The self-organizing evolutionary nature giving rise to complex patterns exhibited by a cellular automaton has led to sustained research in various modelling and simulation problems. Urban growth prediction is one such problem in which the urbanization trend is imperative for city planning officials to effectively forecast and prepare, and establish policies for efficient urban planning. Traditional CA is combined with mining probability rules for transitions and heuristics for calibrating the CA parameters for accurate predictions of urban growth. The input is modeled after the popular SLEUTH and PLUS models with additional region-specific features. The review analyzes and reports the various types of CA, rule mining techniques, and CA calibrations used for urban growth predictions in different geographical regions. GeoCA simulation results are presented to investigate the effects of CA calibration parameters and algorithms. The adequate input factors for non-city predictions and the need for further auxiliary input data for better accuracy for urban and semi-urban predictions are evident from the results. The future research prospects are put forth based on the outcomes of the review findings and insights from the simulation that suggest critical aspects of urban growth planning, including extended research on identifying region-specific parameters and techniques for fine-tuning the CA model and algorithms.