def mean_aggregation(fold):
train_df, train_labels, test_df, test_labels = common.read_fold(path, fold)
ids = test_df.index.get_level_values('id')
labels = test_df.index.get_level_values('label')
predictions = test_df.mean(axis = 1)
diversity = common.diversity_score(test_df.values)
return DataFrame({'id': ids, 'label': labels, 'fold': fold, 'prediction': predictions, 'diversity': diversity})
def select_candidate_sdi(train_df, train_labels, best_classifiers, ensemble, i):
if len(ensemble) >= initial_ensemble_size:
candidates = choice(best_classifiers.index.values, min(max_candidates, len(best_classifiers)), replace = False)
candidate_diversity_scores = [1 - abs(common.diversity_score(train_df[ensemble + [candidate]].values)) for candidate in candidates] # 1 - kappa so larger = more diverse
candidate_scores = [accuracy_weight * best_classifiers.ix[candidate] + (1 - accuracy_weight) * candidate_diversity_scores[candidate_i] for candidate_i, candidate in enumerate(candidates)]
best_candidate = candidates[common.argbest(candidate_scores)]
else:
best_candidate = best_classifiers.index.values[i]
return best_candidate
def select_candidate_drep(train_df, train_labels, best_classifiers, ensemble, i):
if len(ensemble) >= initial_ensemble_size:
candidates = choice(best_classifiers.index.values, min(max_candidates, len(best_classifiers)), replace = False)
candidate_diversity_scores = [abs(common.diversity_score(train_df[ensemble + [candidate]].values)) for candidate in candidates]
candidate_diversity_ranks = array(candidate_diversity_scores).argsort()
diversity_candidates = candidates[candidate_diversity_ranks[:max_diversity_candidates]]
candidate_accuracy_scores = [common.score(train_labels, train_df[ensemble + [candidate]].mean(axis = 1)) for candidate in diversity_candidates]
best_candidate = candidates[common.argbest(candidate_accuracy_scores)]
else:
best_candidate = best_classifiers.index.values[i]
return best_candidate
def mean_aggregation(fold):
train_df, train_labels, test_df, test_labels = common.read_fold(path, fold)
ids = test_df.index.get_level_values('id')
labels = test_df.index.get_level_values('label')
predictions = test_df.mean(axis=1)
diversity = common.diversity_score(test_df.values)
return DataFrame({
'id': ids,
'label': labels,
'fold': fold,
'prediction': predictions,
'diversity': diversity
})
def stacked_generalization(stacker, fold):
train_df, train_labels, test_df, test_labels = common.read_fold(path, fold)
train_df = common.unbag(train_df, 10)
test_df = common.unbag(test_df, 10)
test_predictions = stacker.fit(train_df,
train_labels).predict_proba(test_df)[:, 1]
return DataFrame({
'fold': fold,
'id': test_df.index.get_level_values('id'),
'label': test_labels,
'prediction': test_predictions,
'diversity': common.diversity_score(test_df.values)
})
def get_predictions(df, ensemble, fold, seedval):
ids = df.index.get_level_values('id')
labels = df.index.get_level_values('label')
predictions = df[ensemble].mean(axis=1)
diversity = common.diversity_score(df[ensemble].values)
return DataFrame({
'fold': fold,
'seed': seedval,
'id': ids,
'label': labels,
'prediction': predictions,
'diversity': diversity,
'ensemble_size': len(ensemble)
})
def select_candidate_sdi(train_df, train_labels, best_classifiers, ensemble,
i):
if len(ensemble) >= initial_ensemble_size:
candidates = choice(best_classifiers.index.values,
min(max_candidates, len(best_classifiers)),
replace=False)
candidate_diversity_scores = [
1 -
abs(common.diversity_score(
train_df[ensemble + [candidate]].values))
for candidate in candidates
] # 1 - kappa so larger = more diverse
candidate_scores = [
accuracy_weight * best_classifiers.ix[candidate] +
(1 - accuracy_weight) * candidate_diversity_scores[candidate_i]
for candidate_i, candidate in enumerate(candidates)
]
best_candidate = candidates[common.argbest(candidate_scores)]
else:
best_candidate = best_classifiers.index.values[i]
return best_candidate
def stacked_selection(fold):
seed(seedval)
train_df, train_labels, test_df, test_labels = common.read_fold(path, fold)
