Software Defined Networks have opened the door to statistical and AI-based\ntechniques to improve efficiency of networking. Especially to ensure a certain\nQuality of Service (QoS) for specific applications by routing packets with\nawareness on content nature (VoIP, video, files, etc.) and its needs (latency,\nbandwidth, etc.) to use efficiently resources of a network. Monitoring and\npredicting various Key Performance Indicators (KPIs) at any level may handle\nsuch problems while preserving network bandwidth. The question addressed in\nthis work is the design of efficient, low-cost adaptive algorithms for KPI\nestimation, monitoring and prediction. We focus on end-to-end latency\nprediction, for which we illustrate our approaches and results on data obtained\nfrom a public generator provided after the recent international challenge on\nGNN [12]. In this paper, we improve our previously proposed low-cost estimators\n[6] by adding the adaptive dimension, and show that the performances are\nminimally modified while gaining the ability to track varying networks.\n