ͳ¼Æ¾Ö´óÊý¾Ýͳ¼ÆÆ½Ì¨½¨Éè·½°¸ ÖÇ»Ûͳ¼Æ´óÊý¾ÝÔÆÆ½Ì¨½¨Éè·½°¸ ÏÂÔØ±¾ÎÄ

ÀûÓù¤ÉÌÐÅÏ¢¡¢ÆóÒµ»ù±¾ÐÅÏ¢¡¢Ë°ÎñÐÅÏ¢µÈ¼Ç¼£¬ÎÒÃÇ¿ÉÒÔ¹¹½¨³öÒ»¸öÍøÂçÍØÆË½á¹¹£¬ÒÔ¹¹½¨ÆóÒµµÄÏàÓ¦ÊôÐÔ±êÇ©£¬ÊôÐÔ±êÇ©¿ÉÒÔ·´Ïò±ê¼Ç»ØÆóÒµ£¬´Ó¶øÆóÒµ¾ßÓÐȺÌåÊôÐÔ¡£

ͼ3-14ÆóÒµ¾ÛÀàͼ

¾ÛÀà·ÖÎöʱÊý¾ÝÍÚ¾òµÄÒ»¸öÖØÒª·ÖÖ§£¬Ä¿±êÊǽ«Êý¾Ý¶ÔÏó·Ö×é³ÉΪ¶à¸öÀà»ò´Ø£¬ÔÚͬһ¸öÀàÖеĶÔÏóÖ®¼äÓнϸߵÄÏàËÆ¶È£¬¶ø²»Í¬ÀàÖеĶÔÏó²î±ð½Ï´ó¡£ÔçÆÚµÄ¾ÛÀà¶à²ÉÓÃÎ޼ල·½Ê½£¬µ«Ëæ×ÅÊý¾ÝÍÚ¾òÁìÓòµÄÀ©Õ¹£¬ºÜ¶àµÄÐÐÒµÎÊÌ⣬ÐèÒªÓмලµÄÊý¾ÝÍÚ¾ò¼¼Êõ¡£

ͼ3-15»ùÓÚÄÚ´æ¼ÆËãµÄ·Ö²¼Ê½¾ÛÀàËã·¨

48

Ë㷨Ŀ±êÊÇÄ£¿é¶Èº¯Êý£¬¸Ãº¯Êý¶¨ÒåΪ´ØÄÚʵ¼ÊÁ¬½ÓÊýÄ¿ÓëËæ»úÁ¬½ÓÇé¿öÏÂÝýÄÚÆÚÍûÁ¬½ÓÊýĿ֮²î£¬ÓÃÀ´ºâÁ¿ÆóÒµÏîÄ¿µÄ»®·ÖÖÊÁ¿£¬Õû¸ö¹ý³Ì×Ôµ×ÏòÉϽøÐС£

ͼ3-16 ¾ÛÀàË㷨ģ¿éͼ

ͼ3-17 ¾ÛÀàË㷨ģ¿éͼ

ÄÚ´æ¼ÆË㣨In-Memory Processing£©£¬ÊµÖÊÉϾÍÊÇCPUÖ±½Ó

49

´ÓÄÚ´æ¶ø·ÇÓ²ÅÌÉ϶ÁÈ¡Êý¾Ý£¬²¢¶ÔÊý¾Ý½øÐмÆËã¡¢·ÖÎö¡£ÓÈÆäÊǶÔÓÚ¸´ÔÓµÄÄ£ÐÍ£¬ÍùÍùÐèÒª½Ï¶àµü´ú´ÎÊý²ÅÄÜÊÕÁ²£¬¶ø»ùÓÚ´ÅÅ̵ķÖÎö·½Ê½£¨HadoopÉϵÄMahout£©ÔòIO¿ªÏú¾Þ´ó¡£ £¨2£©»ùÓÚÄÚ´æ¼ÆËãµÄ·Ö²¼Ê½·ÖÀàËã·¨

