def returnXy(mode=0, pX=None, index=None, K = TRAINK): temp = returnDataA(1, 0)[3] L = len(temp) start = 0 ed = int(L * K) X = [] y = [] for ci in range(0, CAN, 1): for xi in range(0, ed-start, 1): X.append([]) for fi in range(TN): arr = returnDataA(ci, fi)[3][start:ed] for xi in range(0, ed-start, 1): X[(ci)*(ed-start)+xi].append(float(arr[xi])) cop = [] cop = returnDataA(ci, TCOP)[3][start:ed] for yi in range(0, ed-start, 1): y.append(cop[yi]) return X, y
def returnXy(mode=0, pX=None, index=None, K = TRAINK):
#print ' K =', K, ' Mode =' , mode
temp = returnDataA(1, 0)[3]
#train index: 0 to L * K
#predict index: L * K to L
L = len(temp)
if mode == 5: #single sample with index
st = index
ed = index + 1
else:
if mode == 0:
st = 0
ed = int(L * K)
elif mode == 1:
st = int(L * K)
ed = L
elif mode == 2:
st = 0
ed = L
elif mode == 3 or mode == 4 or mode == 6 or mode == 7:
st = 0
ed = int(L * K)
#merely chiller 1
X = []
for xi in range(0, ed-st, 1):
X.append([])
for fi in range(TN):
arr = returnDataA(1, fi)[3][st:ed]
for xi in range(0, ed-st, 1):
X[xi].append(arr[xi])
y = []
cop = []
cop = returnDataA(1, TCOP)[3][st:ed]
for yi in range(0, ed-st, 1):
y.append(cop[yi])
#clustered samples
if mode == 3 or mode == 4 or mode == 6 or mode == 7:
X0 = []
y0 = []
X0 = deepcopy(X)
y0 = deepcopy(y)
#http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsRegressor.html
if mode == 3: #single chiller clustering
#print 'clustered X, y' ###
#print X[1:5], y[1:5] ###
#print ###
#from sklearn.neighbors import NearestNeighbors
neigh = NearestNeighbors()
neigh.fit(X0)
d, xis = neigh.kneighbors([pX])
X = []
y = []
for i in range(len(xis)):
for xi in xis[i]:
#print xi
#print X0[xi]
X.append(X0[xi])
y.append(y0[xi]) #share the common index
elif mode == 4: #clustering chillers of the same type
#CN = CSN #chillers of the same type
X = []
y = []
Xt = []
yt = []
for ci in range(CSN):
for xi in range(0, ed-st, 1):
Xt.append([])
for fi in range(TN):
arr = returnDataA(ci, fi)[3][st:ed]
for xi in range(0, ed-st, 1):
Xt[ci*(ed-st)+xi].append(arr[xi])
cop = []
cop = returnDataA(ci, TCOP)[3][st:ed]
for yi in range(0, ed-st, 1):
yt.append(cop[yi])
#from sklearn.neighbors import NearestNeighbors
neigh = NearestNeighbors()#
neigh.fit(Xt)
X0 = []
y0 = []
X0 = deepcopy(Xt)
y0 = deepcopy(yt)
d, xis = neigh.kneighbors([pX])
for i in range(len(xis)):
for xi in xis[i]:
#print xi
#print X0[xi]
X.append(X0[xi])
y.append(y0[xi]) #share the common index
elif mode == 6: #clustering chillers of all types
X = []
y = []
for ci in range(0, CAN, 1):
for xi in range(0, ed-st, 1):
X.append([])
for fi in range(TN):
arr = returnDataA(ci, fi)[3][st:ed]
for xi in range(0, ed-st, 1):
X[(ci)*(ed-st)+xi].append(arr[xi])
cop = []
cop = returnDataA(ci, TCOP)[3][st:ed]
for yi in range(0, ed-st, 1):
y.append(cop[yi])
elif mode == 7: #clustering all chillers
X = []
y = []
Xt = []
yt = []
for ci in range(CAN):
for xi in range(0, ed-st, 1):
Xt.append([])
for fi in range(TN):
arr = returnDataA(ci, fi)[3][st:ed]
for xi in range(0, ed-st, 1):
Xt[ci*(ed-st)+xi].append(arr[xi])
cop = []
cop = returnDataA(ci, TCOP)[3][st:ed]
for yi in range(0, ed-st, 1):
yt.append(cop[yi])
kmeans = KMeans(n_clusters=5)#n_clusters=2, random_state=0
XtCs = kmeans.fit_predict(Xt)
X0 = []
y0 = []
X0 = deepcopy(Xt)
y0 = deepcopy(yt)
pXCs = kmeans.predict([pX])
pXC = pXCs[0] #which cluster is pX
#print X0
#print y0
for xi in range(len(XtCs)):
if XtCs[xi] == pXC: #take the samples from the same cluster
#print xi
#print X0[xi]
X.append(X0[xi])
y.append(y0[xi]) #share the common index
#print 'Length', len(X), len(y)
return X, y