NuSVC Module

A Nu-Support Vector Classification. model.

The doc is here : http://scikit-learn.org/dev/modules/generated/sklearn.svm.NuSVC.html#sklearn.svm.NuSVC

---

Sortie du script

$ python NuSVC.py
Opening the file 'train.csv' and 'test.csv'...
Find the best value for the meta parameter nu, with 5 run for each...
Searching in the range : [0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29]...
Using the first part (67.00%, 596 passengers) of the training dataset as training, 
and the second part (33.00%, 295 passengers) as testing !
For nu=0.01, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 51.01%...
For nu=0.02, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 42.45%...
For nu=0.03, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 40.77%...
For nu=0.04, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 68.79%...
For nu=0.05, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 38.09%...
For nu=0.06, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 55.20%...
For nu=0.07, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 36.41%...
For nu=0.08, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 54.53%...
For nu=0.09, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 38.42%...
For nu=0.1, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 51.85%...
For nu=0.11, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 73.99%...
For nu=0.12, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 69.63%...
For nu=0.13, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 66.61%...
For nu=0.14, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 65.10%...
For nu=0.15, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 77.01%...
For nu=0.16, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 76.01%...
For nu=0.17, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 67.11%...
For nu=0.18, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 82.05%...
For nu=0.19, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 78.02%...
For nu=0.2, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 76.85%...
For nu=0.21, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 81.88%...
For nu=0.22, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 83.22%...
For nu=0.23, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 82.21%...
For nu=0.24, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 82.55%...
For nu=0.25, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 83.72%...
For nu=0.26, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 80.37%...
For nu=0.27, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 82.55%...
For nu=0.28, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 83.72%...
For nu=0.29, learning from the first part of the dataset...
... this value of nu seems to have a (mean) quality = 83.22%...
With trying each of the following nu ([0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29]), each 5 times, the best one is 0.25. (for a quality = 83.72%)
Creating the classifier with the optimal value of nu.
Learning...
 Proportion of perfect fitting for the training dataset = 84.40%
Predicting for the testing dataset
Prediction: wrote in the file csv/NuSVC_best.csv.

Résultats

La soumission du résultat à Kaggle donne ??.??%.

---

NuSVC.list_nu = [0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1, 0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.2, 0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29]

Espace de recherche

NuSVC.Number_try = 5

Nombre de tests utilisés pour méta-apprendre

NuSVC.proportion_train = 0.67

Proportion d’individus utilisés pour méta-apprendre.

NuSVC.best_nu = 0.29

La valeur optimale trouvée pour le paramètre n_estimators

NuSVC.score = 85.970819304152641

The score for this classifier.