<?xml version="1.0" encoding="UTF-8"?><xml><records><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>10</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Mezache Hatem</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Kernel Principal Components Analysis with Extreme Learning Machines for Wind Speed Prediction.</style></title><secondary-title><style face="normal" font="default" size="100%"> IREC2016 Seventh International Renewable Energy Congress</style></secondary-title></titles><dates><year><style  face="normal" font="default" size="100%">2016</style></year></dates><urls><web-urls><url><style face="normal" font="default" size="100%">https://hal.inria.fr/hal-01394000/document</style></url></web-urls></urls><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">&lt;span style=&quot;left:150.81px;top:400.428px;15px;serif;&quot;&gt;N&lt;/span&gt;&lt;span style=&quot;left:161.61px;top:400.428px;15px;serif;transform:scaleX(1.04515);&quot;&gt;owadays&lt;/span&gt;&lt;span style=&quot;left:216.405px;top:400.428px;15px;serif;transform:scaleX(1.076);&quot;&gt;, wind power &lt;/span&gt;&lt;span style=&quot;left:306.21px;top:400.428px;15px;serif;transform:scaleX(1.11686);&quot;&gt;and&lt;/span&gt;&lt;span style=&quot;left:335.205px;top:400.428px;15px;serif;transform:scaleX(1.06508);&quot;&gt;precise forecasting &lt;/span&gt;&lt;span style=&quot;left:460.41px;top:400.428px;15px;serif;transform:scaleX(1.1134);&quot;&gt;are &lt;/span&gt;&lt;span style=&quot;left:61.2px;top:417.633px;15px;serif;transform:scaleX(1.07125);&quot;&gt;of great importance&lt;/span&gt;&lt;span style=&quot;left:202.41px;top:417.633px;15px;serif;transform:scaleX(1.04877);&quot;&gt;for the development of&lt;/span&gt;&lt;span style=&quot;left:365.61px;top:417.633px;15px;serif;transform:scaleX(1.05521);&quot;&gt;modern ele&lt;/span&gt;&lt;span style=&quot;left:440.205px;top:417.633px;15px;serif;transform:scaleX(1.09134);&quot;&gt;ctrical&lt;/span&gt;&lt;span style=&quot;left:61.2px;top:434.838px;15px;serif;transform:scaleX(1.07216);&quot;&gt;grids.&lt;/span&gt;&lt;span style=&quot;left:105px;top:434.838px;15px;serif;transform:scaleX(1.095);&quot;&gt;In this paper&lt;/span&gt;&lt;span style=&quot;left:202.8px;top:434.838px;15px;serif;transform:scaleX(1.02156);&quot;&gt;we&lt;/span&gt;&lt;span style=&quot;left:228.21px;top:434.838px;15px;serif;transform:scaleX(1.06925);&quot;&gt;propose a&lt;/span&gt;&lt;span style=&quot;left:301.605px;top:434.838px;15px;serif;transform:scaleX(1.065);&quot;&gt;prediction system&lt;/span&gt;&lt;span style=&quot;left:426.015px;top:434.838px;15px;serif;transform:scaleX(1.0641);&quot;&gt;for time &lt;/span&gt;&lt;span style=&quot;left:61.2px;top:452.028px;15px;serif;transform:scaleX(1.04488);&quot;&gt;series&lt;/span&gt;&lt;span style=&quot;left:103.605px;top:452.028px;15px;serif;transform:scaleX(1.04604);&quot;&gt;based on&lt;/span&gt;&lt;span style=&quot;left:169.41px;top:452.028px;15px;serif;transform:scaleX(1.08261);&quot;&gt;Kernel&lt;/span&gt;&lt;span style=&quot;left:220.2px;top:452.028px;15px;serif;transform:scaleX(1.09457);&quot;&gt;Principal&lt;/span&gt;&lt;span