Are you sure that reweighting the points will lead to better results? Can
you also post your ROC AUC code?
On Fri, Jan 15, 2016 at 6:27 AM, Mamun Rashid <mamunbabu2...@gmail.com>
wrote:
> Hi everyone,
>
> I am trying to run a RandomForest classifier with and without sample
> weight via a custom 5 fold cross validation.
> My data has two classes. [ 0 = Negative and 1 = Positive ].
>
> The ratio between my positive and negative class is 1/22 (i.e. negative
> class is 22 times of positive class).
> I do not observe any performance difference between weighted vs
> non-weighted run.
>
> Weighted run AUC : Mean AUC = 0.64
> Non Weighted run AUC : Mean AUC = 0.64
>
> *** I ran weighted and non-weighted run separately.
>
> Any comment and suggestion appreciated.
>
> Thanks,
> Mamun
>
> Code Segment
> ==============
>
> clf
> = RandomForestClassifier(n_estimators=250, max_depth=None,
> min_samples_split=2, random_state=0, oob_score=True)
> weight=tr_neg_data.shape[0]/tr_pos_data.shape[0]
> print "Neg to Pos Class Ratio :: ", weight
> fold_sample_weight = np.array([weight if m == 1 else 1 for m in
> df_tr[label_column]])
>
>
> ## Clf Non Weighted Run ##
> y_pred = clf.fit(X_tr, y_tr ).predict(X_te)
> y_prob = clf.fit(X_tr, y_tr ).predict_proba(X_te)
>
> ## Clf in Weighted Run ##
> y_pred = clf.fit(X_tr, y_tr,sample_weight=fold_sample_weight
> ).predict(X_te)
> y_prob = clf.fit(X_tr, y_tr,sample_weight
> =fold_sample_weight).predict_proba(X_te)
>
>
> Version
> ===========
>
> python : 3.5.0 ; Sklearn : 0.17
>
>
>
>
>
> ------------------------------------------------------------------------------
> Site24x7 APM Insight: Get Deep Visibility into Application Performance
> APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
> Monitor end-to-end web transactions and take corrective actions now
> Troubleshoot faster and improve end-user experience. Signup Now!
> http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140
> _______________________________________________
> Scikit-learn-general mailing list
> Scikit-learn-general@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
>
------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=267308311&iu=/4140
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
