As agent-based models (ABMs) become a primary approach for analyzing complex systems, demand increases for an efficient way to explore and characterize their global behavior. In this work, we explore the adaptation of time series analysis (TSA) techniques to the output of ABMs. Specifically, we use two TSA techniques to analyze the behavior of ABMs: dimensionality reduction and multivariate clustering. Our objective is to (1) present a novel approach for approximating ABM trajectories based on initial parameter values, and (2) discover distinct qualitative outcomes of the models via clustering. We use dimensionality reduction, specifically linear modeling of time series with extraction of the slope parameter, as a minimal representation of the trajectory of a simulation. We then assess the ability of classification algorithms to predict the model trajectory based only on initial parameter values. We use multivariate time series clustering to discover distinct classes of outcomes for a given ABM without prior knowledge of the model’s behavior. Although surveys of time series clustering have been published, there is a gap in the literature when it comes to time series data generated by ABMs. We aim to address this gap by conducting the first comparative assessment of clustering methods on ABM-generated time series. We include 9 different clustering techniques on data from three different canonical models, assess direct effectiveness of the clustering techniques, and examine similarities between the methods.