In order to migrate software product variants which are considered similar into a software product line (SPL), it is essential to identify the mandatory and optional features between the product variants. To exploit existing software variants and build a SPL, a feature model of this SPL must be built as a first step. To do so, it is necessary to mine optional and mandatory features from the source code of the software variants. Thus, we propose, in this paper, a new approach to mine features and feature models from the object-oriented source code of a set of software variants based on Formal Concept Analysis and Latent Semantic Indexing.