scikit-learn特征降维

于 2024-08-20 16:39:31 发布
阅读量735 收藏 15
点赞数 22
分类专栏： python 文章标签： scikit-learn python 机器学习
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/m0_52606060/article/details/141340061
版权

为什么要特征降维

特征降维是指将数据集中的特征数（或称为维度）降低的过程，这个过程有助于简化数据结构，同时去除冗余特征和噪声，提高模型的运行效率和预测能力。其优点主要包括以下几点：

  • 减少计算量：降低特征数量会减少模型训练时所需的计算资源，特别是对于高维数据集来说，这可以显著减少训练时间。

  • 防止过拟合：通过去除一些不重要或者冗余的特征，可以减少模型对训练数据的过度依赖，从而提高模型在新数据上的泛化能力。

  • 数据可视化：在高维空间中，数据的可视化非常困难。通过特征降维，可以将数据投影到二维或三维空间中，使得数据的可视化和分析变得更加直观。

  • 去除噪声：高维数据往往包含较多的噪声，特征降维可以帮助去除一些无关紧要的特征，留下对模型最有帮助的信息。

  • 提高模型性能：一些降维技术（如线性判别分析LDA）能够通过寻找最佳的特征组合来提升分类或回归任务的性能。

降维

降维是指在某些限定条件下，降低随机变量（特征）个数，得到一组“不相关”主变量的过程

降维的两种方式

  • 特征选择
  • 主成分分析（可以理解一种特征提取的方式）

什么是特征选择

定义

数据中包含冗余或相关变量（或称特征、属性、指标等），旨在从原有特征中找出主要特征
在这里插入图片描述

方法

Filter过滤式：主要探究特征本身特点、特征与特征和目标值之间关联

  • 方差选择法：低方差特征过滤
  • 相关系数：特征与特征之间的相关程度

Embedded嵌入式：算法自动选择特征（特征与目标值之间的关联）

  • 决策树：信息熵、信息增益
  • 正则化：L1，L2
  • 深度学习：卷积等

模块

sklearn.feature_selection

测试数据

index,pe_ratio,pb_ratio,market_cap,return_on_asset_net_profit,du_return_on_equity,ev,earnings_per_share,revenue,total_expense,date,return
0,000001.XSHE,5.9572,1.1818,85252550922.0,0.8008,14.9403,1211444855670.0,2.01,20701401000.0,10882540000.0,2012-01-31,0.027657228229937388
1,000002.XSHE,7.0289,1.588,84113358168.0,1.6463,7.8656,300252061695.0,0.326,29308369223.2,23783476901.2,2012-01-31,0.08235182370820669
2,000008.XSHE,-262.7461,7.0003,517045520.0,-0.5678,-0.5943,770517752.56,-0.006,11679829.03,12030080.04,2012-01-31,0.09978900335112327
3,000060.XSHE,16.476,3.7146,19680455995.0,5.6036,14.617,28009159184.6,0.35,9189386877.65,7935542726.05,2012-01-31,0.12159482758620697
4,000069.XSHE,12.5878,2.5616,41727214853.0,2.8729,10.9097,81247380359.0,0.271,8951453490.28,7091397989.13,2012-01-31,-0.0026808154146886697
5,000100.XSHE,10.796,1.5219999999999998,17206724233.0,2.245,7.7394,66034033386.1,0.0974,43883757748.0,43092263405.0,2012-01-31,0.13795588072275808
6,000402.XSHE,8.1032,1.0078,18253291248.0,3.2233,10.3965,56511787330.1,0.6,7303544153.27,5440677926.67,2012-01-31,0.08457869004968227
7,000413.XSHE,192.5234,18.3692,4270450000.0,-0.6423,-1.0375,4363236328.83,-0.0059,62630800.67,64784767.7,2012-01-31,0.11926833305744075
8,000415.XSHE,18.0444,1.2301,8699264600.0,1.8112,3.7845,21846364582.5,0.3197,1466188261.63,1036862316.81,2012-01-31,0.188558875748548
9,000423.XSHE,23.3689,7.7135,26409389664.0,16.1098,20.7633,27897612682.3,0.9372,1862915702.43,1157644109.36,2012-01-31,0.08766839826138167
10,000425.XSHE,8.6344,2.1796,33004130464.0,9.9592,21.1151,53846101659.4,1.37,25233044684.7,21955101570.5,2012-01-31,0.0024943086212016125
11,000503.XSHE,226.9945,4.3887,5410909668.0,0.1796,0.3275,6249917784.47,0.0045,99694183.97,135382267.29,2012-01-31,0.16777408637873767
12,000538.XSHE,22.6896,6.1087,33935745494.0,10.2038,17.8649,43071657562.5,1.24,7741652400.02,6769161248.85,2012-01-31,0.0654679840686276
13,000540.XSHE,17.6816,4.0655,8759864200.0,2.3934,17.664,22467665936.4,0.2538,2004355082.02,1749636830.11,2012-01-31,0.19704350036003213
