During the past decades, cellular networks have been greatly developed. An increasing number of devices such as tablets and mobile phones are connected to cellular networks. The heterogeneous networks (HetNets) play an important role in serving users with different requirements and huge data demands. LTE-A HetNets have been extensively studied for many years. However, most research works have focused on theoretic studies of LTE-A HetNets. Only a few researchers have a chance to access and study the actual HetNets. A big gap exists between theories and actual applications for cellular networks. It is essential to understand the mechanism of HetNets in real-world environments for better network performance. Building a traffic model that is more suitable for the real-world environment is necessary not only for network operators to provide better service and save costs, but also for users to have better experience with strong received signals. This thesis analyzes and evaluates measured data from a real-world LTE-A HetNet, models user traffic in the actual environment, and optimizes the HetNet using the developed models. In this thesis, the real-world LTE-A HetNet is studied in detail. Both the aggregate data and UE (user equipment)’s data are investigated. The main goal is to study the actual environment, understand the mechanisms of the actual system, and model and optimize the users’ actual data traffic in the real-world environment in this thesis. The aggregate data (cell level data) for the HetNet at the University of Regina are analyzed and modeled in detail for all the cells in the actual HetNet. Four indicators are introduced to evaluate the performance of cell level data. In addition, a series of data collection activities are performed at the University of Regina to better understand the real-world LTE-A HetNet. These tests are intended to analyze and evaluate the baseline of the network and measure the dynamic response of the system when the network settings are adjusted. The activities include handover tests with adjusting A3 event handover parameters and indoor cell-splitting tests with interference mitigation techniques (e.g., Almost Blank Subframe (ABS)). The characteristics of the actual scheduler of the HetNet are analyzed in depth by comparing allocated resource blocks of each test device. The performance of different typical and popular schedulers (e.g., Proportional Fair) is compared with the measured data from the real-world. A fairness guaranteed scheduler is proposed to maintain the fairness of user throughput since the fairness is a crucial indicator. This innovative scheduler is developed using the generalized proportional fair (PF) scheduler and control theory. A simulation model is developed to predict user downlink data rate in a dynamic environment with algorithms and measurement. Some indicators are also proposed for the model. Furthermore, both enhancement strategies and algorithms are proposed for the HetNet to increase cell throughput of the overall networks. This model is useful to predict user data rate more accurately and to help the network operators produce effective cell planning and provide seamless service to users. Studying the actual cellular networks will bring more insights about how the actual network behaves and will be beneficial for the deployments of 5G networks in the future, because many features in LTE-A (e.g., small cells) are also crucial to the 5G networks.