January 2022 |
| Creators | Skiöld, David, Arora, Shivani |
| Publisher | Malmö universitet, Fakulteten för teknik och samhälle (TS) |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0018 seconds