To　the　problem　of　web　recruitment　enterprise　customers　free　to　pay　prediction,　PCA_BP　algorithm　is　proposed　in　this　paper.　Through　extracted　and　analysis　the　enterprise　customers　online　behavior　log　data　to　build　data　mart　as　prediction　data　source,　The　PCA　method　is　used　to　data　attribute　reduce　and　then　BP　neural　network　algorithm　is　used　for　customers　pay　prediction.　Further　study　shows　that,　PCA_BP　is　better　than　PCA_PBF　to　the　prediction　results.　To　describe　the　PCA_BP　application,　a　well-known　web　recruitment　site　enterprise　customers　to　pay　prediction　is　discussed　as　a　example　and　the　results　proved　that　compared　with　BP　algorithm,　the　PCA_BP　is　more　precise　and　efficient.