style=&quot;left:286.2px;top:452.028px;15px;serif;transform:scaleX(1.04498);&quot;&gt;Component Analysis&lt;/span&gt;&lt;span style=&quot;left:428.61px;top:452.028px;15px;serif;&quot;&gt;(&lt;/span&gt;&lt;span style=&quot;left:433.605px;top:452.028px;15px;serif;transform:scaleX(1.05339);&quot;&gt;KPCA) &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:469.428px;15px;serif;transform:scaleX(1.11686);&quot;&gt;and&lt;/span&gt;&lt;span style=&quot;left:97.1999px;top:469.428px;15px;serif;transform:scaleX(1.08388);&quot;&gt;Extreme&lt;/span&gt;&lt;span style=&quot;left:164.4px;top:469.428px;15px;serif;transform:scaleX(1.09177);&quot;&gt;Learning&lt;/span&gt;&lt;span style=&quot;left:235.605px;top:469.428px;15px;serif;transform:scaleX(1.06902);&quot;&gt;Machine&lt;/span&gt;&lt;span style=&quot;left:303.6px;top:469.428px;15px;serif;&quot;&gt;(&lt;/span&gt;&lt;span style=&quot;left:308.61px;top:469.428px;15px;serif;transform:scaleX(1.07443);&quot;&gt;ELM)&lt;/span&gt;&lt;span style=&quot;left:348px;top:469.428px;15px;serif;transform:scaleX(1.08054);&quot;&gt;. To compare the &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:486.633px;15px;serif;transform:scaleX(1.08745);&quot;&gt;proposed approach, three&lt;/span&gt;&lt;span style=&quot;left:241.41px;top:486.633px;15px;serif;transform:scaleX(1.06526);&quot;&gt;dimensionality reduction techniques &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:503.838px;15px;serif;transform:scaleX(1.06277);&quot;&gt;were used&lt;/span&gt;&lt;span style=&quot;left:129.21px;top:503.838px;15px;serif;&quot;&gt;:&lt;/span&gt;&lt;span style=&quot;left:142.41px;top:503.838px;15px;serif;transform:scaleX(1.02876);&quot;&gt;full &lt;/span&gt;&lt;span style=&quot;left:172.2px;top:503.838px;15px;serif;transform:scaleX(1.05774);&quot;&gt;space (50 variables&lt;/span&gt;&lt;span style=&quot;left:301.605px;top:503.838px;15px;serif;&quot;&gt;)&lt;/span&gt;&lt;span style=&quot;left:306.6px;top:503.838px;15px;serif;transform:scaleX(1.08148);&quot;&gt;, part of space (last four &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:521.028px;15px;serif;transform:scaleX(1.04454);&quot;&gt;variables) and classical Principal Components Analysis (PCA). &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:538.428px;15px;serif;transform:scaleX(1.07256);&quot;&gt;These models were compared using three evaluation criteria: &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:555.633px;15px;serif;&quot;&gt;m&lt;/span&gt;&lt;span style=&quot;left:73.3949px;top:555.633px;15px;serif;transform:scaleX(1.07334);&quot;&gt;ean &lt;/span&gt;&lt;span style=&quot;left:102.6px;top:555.633px;15px;serif;transform:scaleX(1.06993);&quot;&gt;absolute&lt;/span&gt;&lt;span style=&quot;left:162.6px;top:555.633px;15px;serif;transform:scaleX(1.0868);&quot;&gt;error (M&lt;/span&gt;&lt;span style=&quot;left:222.6px;top:555.633px;15px;serif;&quot;&gt;A&lt;/span&gt;&lt;span style=&quot;left:233.4px;top:555.633px;15px;serif;transform:scaleX(1.03846);&quot;&gt;E), &lt;/span&gt;&lt;span style=&quot;left:258.81px;top:555.633px;15px;serif;transform:scaleX(1.08341);&quot;&gt;root &lt;/span&gt;&lt;span style=&quot;left:292.005px;top:555.633px;15px;serif;transform:scaleX(1.06445);&quot;&gt;mean &lt;/span&gt;&lt;span style=&quot;left:333.405px;top:555.633px;15px;serif;transform:scaleX(1.10801);&quot;&gt;square&lt;/span&gt;&lt;span style=&quot;left:383.415px;top:555.633px;15px;serif;transform:scaleX(1.09474);&quot;&gt;error (&lt;/span&gt;&lt;span style=&quot;left:429.015px;top:555.633px;15px;serif;transform:scaleX(1.05862);&quot;&gt;RMS&lt;/span&gt;&lt;span style=&quot;left:462.615px;top:555.633px;15px;serif;transform:scaleX(1.03156);&quot;&gt;E), &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:572.838px;15px;serif;transform:scaleX(1.09254);&quot;&gt;and &lt;/span&gt;&lt;span style=&quot;left:91.9949px;top:572.838px;15px;serif;transform:scaleX(1.07488);&quot;&gt;normalized&lt;/span&gt;&lt;span style=&quot;left:171.405px;top:572.838px;15px;serif;transform:scaleX(1.06994);&quot;&gt;mean square er&lt;/span&gt;&lt;span style=&quot;left:276.405px;top:572.838px;15px;serif;transform:scaleX(1.08652);&quot;&gt;ror (&lt;/span&gt;&lt;span style=&quot;left:309px;top:572.838px;15px;serif;&quot;&gt;N&lt;/span&gt;&lt;span style=&quot;left:319.8px;top:572.838px;15px;serif;transform:scaleX(1.03232);&quot;&gt;MSE). The results show &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:590.028px;15px;serif;transform:scaleX(1.10458);&quot;&gt;that the &lt;/span&gt;&lt;span style=&quot;left:114.6px;top:590.028px;15px;serif;transform:scaleX(1.06779);&quot;&gt;reduction of the original input space affects&lt;/span&gt;&lt;span style=&quot;left:396.615px;top:590.028px;15px;serif;transform:scaleX(1.02625);&quot;&gt;positively &lt;/span&gt;&lt;span style=&quot;left:461.415px;top:590.028px;15px;serif;transform:scaleX(1.06995);&quot;&gt;the &lt;/span&gt;&lt;span style=&quot;left:61.1999px;top:607.428px;15px;serif;transform:scaleX(1.04498);&quot;&gt;prediction output of &lt;/span&gt;&lt;span style=&quot;left:199.41px;top:607.428px;15px;serif;transform:scaleX(1.01654);&quot;&gt;the wind speed&lt;/span&gt;&lt;span style=&quot;left:298.41px;top:607.428px;15px;serif;&quot;&gt;.&lt;/span&gt;&lt;span style=&quot;left:308.01px;top:607.428px;15px;serif;transform:scaleX(1.07706);&quot;&gt;Thus,&lt;/span&gt;&lt;span style=&quot;left:350.205px;top:607.428px;15px;serif;transform:scaleX(1.02059);&quot;&gt;It can be concluded &lt;/span&gt;&lt;span style=&quot;left:61.2px;top:624.633px;15px;serif;transform:scaleX(1.12244);&quot;&gt;that &lt;/span&gt;&lt;span style=&quot;left:92.805px;top:624.633px;15px;serif;transform:scaleX(1.0248);&quot;&gt;the non linear model (KPCA)&lt;/span&gt;&lt;span style=&quot;left:296.205px;top:624.633px;15px;serif;transform:scaleX(1.03267);&quot;&gt;model &lt;/span&gt;&lt;span style=&quot;left:341.01px;top:624.633px;15px;serif;transform:scaleX(1.06387);&quot;&gt;outperform the other &lt;/span&gt;&lt;span style=&quot;left:61.2px;top:641.838px;15px;serif;transform:scaleX(1.06947);&quot;&gt;reduction techniques in terms of p&lt;/span&gt;&lt;span style=&quot;left:279.81px;top:641.838px;15px;serif;transform:scaleX(1.08166);&quot;&gt;rediction performance&lt;/span&gt;&lt;span style=&quot;left:423.405px;top:641.838px;15px;serif;&quot;&gt;.&lt;/span&gt;</style></abstract></record></records></xml>