Abstract: Anomalies in network traffic are detected using machine learning. A plurality of machine learning models is employed to determine whether there are anomalies in network traffic of an MPLS (Multiprotocol Label Switching) network that can affect the performance of devices in the network. A first machine learning model is trained on network traffic passed through network tunnels of a plurality of routers in the network. A second machine learning model is trained on router-specific network traffic passed through router-specific network traffic for a subset of the network tunnels associated with a particular router. The first machine learning model is employed to determine a network anomaly, and the second machine learning model is employed to determine a router-specific anomaly. A router error is identified when both a network anomaly and a router-specific anomaly are determined. An indication of the router error is communicated to a computing device.

Abstract: Anomalies in network traffic are detected using machine learning. A plurality of machine learning models is employed to determine whether there are anomalies in network traffic of an MPLS (Multiprotocol Label Switching) network that can affect the performance of devices in the network. A first machine learning model is trained on network traffic passed through network tunnels of a plurality of routers in the network. A second machine learning model is trained on router-specific network traffic passed through router-specific network traffic for a subset of the network tunnels associated with a particular router. The first machine learning model is employed to determine a network anomaly, and the second machine learning model is employed to determine a router-specific anomaly. A router error is identified when both a network anomaly and a router-specific anomaly are determined. An indication of the router error is communicated to a computing device.

Abstract: Anomalies are detected in network traffic exhibiting a seasonal variation. A neural network is trained using historical network traffic metrics, and as a result, the trained neural network is configured to output a mean error from a network traffic metric input. A decision tree model is trained on a training dataset comprising historical network traffic metric outputs at associated times. To identify an anomaly, network traffic metrics for a particular time are provided as an input to the trained neural network that, in response, outputs the mean error. The particular time is input into the trained decision tree model to output a mean error adjustment. The mean error is adjusted using the mean error adjustment, and the resulting adjusted mean error is compared to a static mean error threshold value to identify the anomaly.

Abstract: A system and method is disclosed to transform Over-The-Top (OTT) call log data collected by telecommunications network operators into a peer-to-peer log in order to determine more accurate information. In this method the IP addresses in the call log records are compared to a list of known mediator server IP addresses and the records in the log that contain mediator server IP addresses are removed from the log to create a peer-to-peer log containing only those records that contain peer IP addresses. The data in the peer-to-peer log can then be categorized with confidence to determine network performance information as well as, for example, the percentage of traffic to a particular country. Conclusions may then be inferred regarding the removed data records based upon the data found in the peer-to-peer log.

Abstract: A system and method is disclosed to transform Over-The-Top (OTT) call log data collected by telecommunications network operators into a peer-to-peer log in order to determine more accurate information. In this method the IP addresses in the call log records are compared to a list of known mediator server IP addresses and the records in the log that contain mediator server IP addresses are removed from the log to create a peer-to-peer log containing only those records that contain peer IP addresses. The data in the peer-to-peer log can then be categorized with confidence to determine network performance information as well as, for example, the percentage of traffic to a particular country. Conclusions may then be inferred regarding the removed data records based upon the data found in the peer-to-peer log.

Abstract: A system and method for improving network coverage is discussed herein. The system and method can improve network coverage based on lack of coverage key performance indicators generated by disconnection data when user equipment (UE) disconnects from and reconnects to a wireless network. The system determines if the disconnection is associated with a lack of coverage by the wireless network or some other reason. When the disconnection is due to lack of coverage, once the UE reconnects to the wireless network, lack of coverage data associated with the lack of coverage can be used to generate key performance indicators which can be used to improve network coverage. For example, an additional network site can be added to the network, a network site operating parameter can be adjusted, or the like.