SVM:网格搜索结果和我的结果不同，整个类(class)缺席和错误预测。为什么？-6ren

SVM:网格搜索结果和我的结果不同，整个类(class)缺席和错误预测。为什么？

转载作者：行者123 更新时间：2023-12-04 05:10:45

我有 155 张图片和 8 个类(class)
前提是特征不在 [0-1] 范围内缩放。

网格搜索交叉验证建议我使用线性内核和 C = 1000 的分数:

         precision    recall  f1-score   support

      1       0.54      0.88      0.67         8
      2       0.73      1.00      0.84         8
      3       1.00      1.00      1.00         6
      4       0.75      0.50      0.60        12
      5       0.83      0.83      0.83         6
      6       0.92      0.65      0.76        17
      7       0.71      0.42      0.53        12
      8       0.60      1.00      0.75         9

avg / total       0.77      0.73      0.72        78

但是当我尝试线性内核和 C=1000 时，我得到:

         precision    recall  f1-score   support

      1       0.00      0.00      0.00         0
      2       1.00      0.70      0.82        10
      3       1.00      1.00      1.00        13
      4       0.73      0.58      0.65        19
      5       1.00      0.95      0.97        19
      6       0.96      0.88      0.92        25
      7       0.82      0.67      0.73        27
      8       0.70      1.00      0.82        16

avg / total       0.88      0.81      0.84       129


Confusion matrix:
[[ 0  0  0  0  0  0  0  0]
[ 0  7  0  0  0  0  3  0]
[ 0  0 13  0  0  0  0  0]
[ 2  0  0 11  0  1  0  5]
[ 0  0  0  1 18  0  0  0]
[ 0  0  0  0  0 22  1  2]
[ 6  0  0  3  0  0 18  0]
[ 0  0  0  0  0  0  0 16]]

为什么第 1 类全为零？

我还看到使用 rbf 内核时我有最好的结果，但在头等舱中总是为零:

         precision    recall  f1-score   support

      1       0.00      0.00      0.00         0
      2       1.00      1.00      1.00        10
      3       1.00      1.00      1.00        13
      4       0.94      0.89      0.92        19
      5       1.00      0.95      0.97        19
      6       0.93      1.00      0.96        25
      7       1.00      0.78      0.88        27
      8       1.00      1.00      1.00        16

avg / total       0.98      0.93      0.95       129


Confusion matrix:
[[ 0  0  0  0  0  0  0  0]
 [ 0 10  0  0  0  0  0  0]
 [ 0  0 13  0  0  0  0  0]
 [ 1  0  0 17  0  1  0  0]
 [ 0  0  0  1 18  0  0  0]
 [ 0  0  0  0  0 25  0  0]
 [ 5  0  0  0  0  1 21  0]
 [ 0  0  0  0  0  0  0 16]]

最后，当我尝试预测训练集的一些相同图像时

print(clf.predict(fv))

其中 fv 是图像特征向量:

[0.16666666666628771, 5.169878828456423e-26, 2.3475644278196356e-21, 1.0, 1.0000000000027285]

并将错误的类分配给特征向量! (即图像拥有 4 个类别，但 predict() 结果是 5 个类别)

重新更新

图像集: https://docs.google.com/file/d/0ByS6Z5WRz-h2V3RkejFkb21Fb0E/edit?usp=sharing

特征图像集: https://docs.google.com/file/d/0ByS6Z5WRz-h2YlhuUmFBaElXVEE/edit?usp=sharing

完整代码:

import os
import glob
import numpy as np
from numpy import array
import cv2

target = [      1,1,1,1,
          1,1,1,1,1,1,1,
          1,1,1,1,1,1,1,
          1,2,2,2,2,2,2,
          2,2,2,2,2,2,2,
          2,2,2,2,3,3,3,
          3,3,3,3,3,3,3,
          3,3,3,4,4,4,4,
          4,4,4,4,4,4,4,
          4,4,4,4,4,4,4,        
          4,5,5,5,5,5,5,
          5,5,5,5,5,5,5,
          5,5,5,5,5,5,6,
          6,6,6,6,6,6,6,
          6,6,6,6,6,6,6,
          6,6,6,6,6,6,6,
          6,6,6,7,7,7,7,               
          7,7,7,7,7,7,7,
          7,7,7,7,7,7,7,
          7,7,7,7,7,7,7,                  
          7,7,8,8,8,8,8,
          8,8,8,8,8,8,8,
          8,8,8,8]