14,000559.XSHE,18.3311,2.3941,9256861365.0,4.7469,9.2005,14628454817.0,0.217,6200943486.15,5822509580.28,2012-01-31,0.13942307692307696
15,000568.XSHE,14.3531,7.2173,51740226954.0,24.1712,34.1465,56573570954.2,1.44,5628271589.15,2856713520.1,2012-01-31,0.0964713454704008
16,000623.XSHE,5.6704,1.7148,15460024303.0,20.7774,23.1428,18804075329.3,2.7168,809481664.08,690911564.42,2012-01-31,0.11214716945348802
17,000625.XSHE,-203.3888,1.3678,20159792217.0,2.7517,7.0257,37217163392.5,0.19,19606549988.7,20002232604.2,2012-01-31,0.1367074829931972
18,000627.XSHE,-16.8256,2.8326,3898338814.0,-2.728,-3.1988,5262950195.63,-0.036000000000000004,797135024.42,832484948.18,2012-01-31,0.1562478197167375
19,000630.XSHE,14.5249,2.8489999999999998,28446350207.0,3.2688,12.367,49261848233.6,0.74,52032837531.5,50898693656.5,2012-01-31,0.09244543062077502
20,000651.XSHE,8.8321,3.1092,54743150990.0,4.9914,25.6552,125069492621.0,1.3388,64074612708.6,61158359767.8,2012-01-31,0.09449989694667722
21,000671.XSHE,54.5916,1.7354,3484036043.0,0.8814,5.9695,11761758216.0,0.19,1828169582.43,1706703767.9,2012-01-31,0.17388986213725338
22,000686.XSHE,-10.5648,2.865,8771366787.0,-0.1107,-0.5469,20765182199.3,-0.03,677592524.5,717833927.89,2012-01-31,0.12170532632190215
23,000709.XSHE,30.7015,0.7431,31749624178.0,0.8678,2.8371,99804099745.6,0.15,107770205166.0,106446201521.0,2012-01-31,0.08023627861186305
24,000723.XSHE,83.1601,5.1914,2447173888.0,3.7628,6.7973,2643902662.49,0.23,1211363701.77,1164178657.88,2012-01-31,0.042778608589422264
25,000725.XSHE,7.8303,0.9565,24473991637.0,-3.8791,-8.9859,59972852071.9,-0.16,8514636554.0,10928202542.0,2012-01-31,0.07731582786287378
26,000728.XSHE,89.0694,1.2107,17853669000.0,1.9189,3.093,27584467396.7,0.2311,1315052595.51,759062567.21,2012-01-31,0.07700600919052664
27,000738.XSHE,48.6018,3.7755,9352957791.0,3.2868,5.6005,12090522078.9,0.1412,1350949551.65,1208450823.91,2012-01-31,0.1542375123375765
28,000750.XSHE,-25.6794,3.0579,7848747888.0,1.0709,1.2462,20293252211.3,0.05,846043430.3,770218550.33,2012-01-31,0.19820142453543432
29,000768.XSHE,-147.1629,2.0306,19102438172.0,0.1931,0.3771,32166350682.9,0.0142,5081888777.39,5059579936.33,2012-01-31,0.1465590484282072
30,000776.XSHE,728.4289,2.3480000000000003,73103249580.0,1.7866,6.13,136014621813.0,0.61,4325459005.79,2377388304.37,2012-01-31,0.07489705895839265
31,000783.XSHE,-45.0677,1.6385,18614185636.0,1.0097,3.1584,38751827912.8,0.14,1337998541.45,967274968.36,2012-01-31,0.12992671244734258
32,000792.XSHE,19.2172,3.7855,51771074558.0,10.514000000000001,19.4208,80915739549.4,1.8385,5539622881.41,2862752126.27,2012-01-31,0.15145999313841346
33,000826.XSHE,23.0132,5.7688,9086509777.0,5.8911,13.970999999999998,12253603099.7,0.462,1024996030.27,798413138.46,2012-01-31,0.08771129113758985
34,000839.XSHE,333.106,1.8536,10756003511.0,0.89,1.5777,18870505927.6,0.0583,1180163422.97,1267800326.91,2012-01-31,0.12829722318773407
