Abstract
Minimising network downtime has been a challenge to all telecommunication service providers. One of the major causes for such downtime is equipment failure at various locations and rectification works are required on ad-hoc basis. Therefore, if these failures can be predicted and rectified, downtime can be reduced. The system activities and operation parameters of these equipment are reported over the network and logged at a monitoring station. By studying these data from the equipment, many of the equipment related failures can be predicted to ensure minimal downtime and increase customer satisfaction. However, these data are massive and generated at very high velocity. A dynamic and adaptive algorithm is needed to process the huge amount of data and generate predictions based on trends and patterns. This paper presents a rule based analysis with regression technique and best-fit line methods to predict the equipment failure. The warning occurrence pattern is studied on daily basis and a threshold for alarm signal triggering can be set. The output of this work suggests that the symptom of a failure started as early as 9 days before the failure while for prediction within 4 days before the failure has an accuracy of up to 99.9%.
Original language | English |
---|---|
Pages (from-to) | 59-69 |
Number of pages | 11 |
Journal | International Journal of Advances in Soft Computing and Its Applications |
Volume | 8 |
Issue number | 3 |
Publication status | Published - 2016 |
Externally published | Yes |
Keywords
- Big data analytics
- Failure prediction
- Hadoop
- Regression technique
- Rule based analysis