features = [ [0.26912666717306399, 0.012738398606387012, 0.011347858467581035, 0.1896938013442868, 2.444553429782046]
,
[0.36793086934925351, 0.034364344308391102, 0.019054536791551006, 0.0076875387476751395, 3.03091214703604]
,
[0.36793086934925351, 0.034364344308391102, 0.019054536791551006, 0.0076875387476751395, 3.03091214703604]
,
[0.30406240228443038, 0.047100329090555518, 0.0049653458889261448, 0.0004618404341300081, 5.987025009738751]
,
[0.36660353297714748, 0.034256126367653919, 0.01892501331178556, 0.007723901183105499, 3.0392760101225234]
,
[0.26708884220978957, 0.012126741224471632, 0.0063753119877062942, 0.0005937801528983894, 2.403113171408598]
,
[0.27070254516425241, 0.01293684867974746, 0.01159661796151442, 0.008380724334031727, 2.4492688425144986]
,
[0.27076540467770038, 0.012502407901054009, 0.011180048331833999, 0.0007116977225672878, 2.4068989750876266]
,
[0.22832314403919951, 0.010491475428909061, 0.0027317652016312271, 0.001417434443656981, 2.6271926274711968]
,
[0.22374814412737717, 0.0095258889624651646, 0.0040833924467236719, 0.1884906960716747, 2.5474055920602514]
,
[0.23860556210266026, 0.0067860933136106557, 0.0052050705189953389, 0.01498751040799334, 2.0545849084769694]
,
[0.32849751530034654, 0.0082079572128769367, 0.017950580842136479, 0.07211170619739862, 1.761646715256231]
,
[0.3536962871782694, 0.04335618127793292, 0.0084705562859388305, 0.003939815915497741, 3.8626463078353632]
,
[0.23642964900011443, 0.0060530993708264348, 0.0041172882106328976, 0.003276003276003276, 1.9809324414862304]
,
[0.35468301957048581, 0.043735489028639378, 0.0085420200506240735, 0.00041124057573680605, 3.873602628153773]
,
[0.35549112610207528, 0.043992218599656373, 0.0086354414147218166, 0.004276259969455286, 3.8781644572829106]
,
[0.97303451800669749, 0.075165987107118692, 0.23350656471824954, 0.04989418850724402, 1.7845923298199189]
,
[0.32292438991638828, 0.0078312712861588109, 0.018256154769458615, 0.05861489639723726, 1.754975905310628]
,
[0.36415716731096714, 0.033783635359516562, 0.0087048690616182353, 0.0007989674881691353, 3.0382507494699778]
,
[0.23247799686964493, 0.023970481957641395, 0.0020180739588722754, 0.2511737089201878, 4.987537342956105]
,
[0.25249755819322928, 0.03355835554037629, 0.0024745974458906918, 0.49168600154679043, 6.286228850887637]
,
[0.25524836990657951, 0.035216193154545015, 0.0023524820730296808, 0.49272798742138363, 6.553001816315555]
,
[0.25226043727172792, 0.033580607886770704, 0.002399474603048905, 0.4913428241631397, 6.310803986284148]
,
[0.2552359153348957, 0.034993472521483299, 0.0024465696242431606, 0.49311565696302123, 6.488164071764478]
,
[0.25249755819322928, 0.03355835554037629, 0.0024745974458906918, 0.49168600154679043, 6.286228850887637]
,
[0.19296658297366265, 0.0073667093687413854, 0.0010128002719554498, 0.20292887029288703, 2.6022382484976103]
,
[0.23130715659438109, 0.023652143308649062, 0.0020734509865596379, 0.2519981194170193, 4.96809084167716]
,
[0.23646940610897133, 0.025909457534721684, 0.0019634358569802723, 0.25097465886939574, 5.263654156113397]
,
[0.61892415483059771, 0.1855733578950316, 0.024118739298890277, 0.00010742003920831431, 5.579333799263049]
,
[0.61892415483059771, 0.1855733578950316, 0.024118739298890277, 0.00010742003920831431, 5.579333799263049]
,
[0.62187109165606835, 0.18810005977070685, 0.060143785970969831, 0.005752046658462197, 5.609811692923419]
,
[0.64410628333823972, 0.20178318336365086, 0.039546324622261202, 8.006565383614564e-05, 5.609490756132282]
,
[0.6214309265075304, 0.18779664186718673, 0.061337975720487534, 0.006350402281839464, 5.608301926807521]
,
[0.20135445416653119, 0.0070220507238874311, 0.0027092098815647042, 0.4125833006664053, 2.4256545571324732]
,
[0.20123494853445922, 0.0069845347246147793, 0.0027020357704780201, 0.4106724003127443, 2.420576584506546]
,
[0.2015816556223165, 0.0070631416111702362, 0.0025149608542164329, 0.4106073986851143, 2.4300340608128606]
,
[0.70115857527896985, 0.35625759453714789, 0.028386898853323388, 0.001234186979327368, 12.446918085552586]
,
[0.68366020888533297, 0.2387861974848598, 0.04047049559400958, 0.0725675987982436, 6.011803834536788]
,
[0.70115857527896985, 0.35625759453714789, 0.028386898853323388, 0.001234186979327368, 12.446918085552586]
