{"id":570671,"date":"2025-05-27T21:02:11","date_gmt":"2025-05-27T20:02:11","guid":{"rendered":"https:\/\/blog.jetbrains.com\/?post_type=pycharm&p=570671"},"modified":"2025-05-27T21:31:29","modified_gmt":"2025-05-27T20:31:29","slug":"anomaly-detection-in-time-series","status":"publish","type":"pycharm","link":"https:\/\/blog.jetbrains.com\/ko\/pycharm\/2025\/05\/anomaly-detection-in-time-series\/","title":{"rendered":"Python\uc744 \ud65c\uc6a9\ud55c \uc2dc\uacc4\uc5f4 \uc774\uc0c1 \ud0d0\uc9c0"},"content":{"rendered":"
\"\"<\/figure>\n

\ub370\uc774\ud130\uc5d0\uc11c \uc911\uc694\ud55c \ubb38\uc81c\ub098 \uc228\uaca8\uc9c4 \uae30\ud68c\ub97c \ub4dc\ub7ec\ub0bc \uc218 \uc788\ub294 \ube44\uc815\uc0c1\uc801\uc778 \ud328\ud134\uc744 \uc5b4\ub5bb\uac8c \uc2dd\ubcc4\ud560 \uc218 \uc788\uc744\uae4c\uc694? \uc774\uc0c1 \ud0d0\uc9c0<\/a>\ub294 \uc815\uc0c1\uc5d0\uc11c \ubc97\uc5b4\ub09c \ub370\uc774\ud130\ub97c \uc2dd\ubcc4\ud558\ub294 \ub370 \ub3c4\uc6c0\uc774 \ub429\ub2c8\ub2e4. \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \uc2dc\uac04\uc758 \ucd94\uc774\uc5d0 \ub530\ub77c \uc218\uc9d1\ub41c \ub370\uc774\ud130\ub85c \uad6c\uc131\ub418\uc5b4 \uc788\uc73c\uba70, \ucd94\uc138\uc640 \uacc4\uc808\uc131 \ud328\ud134\uc774 \ud3ec\ud568\ub41c \uacbd\uc6b0\uac00 \ub9ce\uc2b5\ub2c8\ub2e4. \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc758 \uc774\uc0c1\uc740 \uc774\ub7ec\ud55c \ud328\ud134\uc774 \uae68\uc9c8 \ub54c \ubc1c\uc0dd\ud558\uae30\uc5d0 \uc774\uc0c1 \ud0d0\uc9c0\ub294 \uc601\uc5c5, \uae08\uc735, \uc81c\uc870, \uc758\ub8cc\uc640 \uac19\uc740 \uc0b0\uc5c5\uc5d0\uc11c \uc911\uc694\ud55c \ub3c4\uad6c\uac00 \ub429\ub2c8\ub2e4.<\/p>\n

\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \uacc4\uc808\uc131\uc774\ub098 \ucd94\uc138\uc640 \uac19\uc740 \uace0\uc720\ud55c \ud2b9\uc131\uc744 \uac00\uc9c0\uace0 \uc788\uae30 \ub54c\ubb38\uc5d0 \uc774\uc0c1\uce58\ub97c \ud6a8\uacfc\uc801\uc73c\ub85c \ud0d0\uc9c0\ud558\uae30 \uc704\ud574\uc11c\ub294 \uc804\ubb38\uc801\uc778 \ubc29\ubc95\uc774 \ud544\uc694\ud569\ub2c8\ub2e4. \uc774 \ube14\ub85c\uadf8 \uae00\uc5d0\uc11c\ub294 STL \ubd84\ud574\uc640 LSTM \uc608\uce21 \ub4f1 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc5d0\uc11c \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud558\ub294 \ub370 \ub110\ub9ac \uc0ac\uc6a9\ub418\ub294 \uba87 \uac00\uc9c0 \ubc29\ubc95\uc744 \ub2e4\ub8e8\uba70, \ucc98\uc74c \uc774\ud574\ud560 \ub54c \ub3c4\uc6c0\uc774 \ub420 \uc0c1\uc138\ud55c \ucf54\ub4dc \uc608\uc2dc\ub3c4 \ud568\uaed8 \uc81c\uacf5\ud569\ub2c8\ub2e4.<\/p>\n

\ube44\uc988\ub2c8\uc2a4\uc5d0\uc11c\uc758 \uc2dc\uacc4\uc5f4 \uc774\uc0c1 \ud0d0\uc9c0<\/h2>\n

\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \uc5ec\ub7ec \ube44\uc988\ub2c8\uc2a4 \ubc0f \uc11c\ube44\uc2a4\uc5d0\uc11c \ub9e4\uc6b0 \uc911\uc694\ud569\ub2c8\ub2e4. \ub9ce\uc740 \ube44\uc988\ub2c8\uc2a4\uc5d0\uc11c \uc2dc\uac04 \uacbd\uacfc\uc5d0 \ub530\ub978 \ub370\uc774\ud130\ub97c \ud0c0\uc784\uc2a4\ud0ec\ud504\uc640 \ud568\uaed8 \uae30\ub85d\ud558\uc5ec \uc2dc\uac04 \ucd94\uc774\uc5d0 \ub530\ub978 \ubcc0\ud654\ub97c \ubd84\uc11d\ud558\uace0 \ub370\uc774\ud130\ub97c \ube44\uad50\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc2dc\uacc4\uc5f4\uc740 \ud2b9\uc815 \uae30\uac04 \ub3d9\uc548 \ud2b9\uc815 \uc218\ub7c9\uc744 \ube44\uad50\ud560 \ub54c \uc720\uc6a9\ud569\ub2c8\ub2e4(\uc608: \ub370\uc774\ud130\uac00 \uacc4\uc808\uc801 \ud2b9\uc131\uc744 \ub098\ud0c0\ub0b4\ub294 \uc5f0\ub3c4\ubcc4 \ube44\uad50).<\/p>\n

\ub9e4\ucd9c \ubaa8\ub2c8\ud130\ub9c1<\/strong><\/p>\n

\uacc4\uc808\uc131\uc774 \uc788\ub294 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc758 \uac00\uc7a5 \uc77c\ubc18\uc801\uc778 \uc608\ub294 \ub9e4\ucd9c \ub370\uc774\ud130\uc785\ub2c8\ub2e4. \ub9e4\ucd9c\uc740 \uc5f0\uc911 \uacf5\ud734\uc77c\uacfc \uacc4\uc808\uc758 \uc601\ud5a5\uc744 \ub9ce\uc774 \ubc1b\uae30 \ub54c\ubb38\uc5d0 \uacc4\uc808\uc131\uc744 \uace0\ub824\ud558\uc9c0 \uc54a\uace0 \ub9e4\ucd9c \ub370\uc774\ud130\uc758 \uacb0\ub860\uc744 \ub3c4\ucd9c\ud558\uae30\ub780 \uc5b4\ub824\uc6b4 \uc77c\uc785\ub2c8\ub2e4. \uc774\ub7f0 \uc774\uc720\ub85c \ub9e4\ucd9c \ub370\uc774\ud130\uc5d0\uc11c \uc774\uc0c1\uce58\ub97c \ubd84\uc11d\ud558\uace0 \ubc1c\uacac\ud558\ub294 \uc77c\ubc18\uc801\uc778 \ubc29\ubc95\uc73c\ub85c STL \ubd84\ud574\uac00 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. \uc774\ub294 \uc774 \uae00\uc758 \ud6c4\ubc18\ubd80<\/a>\uc5d0\uc11c \uc790\uc138\ud788 \ub2e4\ub904 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n

\uae08\uc735<\/strong><\/p>\n

\uac70\ub798 \ubc0f \uc8fc\uac00\uc640 \uac19\uc740 \uae08\uc735 \ub370\uc774\ud130\ub294 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc758 \ub300\ud45c\uc801\uc778 \uc608\uc785\ub2c8\ub2e4. \uae08\uc735 \uc5c5\uacc4\uc5d0\uc11c\ub294 \uc774\ub7ec\ud55c \ub370\uc774\ud130\uc758 \uc774\uc0c1\uce58\ub97c \ubd84\uc11d\ud558\uace0 \ud0d0\uc9c0\ud558\ub294 \uac83\uc774 \uc77c\ubc18\uc801\uc778 \uad00\ud589\uc785\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4 \uc790\ub3d9 \uac70\ub798\uc5d0 \uc2dc\uacc4\uc5f4 \uc608\uce21 \ubaa8\ub378\uc744 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774 \uae00\uc758 \ud6c4\ubc18\ubd80<\/a>\uc5d0\uc11c \uc2dc\uacc4\uc5f4 \uc608\uce21\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc8fc\uc2dd \ub370\uc774\ud130\uc758 \uc774\uc0c1\uce58\ub97c \ud30c\uc545\ud558\ub294 \ubc29\ubc95\uc744 \uc0b4\ud3b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n

