机器学习——K最近邻

K鏈€杩戦偦锛圞-Nearest Neighbors锛孠NN锛夋槸涓€绉嶅熀鏈殑鏈哄櫒瀛︿範绠楁硶锛屽睘浜庣洃鐫ｅ涔犱腑鐨勪竴绉嶅垎绫诲拰鍥炲綊鏂规硶銆侹NN鐨勬牳蹇冩€濇兂鏄€氳繃娴嬮噺涓嶅悓鏁版嵁鐐逛箣闂寸殑璺濈鏉ヨ繘琛屽垎绫绘垨鍥炲綊銆傚畠鐨勫熀鏈師鐞嗘槸锛氬鏋滀竴涓牱鏈湪鐗瑰緛绌洪棿涓殑k涓渶鐩搁偦鐨勬牱鏈腑鐨勫ぇ澶氭暟灞炰簬鏌愪竴涓被鍒紝鍒欒鏍锋湰涔熷睘浜庤繖涓被鍒紝鍗抽噰鐢ㄥ鏁拌〃鍐崇殑鏂瑰紡杩涜鍒嗙被銆?/p>

浠€涔堟槸KNN

K鏈€杩戦偦(KNN) 鏄竴绉嶉潪甯哥畝鍗曠殑绠楁硶锛屽叾涓瘡涓娴嬪€奸兘鏄牴鎹叾涓庡叾浠栬娴嬪€肩殑鈥滅浉浼兼€р€濇潵棰勬祴鐨勩€侹NN 鏄竴绉嶅熀浜庤蹇嗙殑绠楁硶锛屼笉鑳界敤灏侀棴褰㈠紡鐨勬ā鍨嬫潵姒傛嫭銆傝繖鎰忓懗鐫€鍦ㄨ繍琛屾椂闇€瑕佽缁冩牱鏈紝骞朵笖鐩存帴鏍规嵁鏍锋湰鍏崇郴杩涜棰勬祴銆傚洜姝わ紝KNN 涔熻绉颁负鎯版€у涔犲櫒銆?/p>

KNN绠楁硶璇嗗埆k 鈥滅浉浼尖€濇垨鏈€鎺ヨ繎琚娴嬬殑鏂拌褰曠殑瑙傚療缁撴灉锛岀劧鍚庝娇鐢ㄨ繖浜涜瀵熺粨鏋滅殑骞冲潎鍝嶅簲鍊硷紙鍥炲綊锛夋垨鏈€甯歌鐨勭被鍒紙鍒嗙被锛塳瑙傚療缁撴灉浣滀负棰勬祴杈撳嚭銆?/p>

濡備綍瀹氫箟璺濈

涓轰簡纭畾鍝簺鏁版嵁鐐规渶鎺ヨ繎缁欏畾鐨勬煡璇㈢偣锛岄渶瑕佽绠楁煡璇㈢偣涓庡叾浠栨暟鎹偣涔嬮棿鐨勮窛绂汇€傝繖浜涜窛绂诲害閲忔湁鍔╀簬褰㈡垚鍐崇瓥杈圭晫锛屽皢鏌ヨ鐐瑰垝鍒嗗埌涓嶅悓鐨勫尯鍩熴€?/p>

娆ф皬璺濈锛坧=2锛夛細杩欐槸鏈€甯哥敤鐨勮窛绂诲害閲忥紝浠呴檺浜庡疄鍊煎悜閲忋€備娇鐢ㄤ笅闈㈢殑鍏紡锛屽畠娴嬮噺鏌ヨ鐐瑰拰鍙︿竴涓娴嬮噺鐐逛箣闂寸殑鐩寸嚎銆?br>

Manhattan distance (p=1)锛?/strong>杩欎篃鏄彟涓€绉嶆祦琛岀殑璺濈搴﹂噺锛屽畠琛￠噺涓ょ偣涔嬮棿鐨勭粷瀵瑰€笺€傚畠涔熻绉颁负鍑虹杞﹁窛绂绘垨鍩庡競琛楀尯璺濈锛屽洜涓哄畠閫氬父鐢ㄧ綉鏍煎彲瑙嗗寲锛岃鏄庝汉浠浣曢€氳繃鍩庡競琛楅亾浠庝竴涓湴鍧€瀵艰埅鍒板彟涓€涓湴鍧€銆?/p>

闂靛彲澶柉鍩鸿窛绂伙細杩欑璺濈搴﹂噺鏄鍑犻噷寰峰拰鏇煎搱椤胯窛绂诲害閲忕殑涓€鑸舰寮忋€備笅闈㈠叕寮忎腑鐨勫弬鏁皃 鍏佽鍒涘缓鍏朵粬璺濈搴﹂噺銆傚綋p绛変簬2鏃舵鍑犻噷寰楄窛绂荤敤杩欎釜鍏紡琛ㄧず锛屾浖鍝堥】璺濈鐢╬绛変簬1琛ㄧず銆?/p>

姹夋槑璺濈锛圚amming Distance锛夛細鏄敤鏉ヨ　閲忎袱涓瓑闀垮瓧绗︿覆鍦ㄥ搴斾綅缃笂涓嶅悓瀛楃鐨勪釜鏁般€傚浜庝袱涓瓑闀跨殑瀛楃涓睞鍜孊锛屾眽鏄庤窛绂绘槸灏咥涓殑姣忎釜瀛楃涓嶣涓殑瀵瑰簲浣嶇疆鐨勫瓧绗﹁繘琛屾瘮杈冿紝缁熻涓嶅悓瀛楃鐨勬暟閲忋€?/p>