,
[0.71378846605495283, 0.37185054375086962, 0.078338189105938844, 0.4899937460913071, 12.727628852581882]
,
[0.72219309919241148, 0.37567368174335658, 0.029371875736917675, 0.48066298342541436, 12.21840343375]
,
[0.84033907078880576, 0.29025638999406633, 0.090118665350957639, 0.00013319126265316994, 4.572824986179928]
,
[0.84033907078880576, 0.29025638999406633, 0.090118665350957639, 0.00013319126265316994, 4.572824986179928]
,
[0.84078478547550572, 0.28881268265635862, 0.092759120470064349, 0.0005334044539271903, 4.542932448095888]
,
[0.86195880470328134, 0.31149212664075476, 0.090341088591145105, 0.00044657097288676234, 4.673692966632184]
,
[0.85542893012496013, 0.29898764801731947, 0.17279563533793374, 0.0005314202205393915, 4.543371196521408]
,
[0.68653873117620423, 0.24135977292901584, 0.031609483792605572, 0.4553053169259345, 6.032229402405299]
,
[0.68937407444389065, 0.2429428175127194, 0.031783181019183315, 0.07118412046543464, 6.017180801429501]
,
[0.66262362984605561, 0.22830191525650573, 0.027222059698182095, 0.4712353884941554, 6.170703008647743]
,
[0.85191326598415906, 0.0066280315423251869, 0.18568977018064967, 0.24070082098793744, 1.211324246965761]
,
[0.41763663758743241, 0.0042550997098748248, 0.01052268995786553, 0.000998003992015968, 1.3702049090803978]
,
[0.47955540731641061, 0.036031336698149265, 0.0037552308556160824, 0.41911764705882354, 2.3102900509255964]
,
[0.28510645493450759, 0.017800467984914338, 0.0013560744373383752, 0.6212718064153067, 2.7591153064421485]
,
[0.28093855472961832, 0.017019535454492932, 0.0025233674347249074, 0.6243626062322947, 2.733908520445971]
,
[0.28510645493450759, 0.017800467984914338, 0.0013560744373383752, 0.6212718064153067, 2.7591153064421485]
,
[0.29957424000441979, 0.020997289413265056, 0.0032514165703168524, 0.002352941176470588, 2.8737257187232768]
,
[0.28093855472961832, 0.017019535454492932, 0.0025233674347249074, 0.6243626062322947, 2.733908520445971]
,
[0.94384505611284442, 0.0070361165614443756, 0.17778161251377933, 0.00013138014845956775, 1.1950816827585424]
,
[1.2480442396269933, 0.013169393067805945, 0.37414805554448649, 0.0018769272020378066, 1.202522486580245]
,
[0.82815785035628164, 0.0071847611802335776, 0.17226935935994725, 0.24680054800013365, 1.2280429227515923]
,
[0.55468014442636804, 0.04844726528488761, 0.074669093941655343, 0.3799483919692869, 2.3157520760049994]
,
[0.85603162865577076, 0.010190325204698992, 0.14635589096917062, 0.00018691588785046728, 1.2673797230628077]
,
[0.55881837183305305, 0.048068057730781634, 0.06639403930381195, 0.3722541921910773, 2.291289872230647]
,
[0.55650701031519434, 0.047379164870780005, 0.075834025272625227, 0.3768812839567851, 2.2847828255276856]
,
[0.59736941845983627, 0.054964632904472815, 0.089651232352172761, 0.0002190940461192967, 2.291980379225357]
,
[0.55468014442636804, 0.04844726528488761, 0.074669093941655343, 0.3799483919692869, 2.3157520760049994]
,
[0.37385965430511475, 0.019136318061858774, 0.0017515265254845647, 0.002456248081056187, 2.1746841721523915]
,
[0.3755068478409902, 0.019166948350188812, 0.0045621553498242356, 0.4868705591597158, 2.1680040687479902]
,
[0.376117657056177, 0.020048016077051325, 0.004081551918441755, 0.48440424204616345, 2.20746211913412]
,
[0.18567611209815035, 0.0017735326711233123, 0.00026719643703200545, 0.37649076434123163, 1.5866887090683386]
,
[0.15935887794419157, 3.0968737461516311e-05, 4.6106803792004044e-06, 7.109594397639615e-05, 1.0723690004464064]
,
[0.1598493732922015, 9.6513614204532248e-05, 1.4807540465080871e-05, 0.020011435105774727, 1.130966420539851]
,
[0.15976502679964721, 9.179670697435723e-05, 1.1098997372160861e-05, 0.027888446215139442, 1.127590980529105]
,
[0.15948519514589277, 8.8904788108173233e-05, 3.0493405326069049e-07, 0.825754804580883, 1.1256719774569757]
,
[0.16617638537179313, 0.0020240604885197228, 3.5948671354276501e-05, 0.00017182868679926113, 1.7424826840700272]
,
[0.16617882105231332, 0.002010285330985506, 3.1650697838912209e-05, 0.00017161489617298782, 1.7390017992958084]
,
[0.16601904246228144, 0.001959487143766989, 3.2733987503779933e-05, 0.10968404829180581, 1.7271461688896599]
,
[0.16628339469915165, 0.0020643314471593802, 1.4502279324313873e-05, 0.14276914653343373, 1.7519319117125625]
