The evolving fifth-generation (5G) cellular networks are intended to overcome the basic challenges of current cellular networks, e.g., higher data-rates, lower latency, network capacity and energy-efficiency (EE). To meet these challenges, 5G networks need to adopt a multi-tier architecture containing different network tiers e.g., macrocells, different types of licensed small cells and device-to-device (D2D) communication to serve their intended users with different quality of service (QoS) demands in an energy-efficient and spectrum-efficient way. The coexistence of different network tiers can be more beneficial if they share the same resources. But, the coexistence of these tiers and sharing the same resources among them may result in severe interference which may degrade the performance of 5G networks. Therefore, a well capable centralized mechanism is required which can efficiently allocate resources to different network tiers in order to achieve the envisioned goals of 5G. To this end, cloud radio access networks (C-RANs) have been presented as effective solutions which can provide high EE, improved network capacity, high network sum-rate, high spectral efficiency and reduced operating expenditures. Therefore, we have combined the potentials of multi-tier architecture and C-RAN and have presented different resource allocation (RA) techniques. xi decomposed into three sub problems namely, sub-channel assignment problem, power allocation problem and RRH-association problem. A three step iterative technique is presented to solve these sub problems in an iterative manner such that sub-channel assignment and power allocation are carried out for the fixed RRH-association in the first and second steps, respectively while in the third step, RRH-association is performed. In addition, two heuristics algorithms for sub-channel assignment namely, individual sub-channel assignment (ISA) and sub-channels block assignment (SBA) algorithms are proposed. Finally, D2D communication enabled SC-FDMA based multi-tier C-RAN is considered and a joint mode selection, sub-channel assignment, power allocation and RRH-association (JMSPR) problem is formulated for EE maximization. The formulated joint optimization problem is MINLP and therefore, it is solved in two stages; mode selection stage and joint SAPARA stage. To solve the mode selection problem, a two-step iterative mode selection algorithm is presented which first identify the potential D2D communication pairs by defining the proximity of the transmitters and then enable D2D communication between the proximate users based on link quality. While, for joint SAPARA, a three-step iterative technique is presented. The performances of the above mentioned proposed techniques are evaluated via exhaustive simulations. It is shown that these techniques significantly enhance the performance of the corresponding multi-tier networks.