Abstract: The large bandwidth\|delay product of high speed networks（HSNs） makes the conventional feedback\|based mechanisms unsuitable for ABR to fast respond to high bursty VBR traffic, and results in a significant rate mismatch . In order to increase the efficiency of feedback\|based control and acquire full access to left\|over bandwidth by high priority traffic, a prediction\|based proactive control scheme is proposed, which predicts future VBR bandwidth occupancy respectively for different feedback delays of ABRs, and assigns the corresponding ABR sources a rate on the basis of the estimate of total ABR capacity. Strong dependency of VBR traffic in HSNs makes possible the relatively exact prediction of time series, and we can take advantage of the existing general\|purpose linear prediction methods to get real\|time results on\|line. The current value of load factor（LF） can be used to adjust ER feedback , thus compensating the forecasting variance and RTT estimate error. The efficiency （i.e., exact rate match achieved） and fairness of the proactive scheme can be seen intuitively from an example, while their general description and simple proof are also presented.