Given the advancement in optical and imaging technology, new projects in astronomy commonly aim to produce a wide-field survey of astronomical objects. In particular, object-specific measurement of observed brightness overtime, so-called light curve, can be exploited to determine an object category, which signifies major properties and behavior. This is normally branded as a classification task, for which several models developed in machine learning community have been explored. Despite reported success in recent literature, imbalance data remains a significant factor that limits actual applications of those models. As such, this paper presents an empirical study based on a published data set of LSST light curve profiles, emphasizing the importance of data-level approach to solve imbalance data. Benchmarking classifiers such as k-nearest neighbor (KNN), decision tree and support vector machine (SVM) are employed in this investigation. Also, the use of PCA as a dimensionality reduction is employed to produce informative variables, which are likely to boost the classification performance.