The　bioinformatics　research　of　lncRNA　has　attracted　much　attention　in　both　academics　and　industry　due　to　the　important　roles　of　gene　expression　in　the　genome.　LncRNA　expression　profile　has　large　number　of　dimensions　and　small　number　of　samples.　In　the　study　of　the　cancer　samples＇　expression　profile,　there　are　a　mass　of　redundant　lncRNAs　which　are　unrelated　to　cancer　classification.　Selecting　key　lncRNAs　which　are　closely　related　to　the　pathophysiology　of　cancer　from　massive　lncRNAs,　can　not　only　find　prognostic　factors　for　survival　in　cancer　patients,　but　also　provide　guidance　for　the　target　of　the　precise　medical　in　the　future　and　greatly　reduce　the　cost　of　biological　analysis.　In　this　paper,　a　new　key　lncRNA　prediction　method　based　on　BPSO　and　ELM　is　proposed,　lncRNAs＇　combination　optimization　problem　is　transformed　into　a　binary　particle　swarm　optimization　model.　The　classification　accuracy　of　ELM　is　used　as　the　evaluation　criterion　of　feature　selection　in　BPSO.　©　2017　IEEE.