35,000858.XSHE,15.3332,5.2524,121432975373.0,16.1359,24.1627,146746453414.0,1.2690000000000001,15648774490.2,9035308334.24,2012-01-31,0.1015933571894008
36,000876.XSHE,7.676,3.0819,29540383370.0,7.4287,14.0505,20968360549.5,1.03,50764868718.2,49120061669.7,2012-01-31,0.06176845044971034
37,000895.XSHE,-1338.4485,10.6955,39377548602.0,4.5209,7.8543,42437138787.3,0.4592,24239106613.4,23337169830.8,2012-01-31,0.062017805916695874
38,000938.XSHE,42.6661,2.8149,2254515200.0,1.6298,4.2112,3529790114.27,0.162,3644404043.75,3629839989.02,2012-01-31,0.1014640474425501
39,000959.XSHE,-26.1079,1.1222,8513929784.0,0.0867,2.0863,21615136271.8,0.0549,9502575535.22,9992265480.22,2012-01-31,0.08711129724636166
40,000961.XSHE,8.7321,1.8054,9809849498.0,2.4954,13.1395,31471593062.6,0.55,7903065272.29,7025337345.32,2012-01-31,0.1964263632991221
41,000963.XSHE,22.4888,6.5556,10022445192.0,7.7243,22.5682,14715366599.7,0.6757,8194944377.07,7738330113.27,2012-01-31,0.08012232979301576
42,000983.XSHE,16.8509,3.6087,50324664000.0,8.237,18.4178,85342188907.5,0.7948,22884339585.8,19205385113.0,2012-01-31,0.0776470137238366
43,002007.XSHE,103.2077,5.627999999999999,13321854976.0,11.9395,13.7822,12957134001.1,0.535,738447615.67,363505122.7,2012-01-31,0.06531210999934314
44,002008.XSHE,9.4767,2.6381,7404771894.0,5.7164,11.6348,11923045459.0,0.2756,2673513831.05,2362243445.18,2012-01-31,0.16502264030126082
45,002024.XSHE,10.8124,2.7260000000000004,60867043234.0,7.0214,17.3106,105525576525.0,0.4892,67625499000.0,63105122000.0,2012-01-31,0.034480285024270016
46,002027.XSHE,-10.2187,2.016,1218410517.0,0.1718,0.257,1805997930.82,0.0056,947381266.9,941843709.87,2012-01-31,0.09178151381773576
47,002044.XSHE,20.4515,5.1103,1935277500.0,6.5053,9.2073,2585540321.99,0.139,579326555.55,536804411.89,2012-01-31,0.11009951302138482
48,002065.XSHE,18.7782,4.3444,9850608640.0,9.892999999999999,14.5688,11373623438.7,0.5093,1606056451.02,1335501978.86,2012-01-31,0.11099090669863453
49,002074.XSHE,34.6532,3.6556,1578482640.0,5.7557,10.5459,2441487011.21,0.18,493069178.84,440614415.3,2012-01-31,0.11717714761670385
50,002081.XSHE,17.0643,5.9058,18377040393.0,9.7896,31.9623,21684655943.0,0.9086,6432177806.42,5839048120.1,2012-01-31,0.139324831207802
51,002142.XSHE,6.6771,1.4932,27944220926.0,1.0298,15.1654,240797776509.0,0.89,5706374000.0,2764872000.0,2012-01-31,0.03405619746902688
52,002146.XSHE,7.3613,2.4509,15039897600.0,3.9236,18.5606,33968718457.5,0.51,6101082478.69,4839724494.1,2012-01-31,0.19950759589290207
53,002153.XSHE,23.5333,6.7642,7397241600.0,15.0778,17.1945,8854453139.38,0.52,459764525.26,302985073.79,2012-01-31,0.11158221038069367
54,002174.XSHE,-135.1836,3.952,902386340.0,0.1001,0.77,1344172473.18,0.021,226840813.26,228325157.3,2012-01-31,0.16452606174928575
55,002230.XSHE,45.8019,6.3889,7516943625.0,6.4937,7.8906,7670455233.62,0.28,342597396.99,285178588.37,2012-01-31,0.31656126844468896
56,002236.XSHE,23.7544,7.819,11774402076.0,11.0706,16.345,12402992370.4,0.73,1460699765.06,1234300793.67,2012-01-31,0.10240671551263997
