def runMultTest(transaction_list, trans4lcm, threshold, set_method, lcm_path, max_comb, outlog, alternative):
max_lambda = maxLambda(transaction_list)
lam_star = 1; func_f = None;
try:
if set_method == "fisher":
func_f = functions4fisher.FunctionOfX(transaction_list, max_lambda, abs(alternative))
elif set_method == "u_test":
func_f = functions4u_test.FunctionOfX(transaction_list, alternative)
elif set_method == "chi":
func_f = functions4chi.FunctionOfX(transaction_list, max_lambda, abs( alternative))
else:
sys.stderr.write("Error: choose \"fisher\", \"chi\" or \"u_test\" by using -p option.\n")
outlog.close()
sys.exit()
lam = max_lambda
# check a MASL of max_lambda
if set_method in BINARY_METHODS:
n1 = func_f.sumValue(transaction_list)
if (n1 < max_lambda):
max_lambda = int( n1 )
lam = int( n1 )
fre_pattern = frequentPatterns.LCM(lcm_path, max_lambda, outlog)
fre_pattern.makeFile4Lem(transaction_list, trans4lcm) # make itemset file for lcm
# If Fisher's exact test or chi-square test is used for computing P-value,
# LCM-LAMP is run to find optimal lambda.
if set_method == "fisher":
neg_size = func_f.getAllSize() - func_f.getN1()
n1 = min( n1, neg_size )
# # of positives == # of negatives, and two.sided hypothesis test.
if ( func_f.getN1() == neg_size ) and (alternative == 0):
fre_pattern, lam_star = depthFirst( trans4lcm, fre_pattern, max_comb, n1, 0.5*threshold, 1 )
else:
fre_pattern, lam_star = depthFirst( trans4lcm, fre_pattern, max_comb, n1, threshold, 1 )
elif set_method == "chi":
neg_size = func_f.getAllSize() - func_f.getN1()
n1 = min( n1, neg_size )
# two-sided hypothesis test
if alternative == 0:
fre_pattern, lam_star = depthFirst( trans4lcm, fre_pattern, max_comb, n1, 0.5*threshold, 2 )
# one-sided
else:
fre_pattern, lam_star = depthFirst( trans4lcm, fre_pattern, max_comb, n1, threshold, 2 )
# If Mann-Whitney U test of Chi-square test is used,
# LAMP ver 1. is run for computing the optimal lambda.
else:
# two-sided hypothesis test
if alternative == 0:
fre_pattern, lam_star = breadthFirst( trans4lcm, fre_pattern, func_f, max_comb, 0.5*threshold, lam, outlog )
# one-sided hypothesis test
else:
fre_pattern, lam_star = breadthFirst( trans4lcm, fre_pattern, func_f, max_comb, threshold, lam, outlog )
except fs.TestMethodError, e:
sys.exit()
sys.exit
except fs.TestMethodError, e:
sys.exit()
try:
lam = max_lambda
# check a MASL of max_lambda
if (set_method == 'fisher') or (set_method == 'chi'):
n1 = func_f.sumValue(transaction_list)
if (n1 < max_lambda):
max_lambda = int(n1)
lam = int(n1)
fre_pattern = frequentPatterns.LCM(lcm_path, max_lambda, outlog)
fre_pattern.makeFile4Lem(transaction_list,
trans4lcm) # make itemset file for lcm
# If file for Lcm exist, comfiem that overwrite the file.
# solve K and lambda
while lam > 1:
outlog.write("--- lambda: " + str(lam) + " ---\n")
# if lambda == 1, all tests which support >= 1 are tested.
if lam == 1:
lam_star = lam
fre_pattern.frequentPatterns(trans4lcm, lam,
max_comb) # line 3 of Algorithm
k = fre_pattern.getTotal(lam)
break
fre_pattern.frequentPatterns(trans4lcm, lam,
def runMultTest(transaction_list, trans4lcm, threshold, set_method, lcm_path,
max_comb, outlog, alternative):
max_lambda = maxLambda(transaction_list)
lam_star = 1
func_f = None
try:
if set_method == "fisher":
func_f = functions4fisher.FunctionOfX(transaction_list, max_lambda,
abs(alternative))
elif set_method == "u_test":
func_f = functions4u_test.FunctionOfX(transaction_list,
alternative)
elif set_method == "chi":
func_f = functions4chi.FunctionOfX(transaction_list, max_lambda,
abs(alternative))
else:
sys.stderr.write(
"Error: choose \"fisher\", \"chi\" or \"u_test\" by using -p option.\n"
)
outlog.close()
sys.exit()
lam = max_lambda
# check a MASL of max_lambda
if set_method in BINARY_METHODS:
n1 = func_f.sumValue(transaction_list)
if (n1 < max_lambda):
max_lambda = int(n1)
lam = int(n1)
fre_pattern = frequentPatterns.LCM(lcm_path, max_lambda, outlog)
fre_pattern.makeFile4Lem(transaction_list,
trans4lcm) # make itemset file for lcm
# If Fisher's exact test or chi-square test is used for computing P-value,
# LCM-LAMP is run to find optimal lambda.
if set_method == "fisher":
neg_size = func_f.getAllSize() - func_f.getN1()
n1 = min(n1, neg_size)
# # of positives == # of negatives, and two.sided hypothesis test.
if (func_f.getN1() == neg_size) and (alternative == 0):
fre_pattern, lam_star = depthFirst(trans4lcm, fre_pattern,
max_comb, n1,
0.5 * threshold, 1)
else:
fre_pattern, lam_star = depthFirst(trans4lcm, fre_pattern,
max_comb, n1, threshold, 1)
elif set_method == "chi":
neg_size = func_f.getAllSize() - func_f.getN1()
n1 = min(n1, neg_size)
# two-sided hypothesis test
if alternative == 0:
fre_pattern, lam_star = depthFirst(trans4lcm, fre_pattern,
max_comb, n1,
0.5 * threshold, 2)
# one-sided
else:
fre_pattern, lam_star = depthFirst(trans4lcm, fre_pattern,
max_comb, n1, threshold, 2)
# If Mann-Whitney U test of Chi-square test is used,
# LAMP ver 1. is run for computing the optimal lambda.
else:
# two-sided hypothesis test
if alternative == 0:
fre_pattern, lam_star = breadthFirst(trans4lcm, fre_pattern,
func_f, max_comb,
0.5 * threshold, lam,
outlog)
# one-sided hypothesis test
else:
fre_pattern, lam_star = breadthFirst(trans4lcm, fre_pattern,
func_f, max_comb,
threshold, lam, outlog)
except fs.TestMethodError as e:
sys.exit()
except frequentPatterns.LCMError as e:
sys.exit()
try:
fre_pattern.frequentPatterns(trans4lcm, lam_star,
max_comb) # P_lambda* at line 13
k = fre_pattern.getTotal(lam_star)
except frequentPatterns.LCMError as e:
sys.exit()
# multiple test by using k and lambda_star
outlog.write("finish calculation of K: %s\n" % k)
# If lam_star > max_lambda, m_lambda set to max_lambda.
# This case cause when optimal solution is found at first step.
outlog.write("%s\n" % lam_star)
if (lam_star > max_lambda):
lam_star = max_lambda
correction_term_time = time.time()
return (fre_pattern, lam_star, max_lambda, correction_term_time, func_f)