The Prediction Method of KPIs by Using LS-TSVR
Shiyang Wang · 2022
Closely monitoring service performance and making predictions of Key Performance Indicators (KPIs) are critical for Internet-based services. However, fast yet accurate prediction of these seasonal KPIs with various patterns and data quality has been a great challenge. This paper tackles this challenge through a novel approach based on auto-regressive Least Square Twin Support Vector Regression (LS-TSVR). As an improved version of SVR, LS-TSVR can handle big data without any external optimization, and meanwhile, the prediction accuracy is better than that of SVR. For seasonal KPI data in a production dataset, our methods satisfy or approximate a mean average error (MAE) of around 0.013, which is significantly lower than the baseline method.