Nationwide　O-3　has　been　deteriorating　and　shows　an　obvious　spatial　aggregation　effect　(SAE),　while　there　are　still　many　cities　not　meeting　Chinese　national　standards　for　PM2.5,　demonstrating　the　urgency　for　the　collaborative　control　of　PM2.5　and　O-3.　This　study　adopted　multiple　mathematic　models　and　data　mining　technologies,　including　Moran＇s　I　(MI),　the　self-organizing　map　(SOM),　the　distributed　lag　nonlinear　model　(DLNM),　multivariate　meta-analysis,　and　univariable　multivariate　meta-regression,　and　aimed　to　explore　the　spatio-temporal　trends　and　influencing　mechanisms　of　PM2.5　and　O-3　in　different　regions.　Results　revealed　that　PM2.5　and　O-3　showed　nonlinear　and　lagged　associations　with　meteorology　and　precursors　and　relatively　large　spatial　heterogeneity　existed　in　the　influencing　mechanisms.　The　eight　clusters,　divided　with　SOM　based　on　air　pollution,　can　explain　a　substantial　part　of　spatial　heterogeneity　in　influencing　mechanisms,　which　means　influencing　mechanisms　are　more　consistent　in　regions　with　similar　pollution　characteristics.　PM2.5　and　O-3　heavily　polluted　regions　showed　strong　SAE　according　to　Moran＇s　I　index　(LMI),　and　showed　sensitive　responses　to　meteorology　and　precursors　according　to　meta-analysis　of　DLNM.　Results　also　suggested　that　simultaneously　mitigating　PM2.5　and　O-3　showed　a　promising　long-term　prospect,　and　that　NOx　reduction　should　be　strengthened　in　PM2.5　dominated　months　and　lightened　in　O-3　dominated　months　at　current　O-3-NOx-VOC　regime.　This　study,　with　multi-technology　fusion,　provides　systematic　understanding　of　PM2.5　and　O-3　pollution　in　China　and　scientifically　backed　support　for　the　next-stage　collaborative　control.