Internet of Things (IoT) has now become a common term in technological communities. Machine-to-machine communications (M2M) is one of the components of IoT which deals with communication of devices to device and device to the network aspects. LTE-A networks have emerged as one of the preferred underlying communication networks to support IoT or M2M traffic. However, the uplink packet scheduling to optimize the QoS (quality of service) of M2M traffic without affecting or least affecting the QoS of Human-to-human (H2H) traffic (traffic generated by regular users (also called as H2H users) such as smartphone traffic, Internet traffic, voice traffic, etc.) is one of the main challenges in LTE-A networks. As a solution, we propose an uplink packet scheduling algorithm, called as enhanced class based dynamic priority (E-CBDP) algorithm, which ensures the QoS of H2H traffic by giving it priority over M2M traffic but, optimizes the QoS of M2M traffic by pushing the scheduling of H2H traffic to their delay boundaries. Further, a dynamic M2M traffic control threshold is defined to enable the operator for proactive regulation of the huge M2M traffic to avoid network congestion as well as to minimise the impact on H2H traffic. We characterize the proposed scheduling algorithm via a mathematical analysis for the metrics such as average length of the H2H queue, average length of the M2M queue, average waiting time of H2H packets, and average waiting time of M2M packets. We compare the performance of E-CBDP algorithm with some recent solutions in terms of average delay, percentage of packets dropped, aggregate throughput, fairness, and energy consumption. Simulation results show that E-CBDP algorithm demonstrates excellent performance in terms of packet drop rate and fairness while provides satisfactory performance in terms of delay, throughput, and energy consumption. © 2018 Elsevier Ltd