\uc81c\uc870<\/strong><\/p>\n

\uc2dc\uacc4\uc5f4 \uc774\uc0c1 \ud0d0\uc9c0\uc758 \ub610 \ub2e4\ub978 \uc0ac\uc6a9 \uc0ac\ub840\ub294 \uc0dd\uc0b0 \ub77c\uc778\uc758 \uacb0\ud568\uc744 \ubaa8\ub2c8\ud130\ub9c1\ud558\ub294 \uac83\uc785\ub2c8\ub2e4. \uae30\uacc4\uac00 \ubaa8\ub2c8\ud130 \uc5ed\ud560\uc744 \ud558\ub294 \uacbd\uc6b0\uac00 \ub9ce\uc544 \uc5ec\uae30\uc5d0\uc11c \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uac00 \uc81c\uacf5\ub418\uae30\ub3c4 \ud569\ub2c8\ub2e4. \uc7a0\uc7ac\uc801\uc778 \uc7a5\uc560\ub97c \uad00\ub9ac\uc790\uc5d0\uac8c \uc54c\ub9ac\ub294 \uac83\uc774 \uc911\uc694\ud558\uba70, \uc774\ub54c \uc774\uc0c1 \ud0d0\uc9c0\uac00 \ud575\uc2ec\uc801\uc778 \uc5ed\ud560\uc744 \ud569\ub2c8\ub2e4.<\/p>\n

\uc81c\uc57d \ubc0f \uc758\ub8cc<\/strong><\/p>\n

\uc81c\uc57d \ubc0f \uc758\ub8cc \uc5c5\uacc4\uc5d0\uc11c\ub294 \uc0ac\ub78c\uc758 \uc0dd\uccb4 \uc2e0\ud638\ub97c \ubaa8\ub2c8\ud130\ub9c1\ud558\uace0 \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\ub294 \uc758\ub8cc \uc5f0\uad6c\uc5d0\uc11c \uc911\uc694\ud558\uc9c0\ub9cc \uc9c4\ub2e8\uc5d0\uc11c\ub3c4 \ub9e4\uc6b0 \uc911\uc694\ud55c \ubb38\uc81c\uc785\ub2c8\ub2e4. \ubcd1\uc6d0\uc5d0 \uc788\ub294 \ud658\uc790\uc758 \ubc14\uc774\ud0c8 \uc0ac\uc778\uc5d0 \uc774\uc0c1\uc774 \ubc1c\uc0dd\ud588\uc744 \ub54c \uc989\uc2dc \uce58\ub8cc\uac00 \uc774\ub8e8\uc5b4\uc9c0\uc9c0 \uc54a\ub294\ub2e4\uba74, \uadf8 \uacb0\uacfc\ub294 \uce58\uba85\uc801\uc77c \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n

\uc2dc\uacc4\uc5f4 \uc774\uc0c1 \ud0d0\uc9c0\uc5d0 \ud2b9\uc218\ud55c \ubc29\ubc95\uc744 \uc0ac\uc6a9\ud574\uc57c \ud558\ub294 \uc774\uc720<\/h2>\n

\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \uacbd\uc6b0\uc5d0 \ub530\ub77c \uae30\ud0c0 \uc720\ud615\uc758 \ub370\uc774\ud130\ucc98\ub7fc \ucde8\uae09\ud560 \uc218 \uc5c6\ub2e4\ub294 \uc810\uc5d0\uc11c \ud2b9\ubcc4\ud558\ub2e4\uace0 \ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc5d0 \ud2b8\ub808\uc774\ub2dd \ud14c\uc2a4\ud2b8 \ubd84\ud560\uc744 \uc801\uc6a9\ud560 \ub54c, \ub370\uc774\ud130\uc758 \uc21c\ucc28\uc801\uc778 \uc5f0\uc18d\uc131 \ub54c\ubb38\uc5d0 \ub370\uc774\ud130\ub97c \uc11e\uc744 \uc218 \uc5c6\uc2b5\ub2c8\ub2e4. \uc774\ub294 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub97c \ub525\ub7ec\ub2dd \ubaa8\ub378\uc5d0 \uc801\uc6a9\ud560 \ub54c\ub3c4 \ub9c8\ucc2c\uac00\uc9c0\uc785\ub2c8\ub2e4. \uc21c\ucc28\uc801 \uad00\uacc4\ub97c \ubc18\uc601\ud558\uae30 \uc704\ud574 \uc21c\ud658 \uc2e0\uacbd\ub9dd(RNN)\uc774 \ud754\ud788 \uc0ac\uc6a9\ub418\uba70, \ud2b8\ub808\uc774\ub2dd \ub370\uc774\ud130\ub294 \uae30\uac04\uc73c\ub85c \uc785\ub825\ub418\uc5b4 \ud574\ub2f9 \uae30\uac04 \ub0b4\uc5d0\uc11c \uc774\ubca4\ud2b8\uc758 \uc21c\uc11c\uac00 \uc720\uc9c0\ub429\ub2c8\ub2e4.<\/p>\n

\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \ubb34\uc2dc\ud560 \uc218 \uc5c6\ub294 \uacc4\uc808\uc131\uacfc \ucd94\uc138\ub97c \uac00\uc9c0\uace0 \uc788\ub294 \uacbd\uc6b0\uac00 \ub9ce\uae30 \ub54c\ubb38\uc5d0 \ub354 \ud2b9\uc218\ud569\ub2c8\ub2e4. \uc774\ub7ec\ud55c \uacc4\uc808\uc131\uc740 24\uc2dc\uac04, 7\uc77c, 12\uac1c\uc6d4 \uc8fc\uae30 \ub4f1\uacfc \uac19\uc774 \uba87\uba87 \ub300\ud45c\uc801\uc778 \ud615\ud0dc\ub85c \ub098\ud0c0\ub0a0 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc544\ub798 \uc608\uc2dc<\/a>\uc5d0\uc11c \ubcfc \uc218 \uc788\ub4ef\uc774 \uacc4\uc808\uc131\uacfc \ucd94\uc138\ub97c \uace0\ub824\ud55c \ud6c4\uc5d0\uc57c \uc774\uc0c1\uce58\ub97c \ud310\ub2e8\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.\u00a0<\/p>\n

\uc2dc\uacc4\uc5f4\uc5d0\uc11c \uc774\uc0c1 \ud0d0\uc9c0\uc5d0 \uc0ac\uc6a9\ub418\ub294 \ubc29\ubc95<\/h2>\n

\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \ud2b9\uc218\ud558\uae30 \ub54c\ubb38\uc5d0 \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud558\uae30 \uc704\ud55c \ud2b9\uc815 \ubc29\ubc95\uc774 \uc788\uc2b5\ub2c8\ub2e4. \ub370\uc774\ud130 \uc720\ud615\uc5d0 \ub530\ub77c \uc774\uc804 \ube14\ub85c\uadf8 \uae00\uc5d0\uc11c \uc774\uc0c1 \ud0d0\uc9c0<\/a>\uc640 \uad00\ub828\ud558\uc5ec \uc5b8\uae09\ud55c \uba87 \uac00\uc9c0 \ubc29\ubc95\uacfc \uc54c\uace0\ub9ac\uc998\uc744 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc5d0 \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \uadf8\ub7ec\ud55c \ubc29\ubc95\uc740 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130 \uc804\uc6a9\uc73c\ub85c \uc124\uacc4\ub41c \uae30\ubc95\uc5d0 \ube44\ud574 \uc774\uc0c1 \ud0d0\uc9c0\uc758 \uc2e0\ub8b0\uc131\uc774 \ub2e4\uc18c \ub5a8\uc5b4\uc9c8 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uacbd\uc6b0\uc5d0 \ub530\ub77c \uc5ec\ub7ec \ud0d0\uc9c0 \ubc29\ubc95\uc744 \uc870\ud569\ud558\uc5ec \ud0d0\uc9c0 \uacb0\uacfc\ub97c \uc7ac\ud655\uc778\ud558\uba74 \uc704\uc591\uc131(FP) \ub610\ub294 \uc704\uc74c\uc131(FN)\uc744 \ubc29\uc9c0\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n

