Question

我有两个列表，其中包含心跳间隔（Y轴，以毫秒为单位；下面的IBI）及其绝对时间点（X轴，以毫秒为单位；下面的RR_times）。存在一些误读，因此第一个列表包含需要删除的异常值，第二个列表包含它们的相应时间点。最好将第一个列表中的异常值设为NaN-ed，以使记录的总时间保持不变。

 RR_times = [411, 827, 1241,  1653,  2066,   2481,   2894,   3308,
     3714,   4126,   4532,   4938,   5343,   5751,   6156,   6552,
     6951,   7346,   7749,   8149,   8546,   8944,   9338,   9735,
    10123,  10511,  10905,  11290,  11675,  12060,  12441,  12825,
    13205,  13581,  13960,  14342,  14717,  15087,  15462,  15829,
    16204,  16531,  16902,  17304,  17670,  18040,  18398,  18762,
    19127,  19465,  19823,  20196,  20554,  20906,  21256,  21609,
    21959,  22264,  22637,  22995,  23308,  23649,  24012,  24352,
    24687,  25026,  25390,  25681,  26014,  26347,  26680,  27330,
    27985,  28628,  28951,  29596,  29915,  30238,  30562,  31191,
    31826,  32141,  32461,  32775,  33095,  33382,  33695,  34029,
    34341,  34654,  34967,  35281,  35595,  36220,  36530,  36844,
    37150,  37462,  37775,  38084,  38395,  38703,  39014,  39324,
    39632,  39937,  40246,  40554,  40862,  41169,  41479,  41787,
    42095,  42406,  42714,  43019,  43330,  43642,  43945,  44254,
    44563,  44871,  45183,  45491,  45796,  46101,  46410,  46713,
    47327,  47632,  47937,  48244,  48555,  48867,  49177,  49488,
    49792,  50094,  50398,  50707,  50993,  51324,  51626,  51931,
    52239,  52550,  52857,  53161,  53773,  54080,  54387,  54693,
    54998,  55311,  55617,  55924,  56235,  56547,  56852,  57159,
    57470,  57781,  58091,  58400,  58709,  59020,  59331,  59644,
    59955,  60265,  60579,  60890,  61206,  61521,  61833,  62149,
    62463,  62772,  63088,  63403,  63716,  64034,  64352,  64665,
    64984,  65624,  65940,  66262,  66578,  66900,  67221,  67543,
    67861,  68179,  68504,  68819,  69145,  69459,  69782,  70111,
    70428,  70747,  71070,  71389,  71710,  72036,  72358,  72680,
    73003,  73326,  73648,  73973,  74296,  74620,  74944,  75269,
    75592,  75916,  76241,  76566,  76889,  77216,  77543,  77869,
    78191,  78518,  78843,  79165,  79496,  79823,  80148,  80479,
    80803,  81128,  81459,  81783,  82110,  82439,  82771,  83095,
    83426,  83757,  84086,  84416,  84741,  85074,  85400,  85729,
    86060,  86390,  86719,  87051,  87380,  87711,  88041,  88373,
    88705,  89029,  89365,  89698,  90023,  90356,  90690,  91019,
    91352,  91684,  92014,  92347,  92681,  93014,  93349,  93678,
    94011,  94344,  94675,  95009,  95339,  95673,  96007,  96341,
    96668,  97002,  97337,  97665,  98003,  98335,  98668,  99003,
    99339,  99673, 100007, 100346, 100684, 101017, 101357, 101693,
   102028, 102368, 102705, 103043, 103380, 103718, 104061, 104403,
   104736, 105077, 105421, 105756, 106096, 106437, 106777, 107118,
   107461, 107800, 108141, 108485, 108822, 109167, 109507, 109848,
   110196, 110538, 110884, 111230, 111571, 111918, 112263, 112606,
   112952, 113639, 113987, 114336, 114680, 115025, 115372, 115722,
   116068, 116418, 116766, 117114, 117464, 117811, 118158, 118511,
   118858, 119208, 119557, 119904, 120257, 120606, 120952, 121303,
   121655, 122003, 122354, 122707, 123057, 123408, 123760, 124114,
   124466, 124815, 125172, 125523, 125879, 126231, 126586, 126946,
   127298, 127653, 128014, 128369, 128724, 129084, 129441, 129794,
   130150, 130504, 130863, 131219, 131576, 131937, 132297, 132653,
   133012, 133375, 133731, 134091, 134455, 134813, 135174, 135534,
   135897, 136258, 136621, 136986, 137349, 137711, 138073, 138439,
   138799, 139164, 139526, 139887, 140253, 140617, 140977, 141344,
   141706, 142071, 142438, 142803, 143170, 143537, 143904, 144274,
   144641, 145011, 145382, 145749, 146124, 146493, 146864, 147235,
   147605, 147977, 148346, 148718, 149085, 149455, 149826, 150195,
   150566, 150936, 151310, 151676, 152048, 152423, 152795, 153167,
   153539, 153916, 154290, 154661, 155036, 155408, 155782, 156159,
   156530, 156905, 157280, 157655, 158029, 158404, 158783, 159157,
   159532, 159910, 160290, 160660, 161037, 161415, 161786, 162161,
   162538, 162913, 163289, 163665, 164040, 164415, 164789, 165164,
   165539, 165911, 166286, 166661, 167040, 167418, 167791, 168169,
   168545, 168922, 169300, 169676, 170053, 170429, 170811, 171195,
   171571, 171952, 172335, 172717, 173098, 173484, 173869, 174254,
   174637, 175020, 175403, 175785, 176167, 176552, 176933, 177316,
   177698, 178080, 178463, 178840, 179224, 179603, 179979, 180360,
   180739, 181114, 181492, 181870, 182248, 182626, 183001, 183378,
   183752, 184128, 184503, 184876, 185252, 185629, 186003, 186384,
   186760, 187134, 187515, 187900, 188281, 188656, 189031, 189415,
   189798, 190176, 190555, 190936, 191313, 191692, 192069, 192448,
   192824, 193203, 193578, 193953, 194330, 194707]

