As　the　population　ages　and　health-care　costs　increase,　smart　environments　can　be　an　effective　and　economical　way　to　provide　care　and　support　for　the　aged　population.　Human　activity　recognition　(HAR),　a　key　element　of　the　smart　environment　research　domain,　has　garnered　a　lot　of　attention　lately.　The　present　work　is　to　provide　a　data-driven　solution　based　on　the　extended　belief　rule　base　(EBRB)　model　for　sensor-based　HAR　in　the　context　of　big　data.　More　specifically,　in　order　to　increase　the　efficiency　of　the　EBRB　model,　this　research　first　offers　a　new　rule　generation　method　based　on　probability　estimation,　which　forms　the　link　between　the　extended　belief　rules　and　human　activities.　The　number　of　extended　belief　rules　used　to　extract　knowledge　from　a　sensor-based　HAR　dataset　is　exactly　equal　to　the　types　of　human　activities,　and　each　rule　can　be　thought　of　as　a　collection　of　class　conditional　probability　distributions.　As　a　result,　it　is　possible　to　create　an　EBRB-BD　model,　an　EBRB　model　for　HAR　using　big　data　that　has　a　compact　but　representative　rule　base.　The　effectiveness　of　the　EBRB-BD　model　is　further　supported　by　case　studies.　Experimental　findings　demonstrate　that　the　modelling　time　of　the　EBRB-BD　model　is　one　in　ten-thousand　of　the　original　EBRB　model,　and　the　EBRB-BD　model　also　achieves　the　best　area　under　the　curve　value　(AUC)　of　94.95%,　surpassing　the　original　EBRB　model　and　some　other　benchmark　classifiers.