#----------------------------------
# NNLS fitting
#----------------------------------
from modeling import nnls_parser
obj_nnls = nnls_parser.BASIC_NNLS(ts_matrix, ts_matrix_title)
fout = open('result_r2.dat', 'a')
fout.write("%s\t%.6f\n" %(settings.TIME_SLOT, obj_nnls.fitted_r2))
fout.close()
np.save('../tmp_data/' + settings.TIME_SLOT+'_obj.npy', obj_nnls)
if pred_folders is not None:
# now have the new selected_symbols
selected_symbols = ts_matrix_title[1:-1]
# use the selected_symbols to get the prediction matrix
pr_stock_matrix = prep_matrix.parser_for_prediction(pred_folders,
selected_symbols, start=settings.Time_Slot, end=settings.E_PRED_DATE)
pr_hs300_matrix = prep_matrix.parser_for_hs300(start=settings.Time_Slot, end=settings.E_PRED_DATE)
pr_matrix = ts_format.ts2npy(logic.merge(pr_stock_matrix, pr_hs300_matrix))
np.save('../tmp_data/'+settings.TIME_SLOT+'_pred.npy', pr_matrix)
obj_nnls.predict(pr_matrix)
def test():
#part2()
import cProfile
cProfile.run('part2()')
def make_integrity(ts_matrix, matrix_title, integrity_type=BIND_TYPE, eliminate_rule=None):
''' convert the ts_matrix to integrity one, use integrity_type
to set up different merge type
------------------------------
integrity_type = 0 : merge all
integrity_type = 1 : merge by eliminate partial columns first, then rows
integrity_type = 2 : merge by eliminate partial rows first, then columns
integrity_type = 3 : merge by eliminate partial columns only
integrity_type = 4 : merge by eliminate partial rows only
------------------------------
'''
matrix = ts_format.ts2npy(ts_matrix) #convert ts matrix to npy
nrows, ncols = matrix.shape
if eliminate_rule is None:
eliminate_or_not = lambda x: x.count('NA') > (len(x)/NA_N_PART)
else:
eliminate_or_not = eliminate_rule
# the function to eliminate the corresponding titles from checker
eliminate_matrix_title = lambda checker: [matrix_title[i] for i in checker]
if integrity_type == 0:
matrix = matrix[~(matrix=='NA').any(axis=1), :] # merge all
eliminated_matrix_title = matrix_title[:]
if 0 in matrix.shape:
print 'Error: Empty integral matrix, change another BIND_TYPE'
exit()
elif integrity_type == 1:
# start from cols, first column is the timeline
checker = [0] # don't eliminate first column
for i in range(1, ncols-1):
if not eliminate_or_not(matrix[:,i].tolist()):
checker.append(i)
else:
pass
pass
checker.append(ncols-1) # don't eliminate the last column
matrix = matrix[:, checker] # remove columns
matrix = matrix[~(matrix=='NA').any(axis=1), :] # remove rows
eliminated_matrix_title = eliminate_matrix_title(checker)
if matrix.shape[0] < 3 or matrix.shape[1] < 3:
print 'Error: Empty integral matrix, change another BIND_TYPE'
exit()
pass
elif integrity_type == 2:
checker = []
for i in range(nrows):
if not eliminate_or_not(matrix[i, :].tolist()): checker.append(i)
#print checker
matrix = matrix[checker, :]
matrix = matrix[:, ~(matrix=='NA').any(axis=0)]
eliminated_matrix_title = \
eliminate_matrix_title(~(matrix=='NA').any(axis=0))
if matrix.shape[0] < 3 or matrix.shape[1] < 3:
print 'Error: Nearly Empty integral matrix, change another BIND_TYPE'
exit()
pass
elif integrity_type == 3:
checker = [0]
for i in range(1, ncols-1):
if not eliminate_or_not(matrix[:,i].tolist()): checker.append(i)
checker.append(ncols-1)
#print checker
matrix = matrix[:, checker] # remove columns only
eliminated_matrix_title = eliminate_matrix_title(checker)
pass
elif integrity_type == 4:
checker = []
for i in range(nrows):
if not eliminate_or_not(matrix[i, :].tolist()): checker.append(i)
#print checker
matrix = matrix[checker, :]
eliminated_matrix_title = matrix_title[:]
pass
return matrix, eliminated_matrix_title
