在線學習運用於4G/LTE網路流量資料辨識

Abstract

透過在線學習演算法可以建立一個動態即時預測模型，而使用kernel的概念可以將線性模型轉換成非線性模型。本研究將著名的在線學習演算法Passive-Aggressive Algorithm結合kernel與Budget的概念，並透過對偶問題的結果，推導出一個可以做資源控管的非線性模型算法並將Passive-Aggressive Algorithm與本論文推倒的演算法運用於4G/LTE網路流量資料辨識的預測上。

Based on online-learning algorithm, we can create a dynamic model, and we can transform a linear predictor into a non-linear one by using kernel functions. This study derives a new non-linear online-learning algorithm by combining the famous online learning algorithm, Passive-Aggressive algorithm with kernels and conception of budget. The new algorithm creates a non-linear predictor, and manage the resources which the model used. Finally both the new algorithm and the original Passive-Aggressive algorithm are perform on the 4G LTE network traffic data.