【发布时间】：2016-12-19 15:58:55
【问题描述】：
我正在尝试将可变长度字符串合并到 python 机器学习数据中。该字符串由 21 个可能的大写字符组成，长度从 3 到 1000 多个，但通常为 50 到 500 个字符。我想将此数据添加到现有的机器学习系统中，因为该字符串是系统中其余数字数据的派生值。我希望通过将这些信息整合到系统中，可以提高预测的准确性。
使用的机器学习系统：scikit-learn 的 SVR、xgboost 的梯度提升随机森林、使用 Theano 和 Keras 组合的神经网络。
示例数据（为清楚起见添加了空格，数千组之一）： 0.20783132530120485, 0.0, 0.14759036144578314, 0.0, 20.500779795353044, -0.012854043345111421, 20.856396736982024, -0.019526697858776032, 0.17055840352519377, MLKQLLTVVLLAICLINVQAQQLTPPAGTFRLGISKGTDSHWLAPQEKVKGIAFRWKALPDTRGFILEVAVTSLQQADTLFWSFGNCQPDMDINVFSVEGQAFTCYYGESMKLRTLQAVTPTDDIRLSNGRQDKTPLLLYESGKRTDRPVLAGRCPLAANSKLYFCFYEQNARADYNYFMLPDLFAKIDESKHSKK, 3907.222610216657, 0.0, 12.957234316695068, 260.35949614307845, 70.22897891511785, 0.0, 3600.1557026363694, 6.5695226674325005, 8.875805301569174E-9, 9.435201047407471E-8, -805.7695207777524, -0.386030775564303, 2.4360867449746193E-4, 0.001535275768898734, -899.103861896121, 0.37012002714844283, 41.30865237441297, 0.6880193813262029, 0.07901855928913903, 0.36786993202927, 0.027022889508663273, 0.20983595671723698, 0.004272043781893587, 2.6548618772402452, 0.8298948072745838, 0.4297709789614357, 0.6592421241850477, 0.7323455585665695, 0.0036084082526088635, 1235.9608595043105, -686.3410939120973, 517.5695296420 419, 0.0, 1383.9587599495007, 137.6709125154875, 48.15897140522527, 11.169320592630035, 0.0017212126730760488, 390.0, 576.0, 162.0, 425.0, -2337.586240324919, -1216.645095553551, -220.7658611143325, -254.87026759361316, -120.44151020211892, -262.1549293391522, -262.70857652215483, -119.78950303227985, -14.056523664351944, - 16.03338970562135, -15.397779250982714, -4.190420980506957, -52.306453723320466, -17.804935707496412, -1602.015046949609, -695.3200007491427, -282.2011792651323, -624.4938669353348, 319.12737432671895, -91.65456051126749, 190.69831510254096, 220.08361973544459, 2971.554863316476, 262.57174547648316, 2708.983117839995, 0.0, 5.482741129097017, -132.68200592716775, -4341.712499207881 , 9.524948063475861, 4.203276705216416, -4.307639899059003, 3.1644632985485313, 2.81419659034428, 2.963504627059134, 3.4913480163824713, 0.0031707417031467916, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0015698345827278798, 0.0016205522602160554, -1.9645139797143648E5, 0.9504047512545211, 0.98335286768 85283, 0.9597468652322548, 0.9865496952192033, 0.9175964036143727, 16312.662271951838, 15062.220268073073, 1250.4420038787648, 0.0, 2.7244897959183674, 0.0, 0.0, 0.0, 10.306122448979592, 0.0, 29.26530612244898, 0.0, 7.797822706065319, 0.0, 228.06859068818272, 0.4027714206386829, 1652.1493757294986, 3410.905281836304, 0.5612244897959183, 0.844845002268259, 0.5834395722203105, 1.0, 1.0, 1797.0, 362.119, 196.0, 1.0, -307.795, 0.000, -847.358, 202.875, -73.825, 2.064, 79.019, 452.437, -10.090, -45.351, -9.292, -36.652, 10.749, -38.050, 23.004, -18.505, 0.837, 0.344
[前9个字段（斜体）是衍生数据，评估其余的数值数据（可能是“Y”），下一个字段（粗体）是需要合并的字符串数据，其余的是机器学习的主要输入（“X”）]
【问题讨论】：
-
您需要找到一种方法来确保学习者获得的信息是固定长度的。即信息的特征化（词袋、k-means 等）