Fault prediction in IPTV using improved AdaBoost algorithm
Yi Qian, Xin Wei, Ruochen Huang, Hao Meng, Qifeng Liu · 2016
IPTV is an increasingly important service for telecom operators. Service providers have to repair IPTV faults quickly in order to provide high quality service. It will be very helpful if faults can be predicted. Considering this situation, we combine status data from the set top box with the data of customer trouble tickets and then build a prediction system with improved AdaBoost to predict faults in IPTV system. First, we clean and conduct some statistical analysis for the dataset. Then, we apply feature selection to the preprocessed data and build the associated models. It is noted that we improve the AdaBoost algorithm by adding F1 measure to cost function and limiting the weights of different classifications. Our evaluation results show that the improved algorithm achieves precision of 88.89% and performs better than original AdaBoost in IPTV fault prediction.