,
[0.16629298316796565, 0.0020800819965552542, 1.9020907349023509e-05, 0.13840607699240376, 1.755817053262183]
,
[0.18572210382333143, 0.0018178104959919194, 0.0002453722722107162, 6.292672183242613e-05, 1.5959450271122788]
,
[0.78164051870269824, 0.051523793666842309, 0.015067726988898911, 4.814636494944632e-05, 1.818489926889651]
,
[0.18566012446433577, 0.0017919804956179246, 0.00018368826559889194, 0.3746835841076679, 1.590696751465318]
,
[0.1593593872646801, 3.0965616570412022e-05, 4.7608077176119086e-06, 0.013757065159432655, 1.072364982247259]
,
[0.15935971192682988, 3.4228786893989237e-05, 2.8175989802780335e-06, 0.011385902663771647, 1.0762239433773122]
,
[0.1593758710624088, 3.1730097257658988e-05, 6.5545372607421827e-06, 0.19480358030830433, 1.0732774861268992]
,
[0.15935651884191823, 3.2075768916173883e-05, 2.6894443902692268e-06, 0.011169712144620248, 1.0736994974496823]
,
[0.1593593872646801, 3.0965616570412022e-05, 4.7608077176119086e-06, 0.013757065159432655, 1.072364982247259]
,
[0.72806364396184653, 0.080927033958709829, 0.082024727906757688, 0.0003304829181641674, 2.282620340759594]
,
[0.34064008340950969, 0.031713563937392303, 0.0223935905703848, 0.5525150905432595, 3.191021756804023]
,
[0.34161716425171257, 0.032414962195661444, 0.023399763826767502, 0.5634559735427863, 3.228573480379]
,
[0.33995795036914717, 0.032291160309302944, 0.014503695651070611, 0.5517519130084575, 3.2425659662137543]
,
[0.53755813910874839, 0.12514260672326116, 0.047097530510313457, 0.0022522522522522522, 4.849281676080233]
,
[0.53892887245870857, 0.12723100136939183, 0.047871070696486759, 0.0003630422944273008, 4.914680204854179]
,
[0.52941013268525083, 0.12033870626971493, 0.044950934295866135, 0.00036251586006887804, 4.801391369341545]
,
[0.5153795221866847, 0.11396653431855266, 0.046028411270117815, 0.0017374383209396067, 4.797613736965006]
,
[0.55889931613495802, 0.13776801275023373, 0.054206231614929122, 0.0003675794890645102, 4.954346523167349]
,
[0.53892887245870857, 0.12723100136939183, 0.047871070696486759, 0.0003630422944273008, 4.914680204854179]
,
[0.53876191407701801, 0.12675358533640296, 0.048092146277654686, 0.0003630422944273008, 4.896575690597256]
,
[0.64579700029686937, 0.053345962571719745, 0.047671705312373282, 0.00021581957483543757, 2.1135534993967275]
,
[0.52907834506993823, 0.11839951044942501, 0.046693278117526091, 0.001802451333813987, 4.720197357775248]
,
[0.62431811267333093, 0.16822847351832676, 0.078460359627903944, 0.0002954864445593558, 4.830349593161275]
,
[0.52957671831590236, 0.1206620716356978, 0.044424337085019652, 0.00036251586006887804, 4.812745400588476]
,
[0.64778861076667615, 0.011264454903514588, 0.26034582337509793, 0.00017355085039916696, 1.3918887090929497]
,
[0.64767923033014785, 0.011511416466409427, 0.26619423461723268, 0.0001713355606956224, 1.3970897837418754]
,
[0.64175254514795532, 0.051344373338613858, 0.047562712202626603, 0.0015838339705079192, 2.091594563276403]
,
[0.74328372556577627, 0.069102582620664751, 0.082952746646336797, 0.0001621665450417579, 2.094372254494601]
,
[0.63983023392719118, 0.050957609005336219, 0.04065234770126492, 0.0002180787264202377, 2.0902782497935077]
,
[0.64175254514795532, 0.051344373338613858, 0.047562712202626603, 0.0015838339705079192, 2.091594563276403]
,
[0.39929495902359424, 0.088487529110910193, 0.022225937358985204, 0.0016210739614994933, 6.842658946475011]
,
[0.40318161986196532, 0.091372930642081962, 0.029342259032521321, 0.0016383370878558263, 6.991543657993919]
,
[0.40286945787178563, 0.092489700223200605, 0.029477042699685527, 0.0008298755186721991, 7.159524821606994]
,
[0.401527045553835, 0.092940206887656154, 0.022384335964308343, 0.0008262755629002272, 7.307506331212089]
,
[0.48221520941584561, 0.080925707098030486, 0.01508266157389335, 0.016811768237766436, 3.877246216887803]
,
[0.23300739937344839, 0.0081726649803679097, 0.00070589920573164966, 0.7233009708737864, 2.267880404181219]
,
[0.4889793426754816, 0.13379642486830962, 0.0079207484968624713, 0.0012550988390335738, 6.938143452247703]
,
[0.50805268679046123, 0.15157146566770596, 0.002286367854475147, 0.0015261350629530714, 7.558054436128668]
,