IBIs = [411,416,414,412,413,415, 413, 414, 406, 412, 406, 406, 405,
   408, 405, 396, 399, 395, 403, 400, 397, 398, 394, 397, 388, 388,
   394, 385, 385, 385, 381, 384, 380, 376, 379, 382, 375, 370, 375,
   367, 375, 327, 371, 402, 366, 370, 358, 364, 365, 338, 358, 373,
   358, 352, 350, 353, 350, 305, 373, 358, 313, 341, 363, 340, 335,
   339, 364, 291, 333, 333, 333, 650, 655, 643, 323, 645, 319, 323,
   324, 629, 635, 315, 320, 314, 320, 287, 313, 334, 312, 313, 313,
   314, 314, 625, 310, 314, 306, 312, 313, 309, 311, 308, 311, 310,
   308, 305, 309, 308, 308, 307, 310, 308, 308, 311, 308, 305, 311,
   312, 303, 309, 309, 308, 312, 308, 305, 305, 309, 303, 614, 305,
   305, 307, 311, 312, 310, 311, 304, 302, 304, 309, 286, 331, 302,
   305, 308, 311, 307, 304, 612, 307, 307, 306, 305, 313, 306, 307,
   311, 312, 305, 307, 311, 311, 310, 309, 309, 311, 311, 313, 311,
   310, 314, 311, 316, 315, 312, 316, 314, 309, 316, 315, 313, 318,
   318, 313, 319, 640, 316, 322, 316, 322, 321, 322, 318, 318, 325,
   315, 326, 314, 323, 329, 317, 319, 323, 319, 321, 326, 322, 322,
   323, 323, 322, 325, 323, 324, 324, 325, 323, 324, 325, 325, 323,
   327, 327, 326, 322, 327, 325, 322, 331, 327, 325, 331, 324, 325,
   331, 324, 327, 329, 332, 324, 331, 331, 329, 330, 325, 333, 326,
   329, 331, 330, 329, 332, 329, 331, 330, 332, 332, 324, 336, 333,
   325, 333, 334, 329, 333, 332, 330, 333, 334, 333, 335, 329, 333,
   333, 331, 334, 330, 334, 334, 334, 327, 334, 335, 328, 338, 332,
   333, 335, 336, 334, 334, 339, 338, 333, 340, 336, 335, 340, 337,
   338, 337, 338, 343, 342, 333, 341, 344, 335, 340, 341, 340, 341,
   343, 339, 341, 344, 337, 345, 340, 341, 348, 342, 346, 346, 341,
   347, 345, 343, 346, 687, 348, 349, 344, 345, 347, 350, 346, 350,
   348, 348, 350, 347, 347, 353, 347, 350, 349, 347, 353, 349, 346,
   351, 352, 348, 351, 353, 350, 351, 352, 354, 352, 349, 357, 351,
   356, 352, 355, 360, 352, 355, 361, 355, 355, 360, 357, 353, 356,
   354, 359, 356, 357, 361, 360, 356, 359, 363, 356, 360, 364, 358,
   361, 360, 363, 361, 363, 365, 363, 362, 362, 366, 360, 365, 362,
   361, 366, 364, 360, 367, 362, 365, 367, 365, 367, 367, 367, 370,
   367, 370, 371, 367, 375, 369, 371, 371, 370, 372, 369, 372, 367,
   370, 371, 369, 371, 370, 374, 366, 372, 375, 372, 372, 372, 377,
   374, 371, 375, 372, 374, 377, 371, 375, 375, 375, 374, 375, 379,
   374, 375, 378, 380, 370, 377, 378, 371, 375, 377, 375, 376, 376,
   375, 375, 374, 375, 375, 372, 375, 375, 379, 378, 373, 378, 376,
   377, 378, 376, 377, 376, 382, 384, 376, 381, 383, 382, 381, 386,
   385, 385, 383, 383, 383, 382, 382, 385, 381, 383, 382, 382, 383,
   377, 384, 379, 376, 381, 379, 375, 378, 378, 378, 378, 375, 377,
   374, 376, 375, 373, 376, 377, 374, 381, 376, 374, 381, 385, 381,
   375, 375, 384, 383, 378, 379, 381, 377, 379, 377, 379, 376, 379,
   375, 375, 377, 377]

绘制整个数据集可得出：

我以前使用过以下过滤器，但不适用于较长的记录，在该记录中，迹线跨越较大的值（在某些记录中，间隔从300（训练过程中）到1500（经过一段休息时间））。

在这种情况下，消除异常值的最佳方法是什么？如何实施它？移动平均，基于stdev的排除，中值过滤器...？

Answer 1

这是一个似乎可行的丑陋方法：

import numpy as np
RR_times = np.array([411, 827, 1241, ...])
IBIs = np.array([411, 416, 414, ...])
diffs = [np.abs(IBIs[i]-IBIs[i+1]) for i in range(len(IBIs)-1)]
IBIs_cleaned = np.full(IBIs.shape, np.nan) # create a array full of NaNs
IBIs_cleaned[0] = IBIs[0] # The first value isn't a outlier

for i in range(1, len(IBIs)):
    if np.abs(IBIs[i]-IBIs[i-1]) < np.mean(diffs) and IBIs[i] < 1.6 * np.mean(IBIs):
        IBIs_cleaned[i] = IBIs[i]

从列表/ XY散点中删除异常值

1 个答案: