The　main　objective　of　this　study　is　to　determine　the　more　appropriate　computational　intelligence　(CI)　model　for　the　prediction　of　air　pollutants　in　urban　areas.　In　this　paper,　in　order　to　emphasize　the　importance　of　short-term　air　quality　(AQ)　prediction,　PM2.5　is　used　as　an　example　to　evaluate　the　concentration　of　pollutants　using　a　variety　of　CI　methods　and　tools.　According　to　the　data　of　air　quality　monitoring　stations,　the　main　air　pollutants　O3,　CO,　NO2,　SO2,　PM10,　PM2.5　and　two　kinds　of　meteorological　factors　temperature　and　humidity　are　selected　as　influencing　factors.　Comparing　with　the　model　of　extreme　learning　machine　(ELM),　fuzzy　neural　network　(FNN)　and　least　squares　support　vector　machine　(LSSVM),　wavelet　Neural　Network　(WNN)　model　is　constructed　for　short　time　prediction　concentration　of　PM2.5.　The　experimental　results　show　that　the　detection　results　based　on　WNN　are　more　accurate,　higher　precision　and　strong　self　-　learning　ability.　©　2018　IEEE.