[0.50504588069443601, 0.14372144884265609, 0.002332370870321935, 0.4888972525404592, 7.020435652999047]
,
[0.49053398407596349, 0.13596678236015974, 0.0068673835378752004, 0.5062523683213338, 7.054927383254023]
,
[0.27047698059047881, 0.02400759815979293, 0.0042725763257732184, 0.1406003159557662, 3.6822411994354223]
,
[0.67217292360607472, 0.21411359416298198, 0.038240138048085716, 0.00030014256771966684, 5.418493234141116]
,
[0.66809561834310183, 0.20843134771175456, 0.055569614057154701, 0.0005965697240865026, 5.316112363334643]
,
[0.69764902288163122, 0.23441611695166623, 0.040989861350760971, 0.00030097817908201655, 5.535854638867057]
,
[0.69337536416934831, 0.23122440548075349, 0.039932976305992858, 0.0011285832518245428, 5.5253522283788445]
,
[0.48053616103332131, 0.078827555080480394, 0.014699769292604886, 0.00040342914775592535, 3.810804845605404]
,
[0.51893243284454049, 0.14486098229876093, 0.007011404157031503, 0.0013995801259622112, 6.503015780005906]
,
[0.51611281879296478, 0.14397569681830566, 0.0063953861901166996, 0.0024067388688327317, 6.552602133840095]
,
[0.52265570318341037, 0.14786059553298658, 0.021856594872657918, 0.002438599547117227, 6.567632701584826]
,
[0.30079480228240624, 0.022512205511218238, 0.00042758792096778651, 0.016516516516516516, 2.990535008572801]
,
[0.30656959740479811, 0.025225633729599333, 0.00052074639660009423, 0.014692653673163419, 3.1500163953105362]
,
[0.36561931104389456, 0.034065616542602442, 0.00073193209081989026, 0.5295319844676067, 3.0388637406298646]
,
[0.30523253105219622, 0.024888851231432006, 0.00049965741600376489, 0.014692653673163419, 3.1395734571173244]
,
[0.30228106501925794, 0.02294279475480349, 0.00029015539686061685, 0.016315633343221597, 3.0087064225809246]
,
[0.48449572183350859, 0.08057148632400099, 0.014649379545360155, 0.0008072653884964682, 3.8293932305935345]
,
[0.48696620229608523, 0.082309882547938931, 0.015050994484143265, 0.0008004802881729037, 3.8679773897921153]
,
[0.28412339537248588, 0.026648939499942827, 0.0040253434951236042, 0.652089407191448, 3.7009800669447657]
,
[0.28496156479277329, 0.02656759057204762, 0.0040076364850396805, 0.6479146459747818, 3.672807600295908]
,
[0.27750673534987835, 0.024513847513161952, 0.0040536738369991365, 0.6610337972166997, 3.589249226795383]
,
[0.23076358836711391, 0.0081276558884353922, 0.0011346229787721842, 0.004830917874396135, 2.2823193871783753]
,
[0.23009954177415121, 0.0067688972295314211, 0.00050627342206410546, 0.7085308056872038, 2.1131083582556887]
,
[0.74667089537876641, 0.017808021782196932, 0.00058715813729321711, 0.20097746402389358, 1.4352297290097118]
,
[0.46459021914407012, 0.015923283050662724, 0.0096104956720461029, 0.07748745012228087, 1.7457829468097172]
,
[0.46915422400128481, 0.016432642899682152, 0.019842029469490614, 0.07962922414422305, 1.75192535048515]
,
[0.46603526212803831, 0.014906446836800192, 0.034027862564102791, 0.00017277871366247678, 1.709953995454109]
,
[0.74667089537876641, 0.017808021782196932, 0.00058715813729321711, 0.20097746402389358, 1.4352297290097118]
,
[0.74667089537876641, 0.017808021782196932, 0.00058715813729321711, 0.20097746402389358, 1.4352297290097118]
,
[0.79130699331490484, 0.018730652999112182, 0.0025843522081647448, 0.18977700753966478, 1.4182461597025509]
,
[0.78526444941147622, 0.019630664985282237, 0.0014735307445837577, 0.19151016964319956, 1.43434362458419]
,
[1.2360091274851985, 0.11319166323186233, 0.037129035449204553, 0.13274704929414488, 1.7480015935515811]
,
[1.2379748172284306, 0.11372770880048684, 0.03880647583352842, 0.1327272446632812, 1.748796779378963]
,
[1.0687065973690613, 0.06124884507730273, 1.0261941877753638, 0.0006237784339002786, 1.6027241684534652]
,
[1.0719786963564104, 0.066016997091209076, 0.67377325492164508, 0.000951317367746205, 1.6304903448777237]
,
[1.0726105544893461, 0.060421845064782209, 0.68185690192755832, 0.0016887717274899085, 1.5946007712754786]
,
[0.46459021914407012, 0.015923283050662724, 0.0096104956720461029, 0.07748745012228087, 1.7457829468097172]
,
[0.46578048886757484, 0.014968271641939066, 0.034797974069617273, 0.009791711402190251, 1.7124764457384176]
,
[0.47125303755595782, 0.015662651510987502, 0.0092152255656893708, 0.00017202081451855674, 1.723199095190638]
             ]




############################ PREDICTION TEST 1 IMAGE ################
print("TRY IMAGE")
import numpy as np
from sklearn import svm, metrics
X = features
y = target
from sklearn.svm import SVC
C = 1000.0
clf = svm.SVC(kernel='rbf', C=C).fit(X, y)
#svm.SVC(kernel='linear', C=C).fit(X, y) #SVC()
#clf.fit(X, y)
print("predizione")

#fv is class 8 but show me 5
fv = [0.16666666666628771, 5.169878828456423e-26, 2.584939414228212e-22, 1.0, 1.0000000000027285]
print(fv)
print(clf.predict([fv]))




############### METRICS ##########


# We learn the digits on the first half of the digits


# Now predict the value of the digit on the second half:
import matplotlib.pyplot as plt

expected = y[26:]
predicted = clf.predict(X[26:])
print("expected")
print(len(expected))
print("predicted")
print(len(predicted))

print "Classification report for classifier %s:\n%s\n" % (
    clf, metrics.classification_report(expected, predicted))
print "Confusion matrix:\n%s" % metrics.confusion_matrix(expected, predicted)

最佳答案

您在完整数据集上训练模型，然后计算训练集子集的分数，即数据集的所有末尾，除了 26 个第一个样本，其中包括来自第 0 类的整个样本集。

你不能这样评估模型:你需要随机打乱数据，然后在训练模型之前分割训练和测试集(否则整个数据集就是训练集，你没有单独的测试集)。如果你这样做:

import numpy as np
from sklearn import svm, metrics
from sklearn.cross_validation import train_test_split
from sklearn.svm import SVC

X = features
y = target

X_train, X_test, y_train, y_test = train_test_split(X, y,
        test_size=0.25, random_state=42)

C = 1000.0
clf = svm.SVC(kernel='rbf', C=C).fit(X_train, y_train)
y_predicted = clf.predict(X_test)

print "Classification report for classifier %s:\n%s\n" % (
    clf, metrics.classification_report(y_test, y_predicted))
print "Confusion matrix:\n%s" % metrics.confusion_matrix(y_test, y_predicted)

print "Predicting on 1 sample"
print "Input features:"
fv = [0.16666666666628771, 5.169878828456423e-26, 2.584939414228212e-22, 1.0, 1.0000000000027285]
print fv
print "Predicted class index:"
print clf.predict([fv])

