def learnerTest( naTrain, naTest ):
'''
@summary: Takes testing and training data and computes average error over the test set
This is compared to a baseline guess which is just the average of the training set
'''
llRange = range(5,51,5)
lfRes = []
for lK in llRange:
cLearn = ftu.createKnnLearner( naTrain, lKnn=lK )
fError = 0.0
naResult = cLearn.query( naTest[:,:-1] )
naError = abs( naResult - naTest[:,-1] )
lfRes.append( np.average(naError) )
''' Generate error of 'dumb' case, just assume average returns every time '''
naGuess = np.ones( naTest.shape[0] ) * np.average( naTrain[:,-1] )
lfGuess = [ np.average( abs(naGuess - naTest[:,-1]) ) ] * len(lfRes)
fAvgRets = np.average(naTest[:,-1])
plt.clf()
plt.plot( llRange, lfRes )
plt.plot( llRange, lfGuess )
plt.title( 'Average error on average returns of %.04lf'%fAvgRets )
plt.legend( ('Learner Predict', 'Average Return Predict') )
plt.xlabel('K value')
plt.ylabel('Error')
#plt.show()
plt.savefig( 'FeatureTest.png', format='png' )
def learnerTest(naTrain, naTest):
'''
@summary: Takes testing and training data and computes average error over the test set
This is compared to a baseline guess which is just the average of the training set
'''
llRange = range(5, 51, 5)
lfRes = []
for lK in llRange:
cLearn = ftu.createKnnLearner(naTrain, lKnn=lK, leafsize=100)
fError = 0.0
naResult = cLearn.query(naTest[:, :-1])
naError = abs(naResult - naTest[:, -1])
lfRes.append(np.average(naError))
''' Generate error of 'dumb' case, just assume average returns every time '''
naGuess = np.ones(naTest.shape[0]) * np.average(naTrain[:, -1])
lfGuess = [np.average(abs(naGuess - naTest[:, -1]))] * len(lfRes)
fAvgRets = np.average(naTest[:, -1])
plt.clf()
plt.plot(llRange, lfRes)
plt.plot(llRange, lfGuess)
plt.title('Average error on average returns of %.04lf' % fAvgRets)
plt.legend(('Learner Predict', 'Average Return Predict'))
plt.xlabel('K value')
plt.ylabel('Error')
plt.show()
plt.savefig('FeatureTest.png', format='png')
def learnerTest( naTrain, naTest ):
kdtlearner = ftu.createKnnLearner( naTrain, 5 )
naResult = kdtlearner.query( naTest[:,:-1] )
correlation = np.corrcoef(naResult,naTest[:,-1])
return correlation
def learnerTest( naTrain, naTest ): #create the learner with the train set with K=5 cLearn = ftu.createKnnLearner( naTrain, lKnn=5 ) #get the Y values predicted by the learner Ypredicted = cLearn.query( naTest[:,:-1] ) #get the actual Y values Y = naTest[:,-1] #calculate the correlation coefficient corrcoef = np.corrcoef(Y,Ypredicted)[0][1] #return the corrcoef return corrcoef
def learnerTest(naTrain, naTest):
#create the learner with the train set with K=5
cLearn = ftu.createKnnLearner(naTrain, lKnn=5)
#get the Y values predicted by the learner
Ypredicted = cLearn.query(naTest[:, :-1])
#get the actual Y values
Y = naTest[:, -1]
#calculate the correlation coefficient
corrcoef = np.corrcoef(Y, Ypredicted)[0][1]
#return the corrcoef
return corrcoef