STL \ubd84\ud574<\/h3>\n

\uacc4\uc808\uc131\uc774 \uc788\ub294 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub97c \uc0ac\uc6a9\ud560 \ub54c \uac00\uc7a5 \uc778\uae30 \uc788\ub294 \ubc29\ubc95 \uc911 \ud558\ub098\ub294 STL \ubd84\ud574\ub85c, LOESS(\uad6d\uc9c0\uc801 \ucd94\uc815 \uc0b0\uc810\ub3c4 \ud3c9\ud65c\ud654)\ub97c \uc0ac\uc6a9\ud55c \uacc4\uc808\uc801 \ucd94\uc138 \ubd84\ud574\uc785\ub2c8\ub2e4. \uc774 \ubc29\ubc95\uc740 \uacc4\uc808\uc131 \ucd94\uc815\uce58(\uc54c\uace0\ub9ac\uc998\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc81c\uacf5\ub418\uac70\ub098 \uacb0\uc815\ub41c \uae30\uac04), \ucd94\uc138(\ucd94\uc815\uce58), \uc794\ucc28(\ub370\uc774\ud130\uc758 \ub178\uc774\uc988)\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc2dc\uacc4\uc5f4\uc744 \ubd84\ud574\ud569\ub2c8\ub2e4. STL \ubd84\ud574 \ub3c4\uad6c<\/a>\ub97c \uc81c\uacf5\ud558\ub294 Python<\/a> \ub77c\uc774\ube0c\ub7ec\ub9ac\ub294 statsmodels<\/a> \ub77c\uc774\ube0c\ub7ec\ub9ac\uc785\ub2c8\ub2e4.<\/p>\n

\"STL<\/figure>\n

\uc794\ucc28\uac00 \ud2b9\uc815 \uc784\uacd7\uac12\uc744 \ucd08\uacfc\ud558\uba74 \uc774\uc0c1\uc774 \ud0d0\uc9c0\ub429\ub2c8\ub2e4.\u00a0<\/p>\n

\ubc8c\uc9d1 \ub370\uc774\ud130\uc5d0 STL \ubd84\ud574 \uc0ac\uc6a9<\/h3>\n

\uc774\uc804 \ube14\ub85c\uadf8 \uae00<\/a>\uc5d0\uc11c\ub294 OneClassSVM<\/a>\uacfc IsolationForest<\/a> \uba54\uc11c\ub4dc\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubc8c\uc9d1\uc5d0\uc11c \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud558\ub294 \ubc29\ubc95\uc744 \uc0b4\ud3b4\ubd24\uc2b5\ub2c8\ub2e4.\u00a0<\/p>\n

\uc774 \ud29c\ud1a0\ub9ac\uc5bc\uc5d0\uc11c\ub294 statsmodels \ub77c\uc774\ube0c\ub7ec\ub9ac\uc758 STL<\/code> \ud074\ub798\uc2a4\ub97c \uc0ac\uc6a9\ud558\uc5ec \ubc8c\uc9d1 \ub370\uc774\ud130<\/a>\ub97c \uc2dc\uacc4\uc5f4\ub85c \ubd84\uc11d\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uc2dc\uc791\ud558\ub824\uba74 \ub2e4\uc74c \ud30c\uc77c(requirements.txt<\/a>)\uc744 \uc0ac\uc6a9\ud558\uc5ec \ud658\uacbd\uc744 \uc124\uc815\ud569\ub2c8\ub2e4.\u00a0<\/p>\n

1. \ub77c\uc774\ube0c\ub7ec\ub9ac \uc124\uce58<\/strong><\/h4>\n

\uc5ec\uae30\uc11c\ub294 Scikit-learn\uc5d0\uc11c \uc81c\uacf5\ud558\ub294 \ubaa8\ub378\ub9cc \uc0ac\uc6a9\ud558\ubbc0\ub85c PyPI\uc5d0\uc11c statsmodels\ub97c \uc124\uce58\ud574\uc57c \ud569\ub2c8\ub2e4. \uc774 \uc791\uc5c5\uc740 PyCharm<\/a>\uc5d0\uc11c \uc27d\uac8c \uc218\ud589\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n

\n
PyCharm Pro \ubb34\ub8cc\ub85c \uc2dc\uc791\ud558\uae30<\/a><\/div>\n<\/div>\n

Python <\/em>Package(\ud328\ud0a4\uc9c0)<\/em><\/a> \ucc3d(IDE \uc67c\ucabd \ud558\ub2e8\uc5d0 \uc788\ub294 \uc544\uc774\ucf58 \uc120\ud0dd)\uc73c\ub85c \uc774\ub3d9\ud558\uc5ec \uac80\uc0c9\ucc3d\uc5d0 statsmodels\ub97c \uc785\ub825\ud569\ub2c8\ub2e4.<\/p>\n

\"PyCharm\uc758<\/figure>\n

\uc6b0\uce21\uc5d0\uc11c \ud328\ud0a4\uc9c0\uc5d0 \ub300\ud55c \ubaa8\ub4e0 \uc815\ubcf4\ub97c \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc124\uce58\ud558\ub824\uba74 Install package(\ud328\ud0a4\uc9c0 \uc124\uce58)<\/em>\ub97c \ud074\ub9ad\ud558\uba74 \ub429\ub2c8\ub2e4.<\/p>\n

2. Jupyter Notebook \ub9cc\ub4e4\uae30<\/strong><\/h4>\n

\ub370\uc774\ud130 \uc138\ud2b8\ub97c \uc790\uc138\ud788 \uc870\uc0ac\ud558\uae30 \uc704\ud574 Jupyter Notebook<\/a>\uc744 \ub9cc\ub4e4\uc5b4 PyCharm\uc758 Jupyter Notebook \ud658\uacbd\uc774 \uc81c\uacf5\ud558\ub294 \ub3c4\uad6c\ub97c \ud65c\uc6a9\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n

\"PyCharm\uc5d0\uc11c<\/figure>\n

pandas<\/a>\ub97c \uac00\uc838\uc640\uc11c .csv<\/code> \ud30c\uc77c\uc744 \ub85c\ub4dc\ud569\ub2c8\ub2e4.<\/p>\n