鎴戜滑浣跨敤璺濈锛堟垨鐩稿紓鎬э級搴﹂噺鏉ヨ绠楄瀵熷€间箣闂寸殑鐩稿宸紓銆傛渶甯歌鐨勮窛绂诲害閲忔槸娆у嚑閲屽痉鍜屾浖鍝堥】璺濈搴﹂噺銆?/p>

鐗瑰緛鐨勬爣鍑嗗寲

鍦ㄥ簲鐢╧nn绠楁硶杩涜寤烘ā鍓嶏紝蹇呴』瑕佸仛鐨勪竴浠朵簨灏辨槸鐗瑰緛鏍囧噯鍖栥€傚洜涓簁nn渚濊禆浜庤窛绂荤殑璁＄畻锛岀幇瀹炴暟鎹腑锛屽悇涓壒寰佺殑鏍囧害涓€鑸槸涓嶄竴鏍风殑锛屽嵆鏁版嵁鑼冨洿鏈夊樊寮傦紝鍋囧鏌愪釜鐗瑰緛鐨勬暟鍊肩壒鍒ぇ锛岄偅涔堣窛绂荤殑搴﹂噺灏变細寮虹儓鍦拌杩欎釜杈冨ぇ鏁板€肩殑鐗瑰緛鎵€鏀厤锛屽叾浠栫壒寰佽捣鍒扮殑浣滅敤灏卞緢灏忎簡锛屾ā鍨嬩篃灏卞け鍘讳簡鎰忎箟銆傚鐗瑰緛鏍囧噯鍖栨湁澶氱鏂规硶锛屽父瑙佺殑鏈塵in-max鏍囧噯鍖栵紝z-score鏍囧噯鍖栫瓑銆?/p>

min-max鏍囧噯鍖?/h3>

X涓虹壒寰佸悜閲忥紝Xnew涓簃in-max鏍囧噯鍖栧悗鐨勭壒寰佸悜閲忋€俶in-max鏍囧噯鍖栧悗鐨勬暟鍊艰寖鍥翠负[0,1]

z-score鏍囧噯鍖?/h3>

饾泹琛ㄧず鐗瑰緛鍚戦噺X鐨勫钩鍧囧€? 饾洈琛ㄧず鐗瑰緛鍚戦噺X鐨勬爣鍑嗗樊銆倆-score鏍囧噯鍖栧悗鐨勬暟鍊艰寖鍥存槸鏃犵晫鐨?鈭掆垶,鈭?

鍝戝彉閲忕紪鐮?/h3>

鍦ㄥ簲鐢╧nn杩涜寤烘ā鏃讹紝鏈変竴涓墠鎻愬氨鏄缁冩暟鎹殑鐗瑰緛蹇呴』閮芥槸鏁板€煎瀷鐨勩€傚亣濡傛鏃惰缁冩暟鎹腑鏈夊悕涔夌壒寰侊紙绫诲埆鍊硷級

閫夋嫨kNN

KNN 绠楁硶涓殑k 鍊煎畾涔変簡灏嗘鏌ュ灏戜釜閭诲眳浠ョ‘瀹氱壒瀹氭煡璇㈢偣鐨勫垎绫汇€備緥濡傦紝濡傛灉k=1锛屽垯瀹炰緥灏嗚鍒嗛厤鍒颁笌鍏舵渶杩戠殑閭诲眳鐩稿悓鐨勭被銆傚畾涔塳 鍙兘鏄竴绉嶅钩琛¤涓猴紝鍥犱负涓嶅悓鐨勫€煎彲鑳戒細瀵艰嚧杩囧害鎷熷悎鎴栨瑺鎷熷悎銆傝緝浣庣殑k 鍊煎彲鑳藉叿鏈夐珮鏂瑰樊锛屼絾浣庡亸宸紝鑰岃緝澶х殑k 鍊煎彲鑳藉鑷撮珮鍋忓樊鍜岃緝浣庣殑鏂瑰樊銆?/p>

搴旂敤

鍩轰簬澧炲己鍥炲綊鏍戝拰k-鏈€杩戦偦鐨勫弽鎺ㄦ柟娉曞畾閲忓湴涓嬫按鐐规簮鍜岄潪鐐规簮纭濋吀鐩愭薄鏌撶殑褰卞搷

KNN绠楁硶涓弬鏁拌皟鏁寸殑缁撴灉

Sari-Neka 鍚按灞傜閰哥洂姹℃煋鐨?KNN 妯″瀷缁撴灉

缁撹锛?/strong>KNN妯″瀷浼拌鍒嗗埆涓?4 涓紙224.6 km2锛夊拰12%锛?16.6 Sari-Neka 鍚按灞傜殑km2) 鍒嗗埆鏄閰哥洂姹℃煋鐨勯珮鍜岄潪甯搁珮涓ょ被鐨勮剢寮卞尯銆?

鏈夐渶瑕佸涔犳満鍣ㄥ涔犵殑鏈嬪弸娆㈣繋浜ゆ祦