Abstract <p>High-accuracy localization is considered critical for resource management algorithms in modern applications such as the Internet of Things (IoT), smart factories, and autonomous vehicles. Indoor positioning presents a challenging process, as achieving higher accuracy is difficult in indoor environments. This paper proposes a novel nonuniform and optimized clustering mechanism (NUOCM) for indoor localization to manage complex datasets acquired from large-scale indoor radio systems. Using NUOCM-based communication, the research aims to improve network lifetime and serve a reasonable number of user equipments (UEs) or devices in multiple-input multiple-output (MIMO) systems. NUOCM generates clusters through a novel method that uses an ensemble of classifiers from nonuniform cluster layers in 5G-based indoor localization. The UEs or devices are divided into sub-datasets based on values of reference signal received power (RSRP). Base classifiers, including support vector machine (SVM), artificial neural network (ANN), decision tree (DT), and random forest (RF), are trained on these nonuniform clusters across several layers to improve spatial diversity. Finally, the optimal number of layers and clusters is defined by a firefly algorithm (FA), which maximizes localization accuracy and spatial diversity. The experiments evaluate the efficacy of the proposed NUOCM, and the results show increased localization accuracy and network lifetime compared to traditional clustering techniques in MIMO systems.</p>