57,002241.XSHE,21.875,8.175,17053408060.0,8.8802,19.7717,20333004124.0,0.47,2753719950.56,2332580270.48,2012-01-31,0.1467561303706585
58,002252.XSHE,29.9917,7.0569,6454560000.0,15.0026,16.5791,6584240682.17,0.51,381696427.88,218311705.2,2012-01-31,0.06362922230950507
59,002292.XSHE,75.7295,7.325,10104832000.0,5.7157,6.7608,7207363453.14,0.22,727006274.15,623049692.79,2012-01-31,0.09363440005958447
60,002294.XSHE,19.4133,4.1442,8836656000.0,12.5764,14.765999999999998,11936854986.3,0.79,1077149060.94,746256481.74,2012-01-31,0.13235003881425728
61,002304.XSHE,19.6996,11.6265,114993000000.0,22.9515,37.134,127056431081.0,3.24,9697320356.4,5655841725.15,2012-01-31,0.16850501935668302
62,002310.XSHE,17.5595,6.0159,11263785183.0,13.5919,22.8921,14699468220.1,2.29,1965187858.05,1566421577.31,2012-01-31,0.2306324150615437
63,002352.XSHE,26.5542,2.1836,1569068525.0,3.2224,4.3585,2073680968.64,0.4,470163824.47,438363251.57,2012-01-31,0.08977645112098892
64,002385.XSHE,16.6065,3.3233,11563080000.0,7.7024,9.8108,14559606885.4,0.78,5279408885.11,4905932347.5,2012-01-31,0.17157582281371295
65,002411.XSHE,27.4205,3.0469,2356830000.0,5.2652,5.4527,3020826515.22,0.17,770044002.13,712509134.65,2012-01-31,0.21476753135044588
66,002415.XSHE,19.359,5.4305,36980000000.0,12.7014,14.6805,40368127810.4,0.87,3453156822.06,2470651269.54,2012-01-31,0.11356624917600534
67,002424.XSHE,25.5245,3.4475,6599712000.0,4.5786,6.8463,8407405957.77,0.27,701417429.65,554485320.02,2012-01-31,0.12688254714847808
68,002426.XSHE,1512.7409,2.2273,2886956100.0,3.7132,5.2927,4088639981.93,0.1697,1115660255.45,1062257040.89,2012-01-31,0.14153668399768915
69,002450.XSHE,35.4532,6.6061,6787200000.0,7.5141,11.0486,7508019924.18,0.3207,1098832203.2,969665965.32,2012-01-31,0.1619329045715278
70,002456.XSHE,-263.2222,2.8871,2808960000.0,-1.4387,-2.6004,4688450853.06,-0.13,648674946.38,677435315.07,2012-01-31,0.5242354899105472
71,002460.XSHE,49.731,4.5287,3279000000.0,5.1303,5.8898,4023862473.74,0.28,327739393.91,282919396.3,2012-01-31,0.08829619033278123
72,002465.XSHE,31.4086,1.655,7075738533.0,2.5843,2.9048,8694467710.3,0.3679,658569640.25,549797985.66,2012-01-31,0.15225456190619813
73,002466.XSHE,97.1092,4.1013,4036620000.0,2.9892,3.3418,5192804839.01,0.22,298312158.14,267370981.94,2012-01-31,0.126367785906334
74,002468.XSHE,-29.5372,2.2865,1735680000.0,3.0888,4.7192,2985511165.06,0.19399999999999998,813493135.78,777625233.64,2012-01-31,0.1946935798624116
75,002470.XSHE,17.198,3.1169,9618000000.0,8.6549,13.7957,13161820051.0,0.57,6334742941.75,5843040739.55,2012-01-31,0.06985058208818411
76,002475.XSHE,20.348,4.4072,8110377000.0,10.4286,11.1446,10445640552.3,0.73,1805745307.42,1506388762.01,2012-01-31,0.11219353791199868
77,002500.XSHE,-299.473,2.6293,15694692000.0,0.8005,2.2936,28146131862.0,0.06,764146400.0,593553103.0,2012-01-31,0.09173205556984823
78,002508.XSHE,17.1347,2.7193,4106240000.0,7.013999999999999,8.7418,4245859793.34,0.48,1066937023.89,928033582.34,2012-01-31,0.02867975889558632