train_distances = 1 - train_df.corr().abs()
train_performance = []
test_performance = []
for n_clusters in range(1, max_clusters + 1):
train_values, train_predictions = stack_function(n_clusters,
train_distances,
train_df,
train_labels,
predict_df=train_df)
test_values, test_predictions = stack_function(n_clusters,
train_distances,
train_df,
train_labels,
predict_df=test_df)
train_performance.append(
get_cluster_performance(train_labels, train_predictions,
n_clusters, fold, seedval))
test_performance.append(
get_cluster_performance(test_labels, test_predictions, n_clusters,
fold, seedval))
best_cluster_size = common.get_best_performer(
DataFrame.from_records(train_performance)).n_clusters.values
test_values, test_predictions = stack_function(best_cluster_size,
train_distances,
train_df,
train_labels,
predict_df=test_df)
return DataFrame({
'fold': fold,
'seed': seedval,
'id': test_df.index.get_level_values('id'),
'label': test_labels,
'prediction': test_predictions,
'diversity': common.diversity_score(test_values),
'metric': common.score.__name__
}), DataFrame.from_records(test_performance)
def select_candidate_drep(train_df, train_labels, best_classifiers, ensemble,
i):
if len(ensemble) >= initial_ensemble_size:
candidates = choice(best_classifiers.index.values,
min(max_candidates, len(best_classifiers)),
replace=False)
candidate_diversity_scores = [
abs(common.diversity_score(train_df[ensemble +
[candidate]].values))
for candidate in candidates
]
candidate_diversity_ranks = array(candidate_diversity_scores).argsort()
diversity_candidates = candidates[
candidate_diversity_ranks[:max_diversity_candidates]]
candidate_accuracy_scores = [
common.score(train_labels,
train_df[ensemble + [candidate]].mean(axis=1))
for candidate in diversity_candidates
]
best_candidate = candidates[common.argbest(candidate_accuracy_scores)]
else:
best_candidate = best_classifiers.index.values[i]
return best_candidate
def stacked_generalization(fold):
train_df, train_labels, test_df, test_labels = common.read_fold(path, fold)
if method == 'aggregate':
train_df = common.unbag(train_df, bag_count)
test_df = common.unbag(test_df, bag_count)
test_predictions = stacker.fit(train_df, train_labels).predict_proba(test_df)[:, 1]
return DataFrame({'fold': fold, 'id': test_df.index.get_level_values('id'), 'label': test_labels, 'prediction': test_predictions, 'diversity': common.diversity_score(test_df.values)})
def get_predictions(df, ensemble, fold, seedval):
ids = df.index.get_level_values('id')
labels = df.index.get_level_values('label')
predictions = df[ensemble].mean(axis = 1)
diversity = common.diversity_score(df[ensemble].values)
return DataFrame({'fold': fold, 'seed': seedval, 'id': ids, 'label': labels, 'prediction': predictions, 'diversity': diversity, 'ensemble_size': len(ensemble)})
def stacked_selection(fold):
seed(seedval)
train_df, train_labels, test_df, test_labels = common.read_fold(path, fold)
train_distances = 1 - train_df.corr().abs()
train_performance = []
test_performance = []
for n_clusters in range(1, max_clusters + 1):
train_values, train_predictions = stack_function(n_clusters, train_distances, train_df, train_labels, predict_df = train_df)
test_values, test_predictions = stack_function(n_clusters, train_distances, train_df, train_labels, predict_df = test_df)
train_performance.append(get_cluster_performance(train_labels, train_predictions, n_clusters, fold, seedval))
test_performance.append(get_cluster_performance(test_labels, test_predictions, n_clusters, fold, seedval))
best_cluster_size = common.get_best_performer(DataFrame.from_records(train_performance)).n_clusters.values
test_values, test_predictions = stack_function(best_cluster_size, train_distances, train_df, train_labels, predict_df = test_df)
return DataFrame({'fold': fold, 'seed': seedval, 'id': test_df.index.get_level_values('id'), 'label': test_labels, 'prediction': test_predictions, 'diversity': common.diversity_score(test_values), 'metric': common.score.__name__}), DataFrame.from_records(test_performance)