def loop_all_combinations_for(dep, correlations, dataset):
best_fit = ('', 0, 0, 9999, 9999, '', [0], [0], 0, (0, 0), 0)
highest = ('', 0, 0, 9999, 9999, '', [0], [0], 0, 0, 0)
subset = correlations
for pca_n in range(configuration.pca_min_n,
min(configuration.pca_max_n + 1, len(correlations))):
print(" generating pca depth ", pca_n)
new = regression(dep, subset, dataset, pca_n)
#util.display_result(dep + " current = ", new)
#print(new[9][0])
# saving if rsquared is better
if (new[1] > highest[1]):
highest = new
# saving if fitness is better
if (new[4] < best_fit[4]):
best_fit = new
# saving if fitness is better
# if(new[9][0] > best_fit[9][0]):
# best_fit = new
util.display_result(dep + " highest squared", highest)
util.display_result(dep + " best fit", best_fit)
def loop_all_combinations_for(dep, correlations, dataset):
best_fit = ('', 0, 0, 9999, 9999, '', [0], [0], 0, (0,0), 0)
highest = ('', 0, 0, 9999, 9999, '', [0], [0], 0, 0, 0)
for length in range(2, min(len(correlations[0]) + 1, configuration.max_dept_logreg +1)):
print(" generating length ", length)
for subset in itertools.combinations(correlations[0], length):
for pca_n in range(configuration.pca_min_n, min(configuration.pca_max_n + 1, len(subset)+1)):
#print(" generating pca depth ", pca_n)
new = nb_model(dep, subset, dataset, pca_n)
#util.display_result(dep + " current = ", new)
#print(new[9][0])
# saving if rsquared is better
if(new[1] > highest[1]):
highest = new
# saving if fitness is better
# if(new[4] < best_fit[4]):
# best_fit = new
# saving if fitness is better
if(new[9][0] > best_fit[9][0]):
best_fit = new
util.display_result(dep + " highest squared", highest)
util.display_result(dep + " best fit", best_fit)
def loop_all_combinations_for(dep, correlations, dataset):
# initialize vars with dummies and sentinels
highest = ('', 0, 0, 9999, 9999, '', [0], [0], 0, 0, 0)
highest_adj = ('', 0, 0, 9999, 9999, '', [0], [0], 0, 0, 0)
best_fit = ('', 0, 0, 9999, 9999, '', [0], [0], 0, 0, 0)
# first generate model with all independent variables
# generating model with all independents
new = regression(dep, configuration.independent, dataset)
highest = new
highest_adj = new
best_fit = new
util.display_result("model with all independents", new)
# second: generate model with only the major independent variables
# generating model with all independents
new = regression(dep, configuration.independent_major, dataset)
util.display_result("model with only major independents", new)
print("creating all combinations for ", dep, "of :")
print(correlations[0])
# generate all combinations for the dimensions that are potentially correlation for dimension <dep>
for length in range(1, min(len(correlations[0]) + 1, configuration.max_depth +1)):
print(" generating length ", length)
for subset in itertools.combinations(correlations[0], length):
independents = " + ".join(subset)
#print(" ", dep, " => ", independents)
# new contains metrics as a tuple
# regression generates models and searches for best fit, highest rsquared, and rsquared_adjusted
new = regression(dep, subset, dataset)
if(new[1] > highest[1]):
highest = new
if(new[2] > highest_adj[2]):
highest_adj = new
if(new[9][0] > best_fit[9][0]):
best_fit = new
util.display_result(dep + " highest squared", highest)
util.display_result(dep + " highest squared_adj", highest_adj)
util.display_result(dep + " best fit", best_fit)
def loop_all_combinations_for(dep, correlations, dataset):
highest = ('', 0, 0, 9999, 9999, '', [0], [0], 0, (0,0), 0)
best_fit = ('', 0, 0, 9999, 9999, '', [0], [0], 0, (0,0), 0)
for length in range(2, min(len(correlations[0]) + 1, configuration.max_dept_poly +1)):
print(" generating length ", length)
for subset in itertools.combinations(correlations[0], length):
for degree in range(configuration.pol_min_degree, configuration.pol_max_degree + 1):
new = regression(dep, subset, dataset, degree)
#util.display_result(dep, new)
# saving if rsquared is better
if(new[1] > highest[1]):
highest = new
# saving if fitness is better
if(new[9][0] > best_fit[9][0]):
best_fit = new
util.display_result(dep + " highest squared", highest)
util.display_result(dep + " best fit", best_fit)
def loop_all_combinations_for(dep, correlations, dataset):
independents = []
independents_all = []
best_fit = ('', 0, 0, 9999, 9999, '', [0], [0], 0)
# first put all possible degree combinations in list ...
for indep in correlations[0]:
independents_all.append(indep)
independents.append(indep)
if not util.hasNumbers(indep):
for degree in range(configuration.pol_min_degree,
configuration.pol_max_degree + 1):
independents_all.append(indep + " ** " + str(degree))
#print(independents)
#print(independents_all)
# ... and now create every combination
#print("independents_all = ", independents_all)
for length in range(
1, min(len(independents_all) + 1,
configuration.max_dept_poly + 1)):
print(" generating length ", length)
for subset in itertools.combinations(independents_all, length):
#print(subset)
new = regression(dep, subset, dataset, independents)
util.display_result(dep, new)
if (new[4] < best_fit[4]):
best_fit = new
print("subset = ", subset)
util.display_result(dep, new)
util.display_result(dep + " best fit", best_fit)