Technology　opportunities　are　important　drivers　of　technological　advances.　Consequently,　methods　for　technology　opportunity　discovery　(TOD)　are　proposed　to　discover　new　types　of　technology　opportunities　and　to　design　criteria　for　defining　and　evaluating　technology　opportunities,　providing　R&D　teams　and　innovators　with　a　plethora　of　inventive　ideas.　However,　current　TOD　methods　have　some　common　limitations.　First,　the　criteria　for　defining　technology　opportunities　are　typically　restrictive,　thus　may　exclude　some　promising　candidates.　Second,　most　criteria　for　evaluating　opportunities　lack　empirical　evidence.　In　this　article,　we　propose　a　general　methodology　for　discovering　technology　opportunities　that　addresses　these　limitations.　We　create　a　less　restrictive　technology　opportunity　space　(TOS),　built　evaluation　models　for　each　candidate　by　learning　from　historical　data,　and　use　optimization　techniques　to　search　the　TOS　for　the　best　technology　opportunities.　We　then　implement　the　proposed　methodology　in　a　case　study　that　discovered　firm-specific　technology　opportunities　in　neural　network　technology.　We　present　technology　opportunities　as　connected　subnetworks　of　subject-action-object　based　knowledge　networks;　designed　industry-level,　firm-specific　and　patent-specific　evaluation　criteria;　use　random　forest　to　develop　the　evaluation　model　from　historical　patents;　and　apply　ant　colony　optimization　to　find　the　best　opportunities.　The　case　shows　the　feasibility　and　effectiveness　of　the　general　methodology　for　TOD.