DASHBOARD DEVELOPMENT OF MACHINE LEARNING BASED ANOMALY DETECTION ON NETWORK MANAGEMENT SYSTEM (NMS)

• Main Author: Veronika, Nadia
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55318
• Institution: Institut Teknologi Bandung

Since 2015, the Indonesian government gas focused on providing internet networks in 3T area (tertinggal, terluar, terdepan). The deployment of this internet network is a project of the Ministry of Communication and Information Technology (Kominfo) in collaboration with other ministries in Indonesia on the Universal Service Obligation (USO) or Kontribusi Kewajiban Pelayanan Universal (KPPU) program. In addition, Kominfo through BAKTI targeting all villages in Indonesia can access 4G network by 2022. 4000 Base Tranceiver Station (BTS) are planned to be built on Papua and Papua Barat. This project is BAKTI’s responsibility to provide internet access using USO funds. To achieve this goal, collaboration between the government and cellular operators is required. In order for the optimal use of USO funds, the government must be able to ascertain whether the infrastructure built by telecommunications operators can operate properly and provide internet services at any time. The service monitoring process can be done by looking at the sensor’s uptime. However, there is an error where the uptime data received from a sensor does not always increaces with the addition of time. This can be caused by the connectivity problems or indeed the sensor is off. There have been a study that discuss these errors and devides them into four cases, but there are still some error that cannot be classified into these four cases, called anomalies. Therefore, we created a system that can detect these anomalies and classify them into new cases using machine learning and display the details in an easily accessible dashboard. From the research that has been done, the system found three new cases (anomalies). The dashboard will display the uptime graph, the number of anomalies detected on each sensor, and the health of the device/sensor. In addition, this system will also send notification to user regarding the anomalies found and the device health condition once a month. The results of this project can help operators to know the condition of each of their devices so that evaluation can be done easily and quickly.