This dissertation proposes a Artificial Intelligence (AI) based pilot allocation scheme for network capacity maximization in massive multiple-input multiple-output (MIMO) systems. First, a deep multi-layer perceptron-based pilot allocation scheme (DL-PAS) is proposed for massive MIMO system that use multiple antennas for multiple users in cases of low-density users. The proposed DL-PAS improves the performance in cellular networks with severe pilot contamination by learning the relationship between pilot assignment and the users' location pattern. In this work, we design a novel supervised learning method where input features and output labels are users' locations in all cells and pilot assignments, respectively. Specifically, pretrained optimal pilot assignments with given users' locations are provided through an exhaustive search method as the training data. Then, the proposed DL-PAS provides a near-optimal pilot assignment from the produced inferred function by analyzing the training data. We implement the proposed scheme using a commercial deep multi-layer perceptron system. Simulation-based experiments show that the proposed scheme achieves almost 99.38 % theoretical upper-bound performance with low complexity, requiring only 0.597 ms computational time. In the other topic of this dissertation, we introduce a novel pilot allocation scheme for a massive multiple-input multiple-output system based on deep-convolutional-neural-network (CNN) learning (DC-PAS). This work is an extension of a prior work on the basic deep learning framework of the pilot allocation problem, the application of which to a high-density user nature is difficult owing to the factorial increase in both input features and output layers. To solve this problem, by adopting the advantages of CNN in learning image data, we design input features that represent users' locations in all the cells as image data with a two-dimensional fixed-size matrix. Furthermore, using a sorting mechanism, we construct output layers with a linear space complexity according to the number of users. We also develop a theoretical framework for the network capacity model of a massive MIMO system and apply it to the training process. Finally, we implement the proposed deep CNN-based pilot allocation scheme using a commercial vanilla CNN system, which takes into account shift invariant characteristics. Through simulation-based experiments, we demonstrate that the proposed scheme realizes almost a 98.00 % theoretical upper-bound performance and an elapsed time of 0.842 ms with low complexity in the case of a high-user-density condition.