import pandas as pd\n\ndf = pd.read_csv('..\/data\/Hive17.csv', sep=\";\")\ndf = df.dropna()\ndf<\/pre>\n
\"PyCharm\uc5d0\uc11c<\/figure>\n
3. \uadf8\ub798\ud504\ub85c \ub370\uc774\ud130 \uac80\uc0ac<\/strong><\/h4>\n
\uc774\uc81c \ub370\uc774\ud130\ub97c \uadf8\ub798\ud504\ub85c \uc0b4\ud3b4\ubcfc \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc5ec\uae30\uc11c\ub294 \uc2dc\uac04 \uacbd\uacfc\uc5d0 \ub530\ub978 hive 17\uc758 \uc628\ub3c4\ub97c \ubcf4\uace0\uc790 \ud569\ub2c8\ub2e4. DataFrame \uac80\uc0ac \ub3c4\uad6c\uc5d0\uc11c Chart view(\ucc28\ud2b8 \ubdf0)<\/em>\ub97c \ud074\ub9ad\ud55c \ub2e4\uc74c \uacc4\uc5f4 \uc124\uc815\uc5d0\uc11c T17<\/em>\uc744 y\ucd95\uc73c\ub85c \uc120\ud0dd\ud569\ub2c8\ub2e4.<\/p>\n
\"PyCharm\uc5d0\uc11c<\/figure>\n
\uc2dc\uacc4\uc5f4\ub85c \ud45c\ud604\ub41c \uc628\ub3c4 \ub370\uc774\ud130\ub294 \ub9ce\uc740 \ub4f1\ub77d\uc744 \ubcf4\uc785\ub2c8\ub2e4. \ub0ae\uacfc \ubc24\uc758 \uc8fc\uae30\uc640 \uac19\uc774, \uc8fc\uae30\uc801\uc778 \uc6c0\uc9c1\uc784\uc744 \ub098\ud0c0\ub0b4\ubbc0\ub85c \uc628\ub3c4\uc5d0 24\uc2dc\uac04 \uc8fc\uae30\uac00 \uc788\ub2e4\uace0 \uac00\uc815\ud574\ub3c4 \ubb34\ubc29\ud569\ub2c8\ub2e4.\u00a0<\/p>\n
\ub2e4\uc74c\uc73c\ub85c \uc2dc\uac04\uc5d0 \ub530\ub77c \uc628\ub3c4\uac00 \ub5a8\uc5b4\uc9c0\ub294 \ucd94\uc138\uac00 \uc788\uc2b5\ub2c8\ub2e4. DateTime<\/em> \uc5f4\uc744 \uc0b4\ud3b4\ubcf4\uba74 8\uc6d4\ubd80\ud130 11\uc6d4\uae4c\uc9c0 \ub0a0\uc9dc \ubc94\uc704\uac00 \uc788\uc74c\uc744 \uc54c \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ub370\uc774\ud130 \uc138\ud2b8\uc758 Kaggle \ud398\uc774\uc9c0<\/a>\ub97c \ubcf4\uba74 \ub370\uc774\ud130\uac00 \ud280\ub974\ud0a4\uc608\uc5d0\uc11c \uc218\uc9d1\ub418\uc5c8\uc73c\ubbc0\ub85c, \uc5ec\ub984\uc5d0\uc11c \uac00\uc744\ub85c\uc758 \uc804\ud658\uc774 \uc2dc\uac04\uc774 \uc9c0\ub0a8\uc5d0 \ub530\ub77c \uc628\ub3c4\uac00 \ub0b4\ub824\uac00\ub294 \ud604\uc0c1\uc744 \uc124\uba85\ud574 \uc900\ub2e4\ub294 \uac83\uc744 \uc54c \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
4. \uc2dc\uacc4\uc5f4 \ubd84\ud574<\/strong><\/h4>\n
\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub97c \uc774\ud574\ud558\uace0 \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud558\uae30 \uc704\ud574 STL \ubd84\ud574\ub97c \uc218\ud589\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uc774\ub97c \uc704\ud574 statsmodels\uc5d0\uc11c STL<\/code> \ud074\ub798\uc2a4\ub97c \ubd88\ub7ec\uc640 \uc628\ub3c4 \ub370\uc774\ud130\uc5d0 \uc801\uc6a9\ud558\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
from statsmodels.tsa.seasonal import STL\n\nstl = STL(df[\"T17\"], period=24, robust=True) \nresult = stl.fit()<\/pre>\n
\ubd84\ud574\uac00 \uc791\ub3d9\ud558\ub824\uba74 \uae30\uac04\uc744 \uc785\ub825\ud574\uc57c \ud569\ub2c8\ub2e4. \uc55e\uc11c \uc5b8\uae09\ud588\ub4ef\uc774, 24\uc2dc\uac04 \uc8fc\uae30\uac00 \uc788\ub2e4\uace0 \uac00\uc815\ud574\ub3c4 \ubb34\ubc29\ud569\ub2c8\ub2e4.<\/p>\n
\ubb38\uc11c\uc5d0 \ub530\ub974\uba74 STL<\/code>\uc740 \uc2dc\uacc4\uc5f4\uc744 \ucd94\uc138, \uacc4\uc808\uc131, \uc794\ucc28\uc758 \uc138 \uac00\uc9c0 \uad6c\uc131 \uc694\uc18c\ub85c \ubd84\ud574\ud569\ub2c8\ub2e4. \ubd84\ud574\ub41c \uacb0\uacfc\ub97c \ub354 \uba85\ud655\ud558\uac8c \ud30c\uc545\ud558\uae30 \uc704\ud574 \uae30\ubcf8 \uc81c\uacf5\ub41c plot<\/code> \uba54\uc11c\ub4dc\ub97c \uc0ac\uc6a9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
result.plot()<\/pre>\n
\"\uc2dc\uacc4\uc5f4<\/figure>\n
Trend(\ucd94\uc138)<\/em> \ubc0f Season(\uacc4\uc808\uc131)<\/em> \ud50c\ub86f\uc774 \uc704\uc758 \uac00\uc815\uacfc \uc798 \ub9de\uc544\ub5a8\uc5b4\uc9c0\ub294 \uac83\uc744 \ubcfc \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ud558\uc9c0\ub9cc, \uc6b0\ub9ac\uac00 \uc8fc\ubaa9\ud558\ub294 \uac83\uc740 \ub9e8 \uc544\ub798\uc5d0 \uc788\ub294 \uc794\ucc28 \ud50c\ub86f\uc73c\ub85c, \uc774\ub294 \uc6d0\ub798 \uc2dc\uacc4\uc5f4\uc5d0\uc11c \ucd94\uc138\uc640 \uacc4\uc808 \ubcc0\ub3d9\uc744 \uc81c\uac70\ud55c \uacb0\uacfc\uc785\ub2c8\ub2e4. \uc794\ucc28\uc5d0\uc11c \uadf9\ub2e8\uc801\uc73c\ub85c \ud06c\uac70\ub098 \uc791\uc740 \uac12\uc740 \uc774\uc0c1\uce58\ub97c \ub098\ud0c0\ub0c5\ub2c8\ub2e4.<\/p>\n
5. \uc774\uc0c1\uce58 \uc784\uacd7\uac12<\/strong><\/h4>\n
\ub2e4\uc74c\uc73c\ub85c, \uc794\ucc28 \uc911 \uc5b4\ub5a4 \uac12\uc744 \uc774\uc0c1\uce58\ub85c \uac04\uc8fc\ud560\uc9c0 \uacb0\uc815\ud574\uc57c \ud569\ub2c8\ub2e4. \uc774\ub97c \uc704\ud574 \uc794\ucc28\uc758 \ud788\uc2a4\ud1a0\uadf8\ub7a8\uc744 \uc0b4\ud3b4\ubcfc \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
result.resid.plot.hist()<\/pre>\n
\"PyCharm\uc5d0\uc11c<\/figure>\n
\uc774\ub294 0\uc744 \uc911\uc2ec\uc73c\ub85c \ud558\ub294 \uc815\uaddc \ubd84\ud3ec\ub85c \ubcfc \uc218 \uc788\uc73c\uba70, 5 \uc774\uc0c1\uacfc -5 \uc774\ud558\uc758 \ub871 \ud14c\uc77c\uc774 \ub098\ud0c0\ub098\ub294 \ud615\ud0dc\uc774\ubbc0\ub85c \uc784\uacd7\uac12\uc744 5\ub85c \uc124\uc815\ud558\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
\uc6d0\ub798\uc758 \uc2dc\uacc4\uc5f4\uc5d0 \uc774\uc0c1\uce58\ub97c \ud45c\uc2dc\ud558\uae30 \uc704\ud574 \uadf8\ub798\ud504\uc5d0\uc11c \ubaa8\ub4e0 \uc774\uc0c1\uce58\ub97c \ub2e4\uc74c\uacfc \uac19\uc774 \ube68\uac04\uc0c9\uc73c\ub85c \ud45c\uc2dc\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
