In　order　to　adapt　to　the　characteristics　of　high　nonlinear　and　time-varying　for　air　pollutants　concentration　and　improve　the　real-time　prediction　accuracy　of　air　pollutants　concentration,　a　forecasting　model　of　air　pollutants　concentration　based　on　accurate　online　support　vector　regression　(AOSVR)　algorithm　is　established　in　this　paper.　According　to　the　hourly　SO2　concentration　and　meteorological　parameters　from　May　2014　to　April　2015　in　Wanliu　Monitoring　Station　of　Beijing　in　China,　the　data　of　2　months　are　selected　as　experimental　samples.　At　the　same　time,　Pearson　correlation　coefficient　method　is　used　to　select　the　exogenous　inputs　which　have　strong　correlation　with　the　output　variable.　The　results　show　that　the　AOSVR　algorithm　can　adjust　the　prediction　model　dynamically,　and　the　prediction　accuracy　is　higher　than　that　of　the　conventional　fixed　support　vector　regression　(SVR)　model.　©　2018　IEEE.