jz1584 / MLProject

MLProject

With TFIDF:

Decision Tree: 71%

Random Froest: 83%

SVM: 81% --> very slow at training (585.426215887) comment:yes, and it even takes much longer time to find the optimal penalty parameter updated: SVM 83.5% with Lambda=5 (Run time: 584.36698103)

Logistics: 84%

KNN: 66% --> very slow at testing

Adaboost: 74%

Linear Regressin: has bug comment: I dont see how linear regression could be used for classification ?? We could use it by mapping score >0.5 --> 1 and <0.5 --> 0

Without TFIDF:

comment: the TFIDF doesn't imporve the performance

Decision Tree: 70%

Random Froest: 82%

SVM:

Logistics: 83%

Logistics with L2 :"testAccuracy": 0.8376196172248804

Logistics with L1: testAccuracy: 0.837320574163

KNN:

Adaboost:

Linear Regressin: has bug comment: I dont see how linear regression could be used for classification ??

#wordDim 2000 --> 4000 dataRate 0.04 The best performance of logistic regression: %85