def create_newdata(header, x):
dlist = []
cnt = 0
tmp = 0
hlist = []
if (V):
print('Reducing attributes by ' + str(options['turfpct']) + '%')
sys.stdout.flush()
for a in table:
if (cnt >= keepcnt):
lost[a[0]] = iteration + 1
hlist.append(a[0]) # append lost attribe names to hlist
i = header.index(a[0])
dlist.append(i)
cnt += 1
header = np.delete(header, dlist).tolist() #remove orphans from header
x = np.delete(x, dlist, axis=1) #remove orphaned attributes from data
x = np.ascontiguousarray(x, dtype=np.double)
if (V):
print('Getting new variables, attributes and distance array')
sys.stdout.flush()
var = cmn.getVariables(header, x, y, options)
attr = cmn.getAttributeInfo(header, x, var, options)
if (V):
print("--------------- Parameters ---------------")
print("datatype: " + var['dataType'])
print("attributes: " + str(var['NumAttributes']))
if (var['dataType'] == 'mixed'):
print(" continuous: " + str(var['cpct'][1]))
print(" discrete: " + str(var['dpct'][1]))
if (var['mdcnt'] > 0):
print("missing: " + str(var['mdcnt']))
print("--------------------------------------------")
sys.stdout.flush()
begin = tm.time()
diffs, cidx, didx = cmn.dtypeArray(header, attr, var)
if (var['mdcnt'] > 0):
import mmDistance as md
distArray = md.getDistances(x[:, cidx], x[:, didx], var,
diffs[cidx])
disttype = "missing"
else:
distArray = cmn.getDistances(x, attr, var, cidx, didx)
disttype = "discrete/continuous/mixed"
if (V):
print(disttype + " distance array elapsed time(sec) = " +
str(tm.time() - begin))
sys.stdout.flush()
return header, x, attr, var, distArray, lost
if (var['classType'] == 'multiclass'):
yset = var['phenoTypeList']
print(" classes: " + str(len(yset)))
print("classname: " + var['phenoTypeName'])
print("algorithm: " + options['algorithm'])
print("--------------------------------------------")
sys.stdout.flush()
#-----------------------------------------------------------------------------#
# create distance array and remove intermediate data
# if missing and/or mixed data use the mixedDistance function
#
begin = tm.time()
diffs, cidx, didx = cmn.dtypeArray(header, attr, var)
if (var['mdcnt'] > 0):
import mmDistance as md
distArray = md.getDistances(x[:, cidx], x[:, didx], var, diffs[cidx])
disttype = "missing"
else:
distArray = cmn.getDistances(x, attr, var, cidx, didx, cheader)
disttype = "discrete/continuous/mixed"
if (V):
ctime = "[" + tm.strftime("%H:%M:%S") + "]"
print(ctime + " " + disttype + " distance array time(sec) = " +
str(tm.time() - begin))
sys.stdout.flush()
#############################################################################
###################################################################################################################################################
def test_relieff_Multiplexer():
""" Test ReliefF on 6-bit Multiplexer"""
yset = var['phenoTypeList']
print(" classes: " + str(len(yset)))
print("datatype: " + var['dataType'])
print("classname: " + var['phenoTypeName'])
print("algorithm: " + options['algorithm'])
print("--------------------------------------------")
sys.stdout.flush()
#-----------------------------------------------------------------------------#
# create distance array and remove intermediate data
# if missing and/or mixed data use the mixedDistance function
#
begin = tm.time()
diffs, cidx, didx = cmn.dtypeArray(header, attr, var)
if(var['mdcnt'] > 0):
import mmDistance as md
distArray = md.getDistances(x[:,cidx], x[:,didx], var, diffs[cidx])
disttype = "missing"
else:
distArray = cmn.getDistances(x, attr, var, cidx, didx, cheader)
disttype = "discrete/continuous/mixed"
if(V):
ctime = "[" + tm.strftime("%H:%M:%S") + "]"
print(ctime + " " + disttype + " distance array time(sec) = "
+ str(tm.time()-begin))
sys.stdout.flush()
#-----------------------------------------------------------------------------#
# get Scores based on algorithm selected (-a and -t)
#
if(turfpct > 0): # Use TURF
import Turf as T
def create_newdata(header, x):
dlist = []
cnt = 0
if(V):
print('Reducing attributes by ' + str(options['turfpct']) + '%')
sys.stdout.flush()
#Go through table with feature sorted by decreasing scores, once we hit keepcnt, we start adding to lost.
for a in table:
if(cnt >= keepcnt):
lost[a[0]] = iteration + 1
i = header.index(a[0])
dlist.append(i) #store position of each feature removed in dlist.
cnt += 1
#update header and dataset to reflect removal of lowest scoring features.
header = np.delete(header,dlist).tolist() #remove orphans from header
x = np.delete(x,dlist,axis=1) #remove orphaned attributes from data
x = np.ascontiguousarray(x, dtype=np.double)
if(V):
print('Getting new variables, attributes and distance array')
sys.stdout.flush()
#Redo data survey (which may save time in downstream distance array calculation (depending on dataset)
var = cmn.getVariables(header, x, y, options)
attr = cmn.getAttributeInfo(header, x, var, options)
cheader = []
for i in header:
if attr[i][0] == 'continuous':
cheader.append(i)
if(V):
print("--------------- Parameters ---------------")
print("datatype: " + var['dataType'])
print("attributes: " + str(var['NumAttributes']))
if(var['dataType'] == 'mixed'):
print(" continuous: " + str(var['cpct'][1]))
print(" discrete: " + str(var['dpct'][1]))
if(var['mdcnt'] > 0):
print("missing: " + str(var['mdcnt']))
print("--------------------------------------------")
sys.stdout.flush()
begin = tm.time()
diffs, cidx, didx = cmn.dtypeArray(header, attr, var)
#Calculate distance array based on present feature types and data missingness.
if(var['mdcnt'] > 0):
import mmDistance as md
distArray = md.getDistances(x[:,cidx], x[:,didx], var, diffs[cidx])
disttype = "missing"
else:
distArray = cmn.getDistances(x, attr, var, cidx, didx, cheader)
disttype = "discrete/continuous/mixed"
if(V):
print(disttype + " distance array elapsed time(sec) = "
+ str(tm.time()-begin))
sys.stdout.flush()
return header, x, attr, var, distArray, lost