79,002555.XSHE,44.4557,1.8685,1429780000.0,6.0338,7.8995,1915721333.26,0.2783,253102516.76,224929800.23,2012-01-31,0.09088986411463024
80,002558.XSHE,27.0092,2.2569,1331610000.0,6.2979,7.4814,1511286005.1,0.48,232718969.01,204381787.74,2012-01-31,0.1045629106528585
81,002572.XSHE,18.3723,2.3778,3242100000.0,9.0828,10.6158,4293748635.4,0.8,629835600.6,531961977.68,2012-01-31,0.1920963172804533
82,002601.XSHE,15.5863,4.3066,8685600000.0,16.0068,27.3354,8390870389.73,4.32,1510362207.38,1126040681.21,2012-01-31,0.049572435656780614
83,002602.XSHE,19.6278,1.5786,2436000000.0,9.5175,12.6759,3363782564.41,0.88,745992572.11,605838766.87,2012-01-31,0.1092073225719181
84,002608.XSHE,21.584,1.617,3198720000.0,3.6252,10.8872,6726447058.26,1.43,1953159642.57,1756742374.55,2012-01-31,0.08272847047550601
85,002624.XSHE,25.5598,3.0336,1475000000.0,12.3847,23.1718,197559732.15,0.66,332677250.53,285246943.45,2012-01-31,0.07322489313273754
86,300003.XSHE,20.9579,4.796,11229960000.0,15.6951,16.8851,13490161497.1,0.4467,678769523.57,268496848.15,2012-01-31,0.05638972305638978
87,300015.XSHE,40.59,6.6348,9073728000.0,8.1282,10.3255,9332816274.92,0.31,966750963.5,777018084.74,2012-01-31,0.029664695572409017
88,300017.XSHE,25.9591,2.3706,1847487146.0,4.5921,4.8207,1922528782.12,0.23,382167189.86,346917794.45,2012-01-31,0.1711732096390993
89,300024.XSHE,26.6647,4.7491,5637680400.0,5.7208,7.3934,6614052715.55,0.27,546261214.18,459245055.06,2012-01-31,0.008442293833007293
90,300027.XSHE,32.433,4.9322,8322048000.0,5.1512,6.5198,9548149074.81,0.17,481297343.95,379519891.28,2012-01-31,0.0915833408401646
91,300033.XSHE,62.4488,1.7983,2010624000.0,4.0142,4.5664,2717977058.33,0.37,167909628.35,118080013.56,2012-01-31,0.2179192555394784
92,300059.XSHE,25.5474,2.1984,3738000000.0,4.8088,5.0589,4223969468.4,0.4,198275042.38,115710250.2,2012-01-31,0.29160031847133777
93,300070.XSHE,21.4611,3.3669,10989132000.0,2.5622,2.7239999999999998,12240238304.7,0.25,368083629.66,275351643.46,2012-01-31,0.16540056076110135
94,300072.XSHE,22.5444,2.6189,2990079800.0,3.6867,5.5583,3763745417.33,0.31,409472534.65,340213144.32,2012-01-31,0.1340454340303897
95,300122.XSHE,61.6859,3.7147,8388000000.0,5.6875,5.9543,11197163180.7,0.32,456652112.14,309831451.78,2012-01-31,0.14450333725952091
96,300124.XSHE,22.6099,4.1521,10396080000.0,10.2343,10.9358,15352248689.5,1.19,789887920.75,498177781.4,2012-01-31,0.09703376716456807
97,300136.XSHE,19.499000000000002,2.9683,2002766800.0,8.8253,9.145,2949590988.74,0.4149,121334857.44,56642914.4,2012-01-31,0.12049127343244985
98,300144.XSHE,26.941,2.8002,7573104000.0,6.309,7.3163,7692714479.5,0.52,387976059.28,152151856.75,2012-01-31,0.112750724162431
99,300251.XSHE,22.2623,2.9725,5315600000.0,8.8568,10.3649,6925068439.68,1.14,413759118.43,290804273.67,2012-01-31,0.16700232378001545
100,600000.XSHG,4.859,1.1557,171985006446.0,0.8475,15.1658,2524302191530.0,1.067,49307426000.0,23080440000.0,2012-01-31,0.034713242753983416

