def performPrediction(allFeatures, allLabels, featureNames):
'''
first do PCA
'''
selected_features = None ## initialization
pcaObj = decomposition.PCA(n_components=pca_comp)
pcaObj.fit(allFeatures)
# variance of features
variance_of_features = pcaObj.explained_variance_
# how much variance is explained each component
variance_ratio_of_features = pcaObj.explained_variance_ratio_
totalvarExplained = float(0)
for index_ in xrange(len(variance_ratio_of_features)):
var_exp_ = variance_ratio_of_features[index_]
totalvarExplained = totalvarExplained + var_exp_
print "Prin. comp#{}, ( indi) explained variance:{}, total explained variance:{}".format(
index_ + 1, var_exp_, totalvarExplained)
no_features_to_use = for_feature_selection
print "Of all the features, we will use:", no_features_to_use
print "-" * 50
pcaObj.n_components = no_features_to_use
selected_features = pcaObj.fit_transform(allFeatures)
print "Selected feature dataset size:", np.shape(selected_features)
print "-" * 50
printPCAInsights(pcaObj, topComponentCount, featureNames)
print "-" * 50
'''
lets start prediction , now that we ahve feature selection otu of the way
'''
sklearn_models.performModeling(selected_features, allLabels, 10)
all_features, all_labels)
### use randomized logi. regression to get the features ::: as this performs worse then l1-penalized , it wil not be used
# selected_indices_for_features = sklearn_models.getElgiibleFeatures(all_features, all_labels)
print "Total selected feature count:", len(selected_indices_for_features)
print "The selected feature names: ", Utility.printFeatureName(
selected_indices_for_features, True) ##True for enbaling steroid headers
print "-" * 50
### select the features based on feature indicies
selected_features = Utility.createSelectedFeatures(
all_features, selected_indices_for_features)
print "Selected feature dataset size:", np.shape(selected_features)
print "Glimpse at selected features (11th entry in label list): \n", selected_features[
glimpseIndex]
print "-" * 50
fold2Use = 10
'''
Single iteration zone : turn off 'performIterativeModeling()'
while running this
'''
# this method runs the classifiers once
sklearn_models.performModeling(selected_features, all_labels, fold2Use)
print "-" * 50
'''
Multiple iteration zone : turn off 'performModeling()'
while running this
'''
# this method runs the classifiers 'iteration' number of times
# iteration=1000
# sklearn_models.performIterativeModeling(selected_features, all_labels, fold2Use, iteration)
# print "-"*50
print "Ended at:", Utility.giveTimeStamp()