Anomaly Detection Utilizing Throughput Predicting Model in LTE Network

• Main Author: Nur Arifin, Hasan
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/40138
• Institution: Institut Teknologi Bandung

Throughput can be interpreted as the amount of data sent from the transmitter to the recipient of the unity of time (bits per second/bps). Network throughput cannot do anything that occurs outside the normal throughput limit. Therefore, the pattern of throughput in an LTE network can generally be used as an indicator to find out an anomaly. In this study, the detection of utility anomalies through downlink peak throughput, uplink peak, downlink average and uplink average of users is done. The approach used is by constructing prediction models The decomposition of the moving average, after getting the best prediction results with the error size method, then evaluating it using the Weight Distance Reference (RDW) where the value of empathy throughput is compared with the expected value of the model to get a threshold value, this threshold value will be tested whether it is effective or not in accordance with the anomaly. The results showed that the study showed the most accurate downlink peak throughput, the average downlink, and the uplink average was per 12 hours while the uplink peak was per 24 hours, the results of anomalous detection showed that the results of the dataset with matrix confusion obtained the results of approving detection the anomaly is 76.48%. The results obtained are quite effective in detecting throughput anomalies on LTE networks.