The feature analysis for feature selection is an important step for defining the inputs of the simulation models for urban growth. The simulation models are valuable tools for describing the spatiotemporal evolution and probable scenarios of the urban development of a region. These models use predictors such as distance to the city center, highways, roads, and water bodies, among others, to forecast land changes. The importance of each characteristic is unknown a priori. Furthermore, whether a predictor provides any information is not evident. In this work, we analyze the predictive response of a set of features to determine whether they are informative and their degree of influence in the prediction. The analysis uses the Total Operating Characteristic (TOC) curve to measure a feature's relevance. The results provide insights into the relation of a feature with the urban land change and the degree of influence as a single predictor. Although a broader study that uses sets of features could provide further conclusions, this study serves as a base comparing method and a starting point for feature selection in urban simulation models.