过滤式

低方差特征过滤

删除低方差的一些特征

  • 特征方差小：某个特征大多样本的值比较相近
  • 特征方差大：某个特征很多样本的值都有差别

API

sklearn.feature_selection.VArianceThreshold(threshold=0.0)

删除所有低方差特征

  • Variance.fit_transform(X)，
  • X：numpy array格式的数据[m_sample,n_features]，
  • 返回值：训练集差异低于threadshold的特征将被删除。默认值是保留非零方差特征，即删除所有样本中具有相同值的特征

数据计算

from sklearn.feature_selection import VarianceThreshold
import pandas as pd


def variance_demo():
    """
    低方差特征过滤
    :return:
    """
    # 1、获取数据
    data = pd.read_csv('factor_returns.csv')
    print('data:\n', data)
    data = data.iloc[:,1:-2]
    print('data:\n', data)

    # 2、实例化一个转换器类
    #transform = VarianceThreshold()
    transform = VarianceThreshold(threshold=10)

    # 3、调用fit_transform
    data_new = transform.fit_transform(data)
    print("data_new\n", data_new, data_new.shape)

    return None

if __name__ == "__main__":
    variance_demo()

在这里插入图片描述

相关系数

皮尔逊相关系数（Pearson Correlation Coefficient）：反映特征之间相关关系密切程度的统计指标

在这里插入图片描述
在这里插入图片描述

特点

相关系数的值介于-1与+1之间，即-1<=r<=+1。其性质如下：

  • 当r>0时，表示两变量正相关；r<0时，两变量为负相关
  • 当|r|=1时，表示两变量为完全相关；当r=0时，表示两变量间无相关关系
  • 当0<|r|<1时，表示两变量存在一定程度的相关。且|r|越接近1，两变量间线性关系越密切；|r|越接近0，表示两变量的线性相关越弱
  • 一般可按三级划分：|r|<0.4为低度相关；0.4<=|r|<0.7为显著相关；0.7<=|r|<1为高维线性相关

API

from scipy.stats import pearsonr
  • x：（N.）array_like
  • y：（N.）array_like Returns:(Perason’s correlation coefficient, p-value)
from sklearn.feature_selection import VarianceThreshold
from scipy.stats import pearsonr
import pandas as pd

def variance_demo():
    """
    低方差特征过滤
    :return:
    """
    # 1、获取数据
    data = pd.read_csv('factor_returns.csv')
    print('data:\n', data)
    data = data.iloc[:,1:-2]
    print('data:\n', data)

    # 2、实例化一个转换器类
    #transform = VarianceThreshold()
    transform = VarianceThreshold(threshold=10)

    # 3、调用fit_transform
    data_new = transform.fit_transform(data)
    print("data_new\n", data_new, data_new.shape)

    # 计算两个变量之间的相关系数
    r = pearsonr(data["pe_ratio"],data["pb_ratio"])
    print("相关系数：\n", r)
    return None


if __name__ == "__main__":
    variance_demo()

相关系数看前面那个
在这里插入图片描述

主成分分析

什么是主成分分析（PCA）

定义：高维数据转换为低维数据的过程，在此过程中可能会舍弃原有数据、创造新的变量
作用：是数据维数的压缩，尽可能降低原数据的维数（复杂度），损失少量信息
应用：回归分析或者聚类分析中

计算案例理解

二维降到一维

API

sklearn.decomposition.PCA(n_components=None)
  • 将数据分解为较低维度空间
  • n_components:
  • 小数：表示保留百分之多少的信息
  • 整数：减少到多少特征
    PCA.fit_transform(X)，X：numpy array格式的数据[N_samples, n_features]，返回值：转换后指定维度的array

数据计算

from sklearn.decomposition import PCA


def pca_demo():
    """
    PCA降维
    :return:
    """
    data = [[2,8,4,5], [6,3,0,8], [5,4,9,1]]
    # 1、实例化一个转换器类
    transform = PCA(n_components=2)  # 4个特征降到2个特征

    # 2、调用fit_transform
    data_new = transform.fit_transform(data)
    print("data_new\n", data_new)

    transform2 = PCA(n_components=0.95)  # 保留95%的信息
    data_new2 = transform2.fit_transform(data)
    print("data_new2\n", data_new2)

    return None

if __name__ == "__main__":
    pca_demo()

在这里插入图片描述

案例：探究用户对物品类别的喜好细分降维

数据：

  • order_prodects_prior.csv：订单与商品信息
    字段：order_id, product_id, add_to_cart_order, reordered
  • products.csv：商品信息
    字段：product_id, product_name, aisle_id, department_id
  • order.csv：用户的订单信息
    字段：order_id, user_id, eval_set, order_number, …
  • aisles.csv：商品所属具体物品类别
    字段：aisle_id, aisle