import matplotlib.pyplot as plt\n\nthreshold = 5\nanomalies_filter = result.resid.apply(lambda x: True if abs(x) > threshold else False)\nanomalies = df[\"T17\"][anomalies_filter]\n\nplt.figure(figsize=(14, 8))\nplt.scatter(x=anomalies.index, y=anomalies, color=\"red\", label=\"anomalies\")\nplt.plot(df.index, df['T17'], color='blue')\nplt.title('Temperatures in Hive 17')\nplt.xlabel('Hours')\nplt.ylabel('Temperature')\nplt.legend()\nplt.show()<\/pre>\n
\"PyCharm\uc5d0\uc11c<\/figure>\n
STL \ubd84\ud574 \uc5c6\uc774 \uc8fc\uae30\uc640 \ucd94\uc138\uac00 \ud3ec\ud568\ub41c \uc2dc\uacc4\uc5f4\uc5d0\uc11c \uc774\ub7ec\ud55c \uc774\uc0c1\uce58\ub97c \uc2dd\ubcc4\ud558\uae30\ub294 \ub9e4\uc6b0 \uc5b4\ub835\uc2b5\ub2c8\ub2e4.<\/p>\n
LSTM \uc608\uce21<\/h3>\n
\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc5d0\uc11c \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud558\ub294 \ub610 \ub2e4\ub978 \ubc29\ubc95\uc740 \ub525\ub7ec\ub2dd \uae30\ubc95\uc744 \uc0ac\uc6a9\ud574 \ud574\ub2f9 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub97c \uc608\uce21\ud558\uace0, \uac01 \ub370\uc774\ud130 \ud3ec\uc778\ud2b8\uc758 \uc608\uc0c1 \uacb0\uacfc\ub97c \ucd94\uc815\ud558\ub294 \uac83\uc785\ub2c8\ub2e4. \ub9cc\uc57d \uc608\uc0c1\uac12\uacfc \uc2e4\uc81c \ub370\uc774\ud130 \ud3ec\uc778\ud2b8 \uc0ac\uc774\uc758 \ucc28\uc774\uac00 \ub9e4\uc6b0 \ud06c\ub2e4\uba74, \uc774\ub294 \uc774\uc0c1 \ub370\uc774\ud130\uc77c \uac00\ub2a5\uc131\uc774 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\uc21c\ucc28 \ub370\uc774\ud130 \uc608\uce21\uc744 \uc218\ud589\ud558\ub294 \ub300\ud45c\uc801\uc778 \ub525\ub7ec\ub2dd \uc54c\uace0\ub9ac\uc998 \uc911 \ud558\ub098\ub294 LSTM \ubaa8\ub378\uc774\uba70, \uc774\ub294 \uc21c\ud658 \uc2e0\uacbd\ub9dd(RNN)\uc758 \ud55c \uc885\ub958\uc785\ub2c8\ub2e4. LSTM \ubaa8\ub378\uc740 \uc785\ub825 \uac8c\uc774\ud2b8, \ub9dd\uac01 \uac8c\uc774\ud2b8, \ucd9c\ub825 \uac8c\uc774\ud2b8\ub85c \uad6c\uc131\ub418\uc5b4 \uc788\uc73c\uba70, \uc774\ub7ec\ud55c \uac8c\uc774\ud2b8\ub294 \ubaa8\ub450 \uc218\uce58 \ud589\ub82c\uc785\ub2c8\ub2e4. \uc774\ub97c \ud1b5\ud574 \uc911\uc694\ud55c \uc815\ubcf4\uac00 \ub2e4\uc74c \ub2e8\uacc4\uc758 \ub370\uc774\ud130\ub85c \uc804\ub2ec\ub420 \uc218 \uc788\ub3c4\ub85d \ud569\ub2c8\ub2e4.<\/p>\n
\"LSTM<\/figure>\n
\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \uc21c\ucc28\uc801 \ub370\uc774\ud130\ub85c, \ub370\uc774\ud130 \ud3ec\uc778\ud2b8\uc758 \uc21c\uc11c\uac00 \uc21c\ucc28\uc801\uc73c\ub85c \uc815\ud574\uc838 \uc788\uc73c\uba70 \uc784\uc758\ub85c \uc11e\uc5ec\uc11c\ub294 \uc548 \ub429\ub2c8\ub2e4. \uc774\ub7ec\ud55c \ud2b9\uc131 \ub54c\ubb38\uc5d0, \ud2b9\uc815 \uc2dc\uc810\uc758 \uacb0\uacfc\ub97c \uc608\uce21\ud558\ub294 \ub370\uc5d0\ub294 LSTM \ubaa8\ub378\uc774 \ud6a8\uacfc\uc801\uc778 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc785\ub2c8\ub2e4. \uc774\ub807\uac8c \uc608\uce21\ud55c \uac12\uc740 \uc2e4\uc81c \ub370\uc774\ud130\uc640 \ube44\uad50\ud560 \uc218 \uc788\uc73c\uba70, \ud2b9\uc815 \uc784\uacd7\uac12\uc744 \uc124\uc815\ud574 \uc2e4\uc81c \ub370\uc774\ud130\uac00 \uc774\uc0c1\uce58\uc778\uc9c0 \uc5ec\ubd80\ub97c \ud310\ub2e8\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\uc8fc\uac00 \ub370\uc774\ud130\uc5d0\uc11c LSTM \uc608\uce21 \ud65c\uc6a9<\/h3>\n
\uc774\uc81c \uc9c0\ub09c 5\ub144\uac04 Apple \uc8fc\uac00 \ub370\uc774\ud130\uc5d0\uc11c \uc774\uc0c1\uce58\ub97c \ud0d0\uc9c0\ud558\uae30 \uc704\ud55c \uc0c8\ub85c\uc6b4 Jupyter \ud504\ub85c\uc81d\ud2b8\ub97c \uc2dc\uc791\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4. \uc774 \uc8fc\uac00 \ub370\uc774\ud130 \uc138\ud2b8<\/a>\ub294 \ucd5c\uc2e0 \ub370\uc774\ud130\ub97c \ubcf4\uc5ec\uc90d\ub2c8\ub2e4. \uc774 \ube14\ub85c\uadf8 \uae00\uc758 \ub0b4\uc6a9\uc744 \ub530\ub77c \ud558\uace0 \uc2f6\ub2e4\uba74 \uc5ec\uae30\uc5d0\uc11c \uc0ac\uc6a9\ub418\ub294 \ub370\uc774\ud130 \uc138\ud2b8\ub97c \ub2e4\uc6b4\ub85c\ub4dc<\/a>\ud558\uba74 \ub429\ub2c8\ub2e4.<\/p>\n
1. Jupyter \ud504\ub85c\uc81d\ud2b8 \uc2dc\uc791<\/strong><\/h4>\n
\uc0c8 \ud504\ub85c\uc81d\ud2b8\ub97c \uc2dc\uc791\ud560 \ub54c \ub370\uc774\ud130 \uacfc\ud559\uc5d0 \ucd5c\uc801\ud654\ub41c Jupyter \ud504\ub85c\uc81d\ud2b8\ub97c \ub9cc\ub4e4\ub3c4\ub85d \uc120\ud0dd\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. New Project(\uc0c8 \ud504\ub85c\uc81d\ud2b8)<\/em> \ucc3d\uc5d0\uc11c Git \uc800\uc7a5\uc18c\ub97c \ub9cc\ub4e4\uace0 \ud658\uacbd \uad00\ub9ac\uc5d0 \uc0ac\uc6a9\ud560 conda \uc124\uce58\ub97c \uacb0\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\"PyCharm\uc5d0\uc11c<\/figure>\n
\ud504\ub85c\uc81d\ud2b8\ub97c \uc2dc\uc791\ud558\uba74 \uc608\uc2dc Notebook\uc774 \ud45c\uc2dc\ub429\ub2c8\ub2e4. \uc774 \uc5f0\uc2b5\uc744 \uc704\ud574 \uc0c8 Jupyter Notebook\uc744 \uc2dc\uc791\ud569\ub2c8\ub2e4.<\/p>\n
\"PyCharm\uc758<\/figure>\n
\uadf8\ub7f0 \ub2e4\uc74c requirements.txt<\/code>\ub97c \uc124\uc815\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4. pandas, matplotlib, \uadf8\ub9ac\uace0 PyPI\uc5d0\uc11c torch\ub77c\ub294 \uc774\ub984\uc73c\ub85c \uc81c\uacf5\ub418\ub294 PyTorch\uac00 \ud544\uc694\ud569\ub2c8\ub2e4. PyTorch\ub294 conda \ud658\uacbd\uc5d0 \ud3ec\ud568\ub418\uc5b4 \uc788\uc9c0 \uc54a\uc73c\ubbc0\ub85c PyCharm\uc5d0\uc11c \ud328\ud0a4\uc9c0\uac00 \ub204\ub77d\ub418\uc5c8\ub2e4\uace0 \uc54c\ub824\uc90d\ub2c8\ub2e4. \ud328\ud0a4\uc9c0\ub97c \uc124\uce58\ud558\ub824\uba74 \uc804\uad6c\ub97c \ud074\ub9ad\ud558\uace0 Install all missing packages(\ub204\ub77d\ub41c \ubaa8\ub4e0 \ud328\ud0a4\uc9c0 \uc124\uce58)<\/em>\ub97c \uc120\ud0dd\ud569\ub2c8\ub2e4.<\/p>\n
