<div style="line-height:1.7;color:#000000;font-size:14px;font-family:Arial"><div style="LINE-HEIGHT: 1.7; FONT-FAMILY: Arial; COLOR: #000000; FONT-SIZE: 14px">
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 16px"><strong>Dear all, </strong></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 16px"><strong>    I faced with an embarrassing situation. I have completed some sensitivity analysis. When I used the script like below, I achieved a discouraging accuracy rate:0.51267.</strong></span></div>
<div>       <span style="FONT-SIZE: 13px">  <span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>clf = LinearCSVMC()</em></span></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>           cv = CrossValidation(clf, NFoldPartitioner(), errorfx = lambda p,t:np.mean(p==t), enable_ca=['stats'],postproc = mean_sample())</em><br><em>           sensana = clf.get_sensitivity_analyzer()</em></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>           cv_sensana = RepeatedMeasure(sensana, ChainNode((NFoldPartitioner(), Splitter('partitions', attr_values=(1,)))))</em><br><em>           acc = cv(ds)</em></span></div>
<div><span style="FONT-SIZE: 13px"><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>           sensmap_cv = cv_sensana(ds</em></span>)</span></div>
<div>   <span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 16px"><strong>Then I want to improve the accuracy. So I firstly select 20% features which contributed most for the predicted accuracy. However, I achieved a lower accuracy rate:0.40268.</strong><br></span>     <em>  </em><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>fsel = SensitivityBasedFeatureSelection(OneWayAnova(),FractionTailSelector(0.2,mode='select',tail='upper'))</em><br><em>         fclf = FeatureSelectionClassifier(clf, fsel)</em></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>         cv_sel = CrossValidation(fclf,NFoldPartitioner(), errorfx = lambda p,t:np.mean(p==t),enable_ca=['stats'],postproc = mean_sample())</em></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>         sensana_sel = fclf.get_sensitivity_analyzer()</em></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>         cv_sensana_sel = RepeatedMeasure(sensana_sel, ChainNode((NFoldPartitioner(), Splitter('partitions', attr_values=(1,)))))</em></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>         acc_sel = cv_sel(ds)</em></span></div>
<div><span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 13px"><em>         sensmap_cv_sel = cv_sensana_sel(ds)</em><br></span>  <span style="FONT-FAMILY: Times New Roman; FONT-SIZE: 16px"><strong>What's wrong? Would you tell me?</strong></span><br></div></div><br><br><span title="neteasefooter"><span id="netease_mail_footer"></span></span></div><br><br><span title="neteasefooter"><span id="netease_mail_footer"></span></span>