Call Incidence Record (CIR) Prediction

Challenge

• Long-standing chronic issue (1.5 years infield) causing random reboots of 70k STBs per month
• Huge data (100 MB log/ STB with 20 TB data/ month) collected, but unused

Solution

• Tata Elxsi’s framework used for pattern recognition in the logs
• Processed 200+ parameters in real-time from the STBs
• Our AI framework predicted the STBs which rebooted before a two-hour window
• We also provided the parameters required to be controlled to avoid the reboot
• “Digital forensic” and “Entropy Analysis” performed for tracking parameter-wise anomaly

Benefits

• Identified potential root causes of failure
• Achieved 67% accuracy in identifying the issue with a false alarm rate of 2%
• Resolution of the chronic issue by the engineering teams, due to the inputs provided