This paper describes the methodology to carry out the analysis of different features to automatically classify seismic signals from the Chilean Llaima volcano. Two types of events related to the magmatic activity inside the volcano-longperiod and volcano-tectonic-are considered together with other kind of events that are not related to volcanic activity. The aim of this research is to analyze the impact on the classifier performance of a set of features used either individually or in groups in order to discriminate the three types of events. Support vector machine is used for classification and a genetic algorithm is employed for the optimization process of the features subset. The results show an important improvement of the classification accuracy when the events are classified using more than one feature. The influence of each feature as well as the importance of features not considered in previous works is discussed in the conclusions.