As　an　important　issue　of　Brain　Informatics　(BI)　methodology,　systematic　brain　data　analysis　has　gained　significant　attractions　in　BI　community.　However,　the　existing　expert-driven　multi-aspect　data　analysis　and　distributed　analytical　platforms　excessively　depend　on　individual　capabilities　and　cannot　be　widely　adopted　in　systematic　human　brain　study.　In　this　paper,　we　propose　a　provenance　driven　approach　for　systematic　brain　data　analysis,　which　is　implemented　by　using　the　Data-Brain,　BI　provenances　and　the　Global　Learning　Scheme　for　BI.　Furthermore,　a　systematic　EEG　data　analysis　for　emotion　recognition　which　is　a　key　issue　of　affective　computing　is　described　to　demonstrate　significance　and　usefulness　of　the　proposed　approach.　Such　a　provenance　driven　approach　reduces　the　dependency　of　individual　capabilities　and　provides　a　practical　way　for　realizing　the　systematic　human　brain　data　analysis　of　BI　methodology.