您将获得以下输出:

Classification report for classifier SVC(C=1000.0, cache_size=200, class_weight=None, coef0=0.0, degree=3,
  gamma=0.0, kernel=rbf, max_iter=-1, probability=False, shrinking=True,
  tol=0.001, verbose=False):
             precision    recall  f1-score   support

          1       0.50      0.25      0.33         4
          2       0.75      1.00      0.86         6
          3       1.00      1.00      1.00         2
          4       0.75      1.00      0.86         3
          5       1.00      0.88      0.93         8
          6       1.00      1.00      1.00         5
          7       0.75      0.75      0.75         8
          8       1.00      1.00      1.00         3

avg / total       0.84      0.85      0.83        39


Confusion matrix:
[[1 1 0 0 0 0 2 0]
 [0 6 0 0 0 0 0 0]
 [0 0 2 0 0 0 0 0]
 [0 0 0 3 0 0 0 0]
 [0 0 0 1 7 0 0 0]
 [0 0 0 0 0 5 0 0]
 [1 1 0 0 0 0 6 0]
 [0 0 0 0 0 0 0 3]]
Predicting on 1 sample
Input features:
[0.1666666666662877, 5.169878828456423e-26, 2.584939414228212e-22, 1.0, 1.0000000000027285]
Predicted class index:
[5]

当然，这是一个单一的随机训练/测试拆分，并且由于您的数据集非常小，因此您获得的分数估计值会受到很大差异的影响。您可以通过迭代交叉验证计算此模型类和参数集的预期平均分数的估计值:

from sklearn.cross_validation import ShuffleSplit
from sklearn.cross_validation import cross_val_score
from scipy.stats import sem

params = dict(kernel='rbf', C=1000)
clf = svm.SVC(**params)
cv = ShuffleSplit(X.shape[0], n_iter=50)
cv_scores = cross_val_score(clf, X, y, cv=cv)

这将输出:

print "Cross Validated test scores for SVC with params {0} on full dataset:".format(params)
print "Mean: {0:.3} +/-{1:.3}".format(np.mean(cv_scores), sem(cv_scores))
print "Standard deviation: {0:.3}".format(np.std(cv_scores))

Cross Validated test scores for SVC with params {'kernel': 'rbf', 'C': 1000} on full dataset:
Mean: 0.834 +/-0.0125
Standard deviation: 0.0872

因此，您可以合理地期望总体上具有 83% 的预测准确度(或者更高一点，因为 CV 程序有点低估了)。

如果您想显着提高这种性能水平，我的第一个建议是收集更多标记的样本以获得更大的数据集。

第二个建议是通过对原始图像应用小扰动(例如小的平移、旋转和一点均匀随机噪声)从现有数据中生成更多标记数据，以便通过提取从现有数据中生成更多标记数据这些附加样本的特征。

编辑 :补充问题:

I also left out 8/10 image samples because I think they do not belong to any class.

您可能应该为不属于其他先前类的所有图像添加一个名为“other”的附加类别。

I should add a new class for each one and create new samples by small translations rotations?

不，目标是通过从现有样本中构建新样本来为每个类添加更多样本，从而提高现有类的分类精度。

i got this error: TypeError: init() got an unexpected keyword argument 'n_iter' at this line cv = ShuffleSplit(X.shape[0], n_iter=50)

n_iter 是 0.13 版本中的新名称。在 0.12 中是 n_iterations :

http://scikit-learn.org/0.12/modules/generated/sklearn.cross_validation.ShuffleSplit.html

关于SVM:网格搜索结果和我的结果不同，整个类(class)缺席和错误预测。为什么？，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/14959980/