需求

  • 需要将user_id和aisle放在同一个表中—合并
  • 找到user_id和aisle----交叉表和透视表
# 1、获取数据
# 2、合并表
# 3、找到suer_id和aisle之间的关系
# 4、PAC降维
import pandas as pd
from sklearn.decomposition import PCA
# 1、获取数据
order_products = pd.read_csv('order_products__prior.csv') #32434489× 4
products = pd.read_csv('products.csv')  # (49688,4)
orders = pd.read_csv('orders.csv')     #3421083 rows × 7 columns
aisles = pd.read_csv('aisles.csv')  #(134,2)
# 2、合并表'
# 合并aisles和products
tab1 = pd.merge(aisles, products, on=["aisle_id", "aisle_id"]) #49688 × 5 c
tab2 = pd.merge(tab1, order_products, on=["product_id", "product_id"])#32434489 ,8
tab3 = pd.merge(tab2, orders, on=["order_id", "order_id"])#32434489 ,14
# tab3.head()
# 3、找到suer_id和aisle之间的关系
table = pd.crosstab(tab3["user_id"], tab3["aisle"]) #206209 rows × 134 columns
data = table[:10000] #10000 rows × 134 columns
print(data)
# 4、PAC降维
# 1)实例化一个转换器类
transfer = PCA(n_components=0.95)  # 保留95%的信息
# 2）调用fit_transform
data_new = transfer.fit_transform(data)  #(10000, 42)，由134个特征降维到42个
print(data_new)
print(data_new.shape)