\"PyCharm\uc5d0\uc11c<\/figure>\n
2. \ub370\uc774\ud130 \ub85c\ub4dc \ubc0f \uac80\uc0ac<\/strong><\/h4>\n
\ub2e4\uc74c\uc73c\ub85c, \ub370\uc774\ud130 \ud3f4\ub354\uc5d0 apple_stock_5y.csv<\/a> \ub370\uc774\ud130 \uc138\ud2b8\ub97c \ub123\uace0 pandas DataFrame\uc73c\ub85c \ub85c\ub4dc\ud558\uc5ec \uac80\uc0ac\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
import pandas as pd\n \ndf = pd.read_csv('data\/apple_stock_5y.csv')\ndf<\/pre>\n
\ub300\ud654\ud615 \ud14c\uc774\ube14\uc744 \uc0ac\uc6a9\ud558\uba74 \ub204\ub77d\ub41c \ub370\uc774\ud130\uac00 \uc788\ub294\uc9c0 \uc27d\uac8c \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\"\"<\/figure>\n
\ub204\ub77d\ub41c \ub370\uc774\ud130\ub294 \uc5c6\uc9c0\ub9cc \ud55c \uac00\uc9c0 \ubb38\uc81c\uac00 \uc788\uc2b5\ub2c8\ub2e4. \uc885\uac00\/\ud604\uc7ac\uac00<\/em>\ub97c \uc0ac\uc6a9\ud558\uace0 \uc2f6\uc9c0\ub9cc \uc22b\uc790 \ub370\uc774\ud130 \uc720\ud615\uc774 \uc544\ub2d9\ub2c8\ub2e4. \ubcc0\ud658\uc744 \uc218\ud589\ud558\uace0 \ub370\uc774\ud130\ub97c \ub2e4\uc2dc \uac80\uc0ac\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
df[\"Close\/Last\"] = df[\"Close\/Last\"].apply(lambda x: float(x[1:]))\ndf<\/pre>\n
\uc774\uc81c \ub300\ud654\ud615 \ud14c\uc774\ube14\uc744 \ud1b5\ud574 \ud574\ub2f9 \uac00\uaca9\uc744 \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc67c\ucabd\uc758 \ud50c\ub86f \uc544\uc774\ucf58\uc744 \ud074\ub9ad\ud558\uba74 \ud50c\ub86f\uc774 \uc0dd\uc131\ub429\ub2c8\ub2e4. \uae30\ubcf8\uc801\uc73c\ub85c Date(\ub0a0\uc9dc)<\/em>\ub97c x\ucd95\uc73c\ub85c, Volume(\uac70\ub798\ub7c9)<\/em>\uc744 y\ucd95\uc73c\ub85c \uc0ac\uc6a9\ud569\ub2c8\ub2e4. Close\/Last(\uc885\uac00\/\ud604\uc7ac\uac00)<\/em>\ub97c \uac80\uc0ac\ud560 \uac83\uc774\ubbc0\ub85c, \uc624\ub978\ucabd\uc758 \ud1b1\ub2c8\ubc14\ud034 \uc544\uc774\ucf58\uc744 \ud074\ub9ad\ud558\uc5ec \uc124\uc815\uc73c\ub85c \uc774\ub3d9\ud55c \ub2e4\uc74c Close\/Last<\/em>\ub97c y\ucd95\uc73c\ub85c \uc120\ud0dd\ud569\ub2c8\ub2e4.<\/p>\n
\"\"<\/figure>\n
3. LSTM\uc744 \uc704\ud55c \ud2b8\ub808\uc774\ub2dd \ub370\uc774\ud130 \uc900\ube44<\/strong><\/h4>\n
\ub2e4\uc74c\uc73c\ub85c LSTM \ubaa8\ub378\uc5d0 \uc0ac\uc6a9\ud560 \ud2b8\ub808\uc774\ub2dd \ub370\uc774\ud130\ub97c \uc900\ube44\ud574\uc57c \ud569\ub2c8\ub2e4. \ub2e4\uc74c \uac00\uaca9\uc744 \uc608\uce21\ud558\uae30 \uc704\ud574, \uac01\uac01 \uae30\uac04\uc744 \ub098\ud0c0\ub0b4\ub294 \ubca1\ud130 \uc2dc\ud000\uc2a4(x\ucd95)\ub97c \uc900\ube44\ud574\uc57c \ud569\ub2c8\ub2e4. \ub2e4\uc74c \uac00\uaca9\uc740 \ub610 \ub2e4\ub978 \uc2dc\ud000\uc2a4(\ubaa9\ud45c\uac12 y)\ub97c \ud615\uc131\ud569\ub2c8\ub2e4. \uc5ec\uae30\uc11c lookback<\/code> \ubcc0\uc218\ub97c \uc0ac\uc6a9\ud574 \uc774 \uae30\uac04\uc758 \uae38\uc774\ub97c \uc120\ud0dd\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc544\ub798 \ucf54\ub4dc\ub294 \uc2dc\ud000\uc2a4 x\uc640 y\ub97c \uc0dd\uc131\ud558\uace0, \uc774\ud6c4 PyTorch tensor\ub85c \ubcc0\ud658\ub429\ub2c8\ub2e4.<\/p>\n
import torch\n\nlookback = 5\ntimeseries = df[[\"Close\/Last\"]].values.astype('float32')\n\nX, y = [], []\nfor i in range(len(timeseries)-lookback):\n    feature = timeseries[i:i+lookback]\n    target = timeseries[i+1:i+lookback+1]\n    X.append(feature)\n    y.append(target)\n    \nX = torch.tensor(X)\ny = torch.tensor(y)\n\nprint(X.shape, y.shape)<\/pre>\n
\uc77c\ubc18\uc801\uc73c\ub85c \uae30\uac04\uc774 \uae38\uc218\ub85d \uc785\ub825 \ubca1\ud130\uac00 \ud06c\uae30 \ub54c\ubb38\uc5d0 \ubaa8\ub378\ub3c4 \ucee4\uc9d1\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \uae30\uac04\uc774 \uae38\uc218\ub85d \uc785\ub825 \uc2dc\ud000\uc2a4\uac00 \uc9e7\uc544\uc9c0\ubbc0\ub85c \uc774 \uac80\ud1a0 \uae30\uac04\uc740 \uade0\ud615\uc744 \ub9de\ucdb0 \uacb0\uc815\ud574\uc57c \ud569\ub2c8\ub2e4. \uc5ec\uae30\uc11c\ub294 5\ub85c \uc2dc\uc791\ud558\uc9c0\ub9cc \ub2e4\ub978 \uac12\uc73c\ub85c \uc790\uc720\ub86d\uac8c \uc2dc\ub3c4\ud558\uc5ec \ucc28\uc774\uc810\uc744 \ud655\uc778\ud574 \ubcf4\uc138\uc694.<\/p>\n
4. \ubaa8\ub378 \ube4c\ub4dc \ubc0f \ud2b8\ub808\uc774\ub2dd<\/strong><\/h4>\n
\ubaa8\ub378\uc744 \ud2b8\ub808\uc774\ub2dd\ud558\uae30 \uc804\uc5d0 PyTorch\uc5d0\uc11c nn \ubaa8\ub4c8<\/a>\uc744 \uc0ac\uc6a9\ud558\uc5ec \ud074\ub798\uc2a4\ub97c \uc0dd\uc131\ud558\uace0 \ubaa8\ub378\uc744 \ube4c\ub4dc\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. nn \ubaa8\ub4c8\uc740 \ub2e4\uc591\ud55c \uc2e0\uacbd\ub9dd \ub808\uc774\uc5b4\uc640 \uac19\uc740 \uad6c\uc131 \uc694\uc18c\ub97c \uc81c\uacf5\ud569\ub2c8\ub2e4. \uc774 \uc5f0\uc2b5\uc5d0\uc11c\ub294 \uac04\ub2e8\ud55c LSTM \ub808\uc774\uc5b4<\/a>\uc640 \uc120\ud615 \ub808\uc774\uc5b4<\/a>\ub97c \ube4c\ub4dc\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
import torch.nn as nn\n\nclass StockModel(nn.Module):\n    def __init__(self):\n        super().__init__()\n        self.lstm = nn.LSTM(input_size=1, hidden_size=50, num_layers=1, batch_first=True)\n        self.linear = nn.Linear(50, 1)\n    def forward(self, x):\n        x, _ = self.lstm(x)\n        x = self.linear(x)\n        return x<\/pre>\n
