The work presented here arises from the need to find a tool for computing the performance metric of a given exploration algorithm under arbitrary test conditions. Such tool would allow to find the most suitable algorithm for a predefined scenario without performing time consuming tests or simulations. The work here is a first approximation to develop the proposed tool, in order to validate the suitability of our approach we choose simple and well-known exploration algorithms and scenarios, but in the future we plan to expand to state-of-the-art algorithms and complex scenarios. A Machine Learning system using supervised learning is proposed as a practical method to predict the desired evaluation metric, such estimation will be used for algorithm comparison instead of the conventional method of performing tests or simulations. Other estimation methods, such as a multivariate regression model and approximation to probability distribution functions, are cumbersome and usually require performing additional experiments to gather further data. For testing and evaluation, we use two sample exploration algorithms based on the principle of two-dimensional random walk and a exploration scenario based on a bi-dimensional rectangular grid.