在这里插入图片描述

在这里插入图片描述

Appreciate(欣赏)
关注
  • 22
    点赞
  • 踩
  • 15
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
基于sklearn分析特征工程（特征预处理、特征选择、降维）
北落师门XY的博客
07-09 4124
特征相关文章推荐：https://www.zhihu.com/question/29316149
机器学习学习笔记（四）之特征降维
qq_42769184的博客
09-07 582
特征降维 降维是指在某些限定的条件下，降低随机变量（特征）的个数，得到一组‘不相关’的主变量的过程 效果：特征与特征之间不相关 1、降维的两种方式 特征选择 主成分分析 2、什么是特征选择？ 定义：数据中包含冗余或相关变量（或称特征、属性、指标等），旨在从原有特征中找出主要特征。 方法： Filter（过滤式）：主要探究特征本身特点、特征与特征和目标值之间关联 方差选择法：低方差特征过滤 相关系数 Embedded（嵌入式）：算法自动选择特征（特征与目标值之间的关联） 决策树：信息熵、信息增熵
机器学习（5）特征降维
weixin_44285715的博客
09-06 579
1 降维： 指在某些限定条件下，降低随机变量（特征）个数，得到一组“不相关”主变量的过程。 相关特征（correlated feature） 相对湿度与降雨量之间的相关 2 什么是特征选择 数据中包含冗余或相关变量（特征、属性、指标），旨在从原有特征中找出主要特征。 2.1 方法 Filter（过滤式）：主要研究特征本身特点，特征与特征和目标值之间关联 方差选择法：低方差特征过滤 相关系数法...
python机器学习之特征降维
柳小葱的博客
06-10 2866
????上次我们学习了数据处理中的特征处理，主要包括归一化、标准化、二值化、独热编码等，对数据做完处理后，接下来就应该进行数据的降维处理了，对于数据处理感兴趣的同学可以查看????: python机器学习之特征工程: python机器学习之数据探索. python机器学习之特征工程: python机器学习之特征处理. ????今天我们要学的是数据的降维部分，数据的降维主要是将高维向量空间的数据点印射到低纬空间中，当然也可以从低纬度转换到高维度中，在实际应用中，高维度空间包含较多的冗余信息与噪声信息，从而
（sklearn）机器学习（六）特征降维
weixin_44525558的博客
06-16 2864
特征降维1 降维 - 降低维度 1 降维 - 降低维度
scikit-learn-1.0.2.tar.gz
01-18
在scikit-learn 1.0.2版本中，开发者可能对模型选择、模型评估和特征选择进行了优化。例如，交叉验证（cross-validation）是评估模型性能的关键工具，该版本可能提升了其稳定性和效率。此外，预处理步骤如标准化和...
scikit-learn-1.3.2.tar.gz
01-17
其次，scikit-learn支持数据预处理和特征选择，这对于任何机器学习项目都至关重要。在新版本中，可能会有新的预处理工具或改进的现有工具，以更好地处理缺失值、异常值，或者更有效地执行特征缩放和编码。例如，`...
scipy+scikit-learn组件
04-20
Scikit-learn提供了各种监督和无监督的学习算法，如分类、回归、聚类和降维等。它还包含了模型选择、数据预处理、评估指标和交叉验证等功能，方便用户进行完整的机器学习流程。例如，支持向量机（SVM）用于分类和...
scikit-learn库来进行PCA降维
08-17
在**Python**中，我们可以利用**scikit-learn**库的`PCA`类来实现PCA降维。以下是一个使用scikit-learn的PCA降维的示例： ```python from sklearn.decomposition import PCA from sklearn.datasets import load_...
scikit-learn学习笔记.pdf
06-06
Scikit-learn是Python编程语言中的一个强大机器学习库，被广泛用于数据挖掘、数据分析以及机器学习项目。这个库提供了一系列高效且易于使用的工具，帮助数据科学家进行预处理、建模、评估等任务。 **主要特点：** 1...
Sklearn - 特征降维
编码时光
03-25 559
Matrix Decomposition Feature Selection 一、利用特征提取进行特征降维 decomposition 1、PCA 主成分 2、对线性不可分数据进行特征降维 核机制 3、通过最大化类间可分性进行特征降维 4、使用矩阵分解进行特征降维 5、对稀疏矩阵进行特征降维 二、使用特征选择 进行降维 feature_selection 1、数值型特征方差的阈值化 2、二值特征的方差阈值化 3、处理高度相关性的特征 4、删除与分类任务不相关的特征 1）分类类型 2）数值型特征 3）Sele
sklearn-降维算法
weixin_58975360的博客
03-31 3231
sklearn中的降维算法 主成分分析PCA class sklearn.decomposition.PCA (n_components=None, copy=True, whiten=False, svd_solver=’auto’, tol=0.0, iterated_power=’auto’, random_state=None) 特征选择是从已存在的特征中选取携带信息最多的，选完之后的特征依然具有可解释性，我们依然知道这个特 征在原数据的哪个位置，代表着原数据上的什么含义。 而PCA，是将
机器学习——特征降维
f18896984569的博客
12-03 1039
提供数据降维的简单信息，并安利整理详细的大神博客
机器学习 - 特征降维
最新发布
weixin_47552266的博客
05-16 1066
特征降维在处理高维数据时尤为重要，可以提高模型的训练速度和泛化性能。在实际应用中，通常需要结合特征选择和特征提取方法，根据具体问题选择合适的降维技术。
机器学习：特征降维，奇异值分解，PCA，LDA
燕双嘤
04-12 1582
1，概述 1.1，维数灾难 维数灾难：通常是指在涉及到向量的计算的问题中，随着维数的增加，计算量呈指数倍增长的一种现象。在很多机器学习问题中，训练集中的每条数据经常伴随着上千、甚至上万个特征。要处理这所有的特征的话，不仅会让训练非常缓慢，还会极大增加搜寻良好解决方案的困难。这个问题就是我们常说的维数灾难。 维数灾难涉及数字分析、抽样、组合、机器学习、数据挖掘和数据库等诸多领域。在机器学习的建模过程中，通常指的是随着特征数量的增多，计算量会变得很大，如特征达到上亿维的话，在进行计算的时候是算不出来
机器学习-特征降维
Ministar的博客
04-22 486
机器学习-特征降维 1 降维--降低维度 1.1 维度:在numpy中ndarray中学习过二维数组,在list列表中学习过二维,三位列表 1.2 维数:嵌套的层数 1.3 特征降维 降的不是嵌套的层数,降的是特征值的个数,就是列数 1.4 降维定义 降维是指在某些限定条件下,降低随机变量(特征值)的个数,得到一组不相关的主变量的过程 1.5 效果 特征与特征之间不相关 相关:比如说相对湿度和降雨量具有相
机器学习特征降维
小森的博客
05-12 980
相关系数的值介于–1与+1之间，当 r>0 时，表示两变量正相关，r<0 时，两变量为负相关，当 |r|=1 时，表示两变量为完全相关，当r=0时，表示两变量间无相关关系。斯皮尔曼相关系数与之前的皮尔逊相关系数大小性质一样，取值 [-1, 1]之间，斯皮尔曼相关系数比皮尔逊相关系数应用更加广泛。PCA 通过对数据维数进行压缩，尽可能降低原数据的维数(复杂度)损失少量信息，在此过程中可能会舍弃原有数据、创造新的变量。通过正交变换将一组可能存在相关性的变量转换为一组线性不相关的变量，转换后的这组变。
机器学习-特征工程中的特征降维
数据掘金
08-19 5091
特征降维
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值