In this thesis, we present proof-of-concept studies that describe how data-driven techniques can be applied to observational data in order to detect and estimate properties of galaxy clusters, which are the largest gravitationally bound objects to have assembled in the Universe. Given the significance of clusters in astrophysics and cosmology, it is important to develop automated methods that are able to efficiently detect and process a large sample of clusters from existing photometric datasets, in preparation for upcoming large-scale galaxy surveys. This can be achieved by employing machine learning algorithms that are suited at solving the tasks at hand. In particular, algorithms that can self-learn the importance of features from the labelled data of known clusters, which minimises the amount of manual input required to make accurate predictions.

Initially, we demonstrate how a popular object detection algorithm can be applied to wide-field colour images to identify and predict the astronomical coordinates of clusters. We then demonstrate how a novel ensemble regression algorithm can be applied to line-of-sight galaxies within colour-magnitude space to estimate the photometric redshift of clusters. Finally, we present a hybrid empirical and analytical model that performs background subtraction of field galaxies along the line-of-sight of clusters within colour-magnitude space and then estimates the richness of clusters within a characteristic radius.

We also compare our findings with the results of existing conventional techniques to examine the overall predictive performance of our methods at generalising to unseen instances. Furthermore, we note that our methods can be combined together into a sequential data pipeline to create a comprehensive catalogue that contains key characteristics (e.g. position, distance, mass) of observed clusters for conducting astrophysical and cosmological research.