An Autoencoder-Based Dimensionality Reduction Algorithm for Intelligent Clustering of Mineral Deposit Data

Currently, there has been a dramatic growth of data size and data dimension in geophysics, while achieving the rapid advancements of geological big data technologies with the support of various detection methods. Then, high-dimensional and massive geological data impose very challenging obstacles to traditional data analysis approaches. Given the success of deep learning methods and techniques in big data analysis applications, it is expected that they are also able to achieve the satisfactory performance in dealing with high-dimensional complex geological data. Hence, through the combination of one of the effective implementations of deep learning, i.e., autoencoder, and a clustering algorithm, i.e., K-means, in this paper we achieve the dimensionality reduction for complex data, so as to extract useful data features from mineral deposit data, with the purpose of improving computational efficiency. The experimental results demonstrate the effectiveness our developed method.