\ub2e4\uc74c\uc73c\ub85c \ubaa8\ub378\uc744 \ud2b8\ub808\uc774\ub2dd\ud558\uaca0\uc2b5\ub2c8\ub2e4. \ubaa8\ub378\uc744 \ud2b8\ub808\uc774\ub2dd\ud558\uae30 \uc804\uc5d0 \uc635\ud2f0\ub9c8\uc774\uc800, \uc608\uce21\ub41c y \uac12\uacfc \uc2e4\uc81c y \uac12 \uc0ac\uc774\uc758 \uc190\uc2e4\uc744 \uacc4\uc0b0\ud558\ub294 \ub370 \uc0ac\uc6a9\ub418\ub294 \uc190\uc2e4 \ud568\uc218<\/a>, \uadf8\ub9ac\uace0 \ud559\uc2b5 \ub370\uc774\ud130\ub97c \uc785\ub825\ud560 \ub370\uc774\ud130 \ub85c\ub354<\/a>\ub97c \ub9cc\ub4e4\uc5b4\uc57c \ud569\ub2c8\ub2e4.<\/p>\n
import numpy as np\nimport torch.optim as optim\nimport torch.utils.data as data\n\nmodel = StockModel()\noptimizer = optim.Adam(model.parameters())\nloss_fn = nn.MSELoss()\nloader = data.DataLoader(data.TensorDataset(X, y), shuffle=True, batch_size=8)<\/pre>\n
\uc774\ubbf8 \uae30\uac04\uc774 \uc0dd\uc131\ub418\uc5b4 \uc788\uae30 \ub54c\ubb38\uc5d0 \ub370\uc774\ud130 \ub85c\ub354\uac00 \uc785\ub825\uc744 \uc11e\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\uc5d0 \ub530\ub77c \uac01 \uae30\uac04\uc5d0\uc11c \uc21c\ucc28\uc801 \uad00\uacc4\uac00 \uc720\uc9c0\ub429\ub2c8\ub2e4.<\/p>\n
\ud2b8\ub808\uc774\ub2dd\uc740 \uac01 \uc5d0\ud3ec\ud06c\ub97c \uc21c\ud68c\ud558\ub294 for<\/code> \ub8e8\ud504\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc218\ud589\ub429\ub2c8\ub2e4. 100 \uc5d0\ud3ec\ud06c\ub9c8\ub2e4 \uc190\uc2e4\uc744 \ucd9c\ub825\ud558\uace0 \ubaa8\ub378\uc774 \uc218\ub834\ud558\ub294 \uacfc\uc815\uc744 \uad00\ucc30\ud560 \uac83\uc785\ub2c8\ub2e4.<\/p>\n
n_epochs = 1000\nfor epoch in range(n_epochs):\n    model.train()\n    for X_batch, y_batch in loader:\n        y_pred = model(X_batch)\n        loss = loss_fn(y_pred, y_batch)\n        optimizer.zero_grad()\n        loss.backward()\n        optimizer.step()\n    if epoch % 100 != 0:\n        continue\n    model.eval()\n    with torch.no_grad():\n        y_pred = model(X)\n        rmse = np.sqrt(loss_fn(y_pred, y))\n    print(f\"Epoch {epoch}: RMSE {rmse:.4f}\")<\/pre>\n
1000 \uc5d0\ud3ec\ud06c\ub85c \uc2dc\uc791\ud588\uc9c0\ub9cc \ubaa8\ub378\uc740 \ube44\uad50\uc801 \ube60\ub974\uac8c \uc218\ub834\ud569\ub2c8\ub2e4. \ucd5c\uc801\uc758 \uacb0\uacfc\ub97c \uc5bb\uae30 \uc704\ud574 \ud2b8\ub808\uc774\ub2dd\uc6a9 \uc5d0\ud3ec\ud06c \uc218\ub294 \uc790\uc720\ub86d\uac8c \uc870\uc815\ud574 \ubcf4\uc154\ub3c4 \uc88b\uc2b5\ub2c8\ub2e4.<\/p>\n
\"\ud2b8\ub808\uc774\ub2dd\uc6a9<\/figure>\n
PyCharm\uc5d0\uc11c\ub294 \uc2e4\ud589\uc5d0 \uc2dc\uac04\uc774 \uac78\ub9ac\ub294 \uc140\uc758 \uacbd\uc6b0, \ub0a8\uc740 \uc2dc\uac04\uc744 \uc54c\ub824\uc8fc\ub294 \uc54c\ub9bc\uacfc \ud574\ub2f9 \uc140\ub85c \ubc14\ub85c \uc774\ub3d9\ud560 \uc218 \uc788\ub294 \ubc14\ub85c\uac00\uae30\ub97c \uc81c\uacf5\ud569\ub2c8\ub2e4. \uc774 \uae30\ub2a5\uc740 Jupyter Notebook\uc5d0\uc11c \uba38\uc2e0\ub7ec\ub2dd \ubaa8\ub378, \ud2b9\ud788 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc744 \ud2b8\ub808\uc774\ub2dd\ud560 \ub54c \ub9e4\uc6b0 \uc720\uc6a9\ud569\ub2c8\ub2e4.<\/p>\n
5. \uc608\uce21 \ud50c\ub86f \ubc0f \uc624\ub958 \ucc3e\uae30<\/strong><\/h4>\n
\ub2e4\uc74c\uc73c\ub85c \uc608\uce21\uc744 \uc0dd\uc131\ud558\uace0 \uc2e4\uc81c \uc2dc\uacc4\uc5f4\uacfc \ud568\uaed8 \uc774\ub97c \ud50c\ub86f\uc73c\ub85c \uc0dd\uc131\ud569\ub2c8\ub2e4. \uc774\ub54c \uc2e4\uc81c \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc640 \ud615\uc2dd\uc744 \ub9de\ucd94\uae30 \uc704\ud574 2D np \uacc4\uc5f4\uc744 \uc0dd\uc131\ud574\uc57c \ud569\ub2c8\ub2e4. \uc2e4\uc81c \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \ud30c\ub780\uc0c9, \uc608\uce21 \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \ube68\uac04\uc0c9\uc73c\ub85c \ud45c\uc2dc\ub429\ub2c8\ub2e4.<\/p>\n
import matplotlib.pyplot as plt\n\nwith torch.no_grad():\n    pred_series = np.ones_like(timeseries) * np.nan\n    pred_series[lookback:] = model(X)[:, -1, :]\n\nplt.plot(timeseries, c='b')\nplt.plot(pred_series, c='r')\nplt.show()<\/pre>\n
\"\uc608\uce21<\/figure>\n
\uc8fc\uc758 \uae4a\uac8c \uad00\ucc30\ud558\uba74 \uc608\uce21 \uac12\uacfc \uc2e4\uc81c \uac12\uc774 \uc644\ubcbd\ud558\uac8c \uc77c\uce58\ud558\uc9c0 \uc54a\ub294\ub2e4\ub294 \uac83\uc744 \uc54c \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uadf8\ub7ec\ub098 \ub300\ubd80\ubd84\uc758 \uc608\uce21\uc740 \ube44\uad50\uc801 \uc798 \ub9de\ub294 \ud3b8\uc785\ub2c8\ub2e4.<\/p>\n
\uc624\ucc28\ub97c \uba74\ubc00\ud788 \uac80\uc0ac\ud558\uae30 \uc704\ud574 \uc624\ucc28 \uc2dc\uacc4\uc5f4\uc744 \ub9cc\ub4e4\uace0 \ub300\ud654\ud615 \ud14c\uc774\ube14\uc744 \uc0ac\uc6a9\ud558\uc5ec \uad00\ucc30\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc5ec\uae30\uc11c\ub294 \uc808\ub300 \uc624\ucc28\ub97c \uc0ac\uc6a9\ud569\ub2c8\ub2e4.<\/p>\n
error = abs(timeseries-pred_series)\nerror<\/pre>\n
\uc124\uc815\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc808\ub300 \uc624\ucc28 \uac12\uc744 x\ucd95\uc73c\ub85c \ud558\uace0 \ud574\ub2f9 \uac12\uc758 \uac1c\uc218\ub97c y\ucd95\uc73c\ub85c \ud558\ub294 \ud788\uc2a4\ud1a0\uadf8\ub7a8\uc744 \ub9cc\ub4ed\ub2c8\ub2e4.<\/p>\n
\"\"<\/figure>\n
6. \uc774\uc0c1 \uc784\uacd7\uac12\uc744 \uacb0\uc815\ud558\uace0 \uc2dc\uac01\ud654<\/strong><\/h4>\n
