5G ultra-dense multi-Radio access technology (Multi-RAT) HetNets are considered recently, by industrials and mobile network operators, as a key solutions for boosting network capacity. Hence, supporting the exponentially increasing demand of data traffic. However, improving multiple conflicting metrics for users, with distinctive quality of service (QoS) and quality of experience (QoE) requirements, remains the main challenge in these network environments. This is due to the inappropriate adaptation of the current user association schemes. To overcome this challenge, we formulated the user association problem, in 5G ultra-dense multi-RAT HetNets, as a multi-objective optimization problem (MOOP), solved by the weighted sum method. Then, we proposed the multi-objective genetic algorithm (MOGA) to reach suitable associations, that respect the individual requirements of each user. Numerical results show the potency of the MOGA with decoupled access (MOGA-DA) compared to the MOGA with coupled access (MOGA-CA) and Max-SINR (Signal-to-Interference-plus-Noise Ratio) association schemes in terms of energy efficiency (EE), cost efficiency (CE) and QoS uplink/downlink throughput.