<p>Software-defined networking (SDN) is a promising network model for the future Internet. SDN uses OpenFlow switches to forward the traffic across the network. However, these switches have a limited table size, which can result in table overflow if the space is not managed efficiently. Exceeding the flow table capacity can severely slow down the network performance by blocking the new flows. The current methods of managing new flow setup requests use a fixed conventional timeout, which is unsuitable for dynamic traffic flows. To mitigate this overflow problem, the arrival time of packets at the SDN switch should be considered, as it influences the idle timeout. Short- and long-lived flows should have small and large timeouts, respectively. In addition, early eviction of flows should occur once the flow table capacity reaches a safe threshold to prevent overflow. This paper proposes a novel Dynamic-Timeout Prediction and Proactive Eviction approach. The dynamic timeout is varied using suitable timeout prediction based on flow characteristics in the Decision Tree model. The controller assigns the dynamic timeout to flow entries according to the packet inter-arrival time. At the same time, proactive eviction is used to evict the unused flow entries early if the available space inside the flow table is insufficient to accommodate the new flows. The current flow table occupancy status is continuously collected. Simulation findings show that the proposed technique works better than the existing reactive methods of dynamic timeout in terms of latency, packet-in, flow table occupancy, and packet-drop for different datasets.</p>