\ub300\ubd80\ubd84\uc758 \ud3ec\uc778\ud2b8\ub294 6 \ubbf8\ub9cc\uc758 \uc808\ub300 \uc624\ucc28\ub97c \uac00\uc9c0\ubbc0\ub85c \uc774\ub97c \uc774\uc0c1\uce58 \uc784\uacd7\uac12\uc73c\ub85c \uc124\uc815\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ubc8c\uc9d1 \ub370\uc774\ud130\uc758 \uc774\uc0c1\uce58 \uc791\uc5c5<\/a>\uacfc \uc720\uc0ac\ud55c \ubc29\uc2dd\uc73c\ub85c \uadf8\ub798\ud504\uc5d0 \uc774\uc0c1 \ub370\uc774\ud130 \ud3ec\uc778\ud2b8\ub97c \ud50c\ub86f\uc73c\ub85c \uc0dd\uc131\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
threshold = 6\nerror_series = pd.Series(error.flatten())\nprice_series = pd.Series(timeseries.flatten())\n\nanomalies_filter = error_series.apply(lambda x: True if x > threshold else False)\nanomalies = price_series[anomalies_filter]\n\nplt.figure(figsize=(14, 8))\nplt.scatter(x=anomalies.index, y=anomalies, color=\"red\", label=\"anomalies\")\nplt.plot(df.index, timeseries, color='blue')\nplt.title('Closing price')\nplt.xlabel('Days')\nplt.ylabel('Price')\nplt.legend()\nplt.show()<\/pre>\n
\"\uadf8\ub798\ud504\uc5d0<\/figure>\n
\uc694\uc57d<\/h2>\n
\uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\ub294 \ube44\uc988\ub2c8\uc2a4 \ubc0f \uacfc\ud559 \uc5f0\uad6c\ub97c \ud3ec\ud568\ud55c \ub9ce\uc740 \ubd84\uc57c\uc5d0\uc11c \uc0ac\uc6a9\ub418\ub294 \uc77c\ubc18\uc801\uc778 \ud615\ud0dc\uc758 \ub370\uc774\ud130\uc785\ub2c8\ub2e4. \uc2dc\uacc4\uc5f4 \ub370\uc774\ud130\uc758 \uc21c\ucc28\uc801 \ud2b9\uc131\uc73c\ub85c \uc778\ud574 \uc774\uc0c1 \ud30c\uc545\uc744 \uc704\ud55c \ud2b9\ubcc4\ud55c \ubc29\ubc95\uacfc \uc54c\uace0\ub9ac\uc998\uc774 \uc0ac\uc6a9\ub429\ub2c8\ub2e4. \uc774 \ube14\ub85c\uadf8 \uae00\uc5d0\uc11c\ub294 \uacc4\uc808\uc131\uacfc \ucd94\uc138\ub97c \uc81c\uac70\ud558\uae30 \uc704\ud574 STL \ubd84\ud574\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc774\uc0c1\uce58\ub97c \uc2dd\ubcc4\ud558\ub294 \ubc29\ubc95\uc744 \ubcf4\uc5ec \ub4dc\ub838\uc2b5\ub2c8\ub2e4. \ub610\ud55c \ub525\ub7ec\ub2dd\uacfc LSTM \ubaa8\ub378\uc744 \ud65c\uc6a9\ud574 \uc608\uce21\ub41c \ucd94\uc815\uce58\uc640 \uc2e4\uc81c \ub370\uc774\ud130\ub97c \ube44\uad50\ud558\uc5ec \uc774\uc0c1\uce58\ub97c \ud30c\uc545\ud558\ub294 \ubc29\ubc95\ub3c4 \uc124\uba85\ud588\uc2b5\ub2c8\ub2e4.<\/p>\n
PyCharm\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc774\uc0c1 \ud0d0\uc9c0<\/h2>\n
PyCharm Professional\uc758 Jupyter \ud504\ub85c\uc81d\ud2b8\ub97c \uc0ac\uc6a9\ud558\uba74 \uc218\ub9ce\uc740 \ub370\uc774\ud130 \ud30c\uc77c\uacfc Notebook\uc73c\ub85c \uc774\uc0c1 \ud0d0\uc9c0 \ud504\ub85c\uc81d\ud2b8\ub97c \uc27d\uac8c \uad6c\uc131\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\uc0c1\uce58\ub97c \ud655\uc778\ud558\uae30 \uc704\ud55c \uadf8\ub798\ud504 \ucd9c\ub825\uc744 \uc0dd\uc131\ud560 \uc218 \uc788\uc73c\uba70, PyCharm\uc5d0\uc11c\ub294 \uc774\ub7ec\ud55c \ud50c\ub86f\uc744 \ub9e4\uc6b0 \uc27d\uac8c \ud655\uc778\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc790\ub3d9 \uc644\uc131 \uc81c\uc548\uacfc \uac19\uc740 \ub2e4\ub978 \uae30\ub2a5\uc744 \uc0ac\uc6a9\ud558\uba74 \ubaa8\ub4e0 Scikit-learn \ubaa8\ub378\uacfc Matplotlib \ud50c\ub86f \uc124\uc815\uc744 \uc544\uc8fc \uc27d\uac8c \ud0d0\uc0c9\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
PyCharm\uc744 \uc0ac\uc6a9\ud558\uc5ec \ub370\uc774\ud130 \uacfc\ud559 \ud504\ub85c\uc81d\ud2b8\ub97c \uac15\ud654\ud558\uace0 \ub370\uc774\ud130 \uacfc\ud559 \uc6cc\ud06c\ud50c\ub85c\ub97c \uac04\uc18c\ud654\ud558\uae30 \uc704\ud574 \uc81c\uacf5\ub418\ub294 \ub370\uc774\ud130 \uacfc\ud559 \uae30\ub2a5\ub3c4 \ud655\uc778<\/a>\ud574 \ubcf4\uc138\uc694.<\/p>\n
\n
PyCharm Pro \ubb34\ub8cc\ub85c \uc2dc\uc791\ud558\uae30<\/a><\/div>\n<\/div>\n
\u00a0<\/div>\n
\uac8c\uc2dc\ubb3c \uc6d0\ubb38 \uc791\uc131\uc790<\/em><\/div>\n\n    
\n        
\n            
\n                
\n                    \"Cheuk\n                <\/div>\n                
\n                                            
Cheuk Ting Ho<\/h4>\n                                                        <\/div>\n            <\/div>\n        <\/div>\n    <\/div>\n","protected":false},"author":1191,"featured_media":570725,"comment_status":"closed","ping_status":"closed","template":"","categories":[952,1401],"tags":[8670],"cross-post-tag":[],"acf":[],"_links":{"self":[{"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/pycharm\/570671"}],"collection":[{"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/pycharm"}],"about":[{"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/types\/pycharm"}],"author":[{"embeddable":true,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/users\/1191"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/comments?post=570671"}],"version-history":[{"count":9,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/pycharm\/570671\/revisions"}],"predecessor-version":[{"id":570753,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/pycharm\/570671\/revisions\/570753"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/media\/570725"}],"wp:attachment":[{"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/media?parent=570671"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/categories?post=570671"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/tags?post=570671"},{"taxonomy":"cross-post-tag","embeddable":true,"href":"https:\/\/blog.jetbrains.com\/ko\/wp-json\/wp\/v2\/cross-post-tag?post=570671"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}