文章推荐： spring - JDBC 在迭代 ResultSet 时丢失了连接

文章推荐： asp.net - ASP.net MVC中的自定义表单例份验证/授权方案

文章推荐： performance - R: tm Textmining 包:Doc-Level 元数据生成很慢

R 预测 - 如何仅绘制子集？
我正在使用 R 预测包拟合模型，如下所示: fit <- auto.arima(df) plot(forecast(fit,h=200)) 打印原始数据框和预测。当 df 相当大时，这
r - 预测-回归的神经网络预测相同的值
我正在尝试预测自有住房的中位数，这是一个行之有效的例子，给出了很好的结果。 https://heuristically.wordpress.com/2011/11/17/using-neural-ne
r - 预测()函数的类型参数
type="class"函数中的type="response"和predict有什么区别？例如： predict(modelName, newdata=testData, type = "class
python - 如何以图像的形式保存CNN模型的输出(预测)？
我有一个名为 Downloaded 的文件夹，其中包含经过训练的 CNN 模型必须对其进行预测的图像。下面是导入图片的代码: import os images = [] for filename i
区间内的 R 预测
关于预测的快速问题。我尝试预测的值是 0 或 1(它设置为数字，而不是因子)，因此当我运行随机森林时: fit , data=trainData, ntree=50) 并预测: pred, data
python - 预测，(找到正确的模型)
使用 Python，我尝试使用历史销售数据来预测产品的 future 销售数量。我还试图预测各组产品的这些计数。例如，我的专栏如下所示: Date Sales_count Department It
R SVM 预测
我是 R 新手，所以请帮助我了解问题所在。我试图预测一些数据，但预测函数返回的对象(这是奇怪的类(因子))包含低数据。测试集大小为 5886 obs。 160 个变量，当预测对象长度为 110 时..
java - 预测/识别电话号码的国家代码
关闭。这个问题需要更多focused .它目前不接受答案。想改进这个问题吗？更新问题，使其只关注一个问题 editing this post . 关闭 6 年前。 Improve this qu
python - 您如何从训练有素的网络对给定输入进行预测(预测)？
下面是我的神经网络代码，有 3 个输入和 1 个隐藏层和 1 个输出: #Data ds = SupervisedDataSet(3,1) myfile = open('my_file.csv','r
php - 预测/纠正全文搜索
我正在开发一个 Web 应用程序，它具有全文搜索功能，可以正常运行。我想对此进行改进并向其添加预测/更正功能，这意味着如果用户输入错误或结果为 0，则会查询该输入的更正版本，而不是查询结果。基本上类似
python - 具有单一分类特征的 LSTM 预测
我对时间序列还很陌生。这是我正在处理的数据集: Date Price Location 0 2012-01-01 1771.0
sequence - 如何使用隐马尔可夫模型进行 future 预测
我有许多可变长度的序列。对于这些，我想训练一个隐马尔可夫模型，稍后我想用它来预测(部分)序列的可能延续。到目前为止，我已经找到了两种使用 HMM 预测 future 的方法: 1) 幻觉延续并获得该延
映射到标签的 Tensorflow Serving 预测
我正在使用 TensorFlow 服务提供初始模型。我在 Azure Kubernetes 上这样做，所以不是通过更标准和有据可查的谷歌云。无论如何，这一切都在起作用，但是我感到困惑的是预测作为浮点
r - AWS 预测。项目数量的观察值太少
我正在尝试使用 Amazon Forecast 进行一些测试。我现在尝试了两个不同的数据集，它们看起来像这样: 13,2013-03-31 19:25:00,93.10999 14,2013-03-3
python - 预测 ufunc 输出的内存布局
使用 numpy ndarray大多数时候我们不需要担心内存布局的问题，因为结果并不依赖于它。除非他们这样做。例如，考虑这种设置 3x2 矩阵对角线的稍微过度设计的方法 >>> a = np.zer
R:如何在同一时间序列上绘制多个 ARIMA 预测
我想在同一个地 block 上用不同颜色绘制多个预测，但是，比例尺不对。我对任何其他方法持开放态度。可重现的例子: require(forecast) # MAKING DATA data
r - 通过分类变量和连续变量的交互可视化 GLMM 预测
我正在 R 中使用 GLMM，其中混合了连续变量和 calcategories 变量，并具有一些交互作用。我使用 MuMIn 中的 dredge 和 model.avg 函数来获取每个变量的效果估计。
output - 在命令行中导出 Weka 预测
我能够在 GUI 中成功导出分类器错误，但无法在命令行中执行此操作。有什么办法可以在命令行上完成此操作吗？我使用的是 Weka 3.6.x。在这里，您可以右键单击模型，选择“可视化分类器错误”并从那
R:如何在同一时间序列上绘制多个 ARIMA 预测
我想在同一个地 block 上用不同颜色绘制多个预测，但是，比例尺不对。我对任何其他方法持开放态度。可重现的例子: require(forecast) # MAKING DATA data
r - 预测 R 中的内存使用情况
我从 UCI 机器学习数据集库下载了一个巨大的文件。 (~300mb)。有没有办法在将数据集加载到 R 内存之前预测加载数据集所需的内存？ Google 搜索了很多，但我到处都能找到如何使用 R-p

行者123

个人简介

我是一名优秀的程序员,十分优秀！

作者热门文章

滴滴打车优惠券免费领取

全站热门文章

首页

博学

6Ren·AI

商城

SVM:网格搜索结果和我的结果不同，整个类(class)缺席和错误预测。为什么？