·ÖÀàÎÊÌâÊÇ»úÆ÷ѧϰÖеľ­µäÎÊÌ⣬Îı¾¡¢ÓïÒô¡¢Í¼ÏñµÈ¸÷ÖÖÁìÓò¶¼Óй㷺ӦÓá£ÄâÕë¶Ô¾ß±¸ºËÃܶȼ¼ÊõµÄ·ÖÀà·½·¨ÊµÏÖ»ùÓÚÄÚ´æµÄ·Ö²¼Ê½Ëã·¨¡£Ëù²Î¿¼µÄ·½·¨ÊÇDLR:Density-based Logistic Regression (Chen et al. KDD 2013)

ͼ3-18 ·ÖÀàË㷨ģ¿éͼ

£¨3£©¸ßά¶ÈÒì¹¹Êý¾ÝµÄ½µÎ¬Ëã·¨

Êý¾ÝºÍάÊýÖ®¼äÍùÍùÓÐÏà¹ØÐÔ£¬Óü¸¸ö½ÏÉÙµÄ×ÛºÏÖ¸±êÀ´´úÌæÔ­À´½Ï¶àµÄÖ¸±ê£¬¶øÕâЩ½ÏÉÙµÄ×ÛºÏÖ¸±ê¼ÈÄܾ¡¿ÉÄÜ¶àµØ·´Ó³Ô­À´½Ï¶àÖ¸±êµÄÓÐÓÃÐÅÏ¢£¬ÇÒÏ໥֮¼äÓÖÊÇÎ޹صġ£

50

ͼ3-19¸ßγ¶ÈÒì¹¹Êý¾ÝµÄ½µÎ¬Ëã·¨

3.7´óÊý¾Ý´æ´¢Éè¼Æ

3.7.1Êý¾Ý·Ö¼¶´æ´¢ £¨1£©Êý¾Ý·Ö¼¶´æ´¢Ô­Ôò

Êý¾ÝÉúÃüÖÜÆÚÖÐÔÚÏßÊý¾Ý¶Ô¸ßÐÔÄÜ´æ´¢µÄÐèÇó£¬ÒÔ¼°Ëæ×ÅÊý¾ÝÉúÃüÖÜÆÚµÄ±ä¸ü£¬Öð½¥ÏòÒ»°ãÐÔÄÜ´æ´¢µÄÇ¨ÒÆ£¬ÊÇ·Ö¼¶´æ´¢¹ÜÀíµÄÒ»ÌõÖ÷Ïß¡£Í¬Ê±¼æ¹Ë¿¼ÂÇÆäËû·Ö¼¶Ô­Ôò£¬¹²Í¬×÷ÓÃÓ°ÏìÊý¾ÝÇ¨ÒÆ»úÖÆ¡£

£¨2£©Êý¾ÝÈÚºÏÓë·Ö¼¶´æ´¢ÊµÊ©

½«ºËÐÄÄ£ÐÍ£¨¼´ÖжȻã×ܵÄÄ£ÐÍ£©Í¨¹ý¸ÄÔìÈÚÈëµ½ÏÖÓÐÖ÷Êý¾Ý²Ö¿âµÄºËÐÄÄ£ÐÍÖУ¬¼õÉÙÊý¾ÝÈßÓ࣬ÌáÉýÊý¾ÝÖÊÁ¿¡£

½«Ö÷Êý¾Ý²Ö¿âÖеÄÀúÊ·Êý¾ÝºÍÇåµ¥Êý¾ÝÇ¨ÒÆµ½µÍ³É±¾·Ö²¼Ê½Êý¾Ý¿â£¬¼õÇáÖ÷Êý¾Ý²Ö¿âµÄ¼ÆËãÓë´æ´¢Ñ¹Á¦²¢Ö§³ÅÉî¶ÈÊý¾Ý·ÖÎö¡£

51