In cellular networks, interference is identified as the major performance bottleneck. Attempts are made in 4G and 5G systems to address this by pooling base stations together to form a network multiple-input multiple-output (MIMO) system. Global coordination in network MIMO systems is known to be highly complex and costly. In this thesis, a scalable solution is proposed by clustering the network into groups of base stations. Interference within the cluster is mitigated by performing network MIMO based signal processing in each cluster independently. Interference between clusters is then cancelled by applying fractional frequency reuse (FFR) on a cluster scale. In FFR systems, a greater reuse factor is used for users near the cell or cluster edge since they are more prone to interference. An important problem in FFR systems is classifying the user location as either being in the centre or near the edge. The conventional technique of using a one-dimensional signal-to-interference-and-noise ratio (SINR) threshold is highly inaccurate and an improved machine learning approach based on logistic regression is proposed. It is shown to improve the accuracy to at least 80% and the cell sum rate performance is shown to be comparable to that of a system with 100% accurate classification.