def theProbabilities1(): print '## Graphing the Probabilities 概率密度' reload(numpredict) data1=numpredict.wineset3() numpredict.cumulativegraph(data1,(1,1),120) reload(numpredict) numpredict.probabilitygraph(data1,(1,1),6)
def unevenDistributions(): ''' 对于某些关键特征，数据中并未体现，需要有种方式进行观测,比如部分数据是通过折扣店购买的！''' print '## Uneven Distributions 不对称性' reload(numpredict) data=numpredict.wineset3() print numpredict.wineprice(99.0,20.0) print numpredict.weightedknn(data,[99.0,20.0]) print numpredict.crossvalidate(numpredict.weightedknn,data)
def knn2(d, v):
return numpredict.weightedknn(d, v, k=3)
print(numpredict.crossvalidate(knn1, data))
print(numpredict.crossvalidate(knn2, data))
# Parameter tuning
import optimization
print('<----Parameter tuning---->')
data2 = numpredict.wineset2()
print(numpredict.crossvalidate(knn1, data2))
print(numpredict.crossvalidate(numpredict.weightedknn, data2))
print('<--scaling-->')
sdata = numpredict.rescale(data2, [10, 10, 0, 0.5])
print(numpredict.crossvalidate(knn1, sdata))
print('<--optimizing-->')
costf = numpredict.createcostfunction(numpredict.knnestimate, data2)
print(optimization.geneticoptimize(numpredict.weightdomain, costf, step=2))
# Probability Density
print('\n<----Probability Density---->')
data3 = numpredict.wineset3()
numpredict.cumulativegraph(data3, (99, 20), 120)
numpredict.probabilitygraph(data3, (99, 20), 120)
# eBay API
# remaining problems
# import ebaypredict
# laptops = ebaypredict.doSearch('laptop')
# print(laptops[:10])
]
# gaussian smooth with nearby points
smoothed = []
for i in range(len(probs)): # O(n^2)
sv = 0.0
for j in range(0, len(probs)):
dist = abs(i - j) * 0.1
weight = numpredict.gaussianweight(dist, sigma=ss)
sv += weight * probs[j]
smoothed.append(sv)
plot(t1, array(probs))
plot(t1, array(smoothed))
if __name__ == '__main__':
s = numpredict.wineset3(k=500)
import pprint
pprint.pprint(s)
#wine = [99.0, 20.0] # choose params were the sample data is somewhat dense
wine = [99.0, 20.0] # choose params were the sample data is somewhat dense
# Draw graph that shows p(price | wine)
print 'Real price 1:', numpredict.wineprice(wine[0], wine[1])
print 'Real price 2:', 0.6 * numpredict.wineprice(wine[0], wine[1])
cumulativegraph(s, wine, 120)
probabilitygraph(s, wine, 120)
show() # this does user interaction
import numpredict import optimization #print numpredict.wineprice(95.0, 3.0) #data = numpredict.wineset1() #print data[0],data[1] #print numpredict.knnestimate(data,(99.0,50),k = 1) #print numpredict.knnestimate(data,(99.0,50)) #print numpredict.weightedknn(data,(99.0,5.0)) #print numpredict.crossvalidate(numpredict.knnestimate,data) #data = numpredict.wineset2() #costf = numpredict.createcostfunction(numpredict.knnestimate, data) #print optimization.annealingoptimize(numpredict.weightdomain, costf, step = 2) data = numpredict.wineset3() #print numpredict.probguess(data, [99,20],40,80) numpredict.probabilitygraph(data, (99,30), 120)
probs = [numpredict.probguess(data, vec1, v, v+S, k, weightfun) for v in t1]
# gaussian smooth with nearby points
smoothed = []
for i in range(len(probs)): # O(n^2)
sv = 0.0
for j in range(0, len(probs)):
dist = abs(i - j)*0.1
weight = numpredict.gaussianweight(dist, sigma=ss)
sv += weight*probs[j]
smoothed.append(sv)
plot(t1, array(probs))
plot(t1, array(smoothed))
if __name__ == '__main__':
s = numpredict.wineset3(k=500)
import pprint
pprint.pprint(s)
#wine = [99.0, 20.0] # choose params were the sample data is somewhat dense
wine = [99.0, 20.0] # choose params were the sample data is somewhat dense
# Draw graph that shows p(price | wine)
print 'Real price 1:', numpredict.wineprice(wine[0], wine[1])
print 'Real price 2:', 0.6 * numpredict.wineprice(wine[0], wine[1])
cumulativegraph(s, wine, 120)
probabilitygraph(s, wine, 120)
show() # this does user interaction
