Long non-coding RNAs (lncRNAs) make up a significant portion of non-coding RNAs and are involved in a variety of biological processes. Accurate identification/annotation of lncRNAs is the primary step for gaining deeper insights into their functions. In this study, we report a novel tool, PLncPRO, for prediction of lncRNAs in plants using transcriptome data. PLncPRO is based on machine learning and uses random forest algorithm to classify coding and long non-coding transcripts. PLncPRO has better prediction accuracy as compared to other existing tools and is particularly well-suited for plants. We developed consensus models for dicots and monocots to facilitate prediction of lncRNAs in non-model/orphan plants. The performance of PLncPRO was quite better with vertebrate transcriptome data as well. Using PLncPRO, we discovered 3714 and 3457 high-confidence lncRNAs in rice and chickpea, respectively, under drought or salinity stress conditions. We investigated different characteristics and differential expression under drought/salinity stress conditions, and validated lncRNAs via RT-qPCR. Overall, we developed a new tool for the prediction of lncRNAs in plants and showed its utility via identification of lncRNAs in rice and chickpea. © The Author(s) 2017. Published by Oxford University Press on behalf of Nucleic Acids Research.