import pandas as pd
#=====================META-FEATURE EXTRACTION==================================
with open(str(p.parents[5]) + '/actual/wine_metafeatures_202.pickle',
'rb') as handle:
meta_features = pickle.load(handle)
#============Load the meta-dataset and model performance of each dataset=======
df_results = pd.read_pickle(str(p.parents[7]) + '/test/df_results.plk')
#nested_results is a nested dictionary with all the AUC-ROC performances for each dataset and all models
with open(str(p.parents[6]) + '/nested_results_prc.pickle', 'rb') as handle:
nested_results_prc = pickle.load(handle)
#========================META-LEARNING: RANKING================================
#KNN Ranking Method
top1, top2, top3 = KNN_ranking(df_results, meta_features, nested_results_prc)
print("==========================================")
print(" AUC-PRC ")
print("==========================================")
print("Top 1 predicted model: " + top1)
print("Top 2 predicted model: " + top2)
print("Top 3 predicted model: " + top3)
#Actual results
with open(str(p.parents[5]) + '/actual/wine_top_3_prc.pickle', 'rb') as handle:
actual_results = pickle.load(handle)
print("==========================================")
print("Top 1 ACTUAL model: " + actual_results[0])
print("Top 2 ACTUAL model: " + actual_results[1])
print("Top 3 ACTUAL model: " + actual_results[2])
Remove the meta-features which are not in the meta-dataset
(i.e. the features which have not been selected in the feature selection process)
"""
metafeatures_to_be_removed = []
for metafeature in meta_features.keys():
if metafeature in metadataset_feature_selected.columns:
pass
else:
metafeatures_to_be_removed.append(metafeature)
[meta_features.pop(key) for key in metafeatures_to_be_removed]
#========================META-LEARNING: RANKING================================
#KNN Ranking Method
top1, top2, top3 = KNN_ranking(metadataset_feature_selected, meta_features,
nested_results_prc)
print("==========================================")
print(" AUC-PRC ")
print("==========================================")
print("Top 1 predicted model: " + top1)
print("Top 2 predicted model: " + top2)
print("Top 3 predicted model: " + top3)
#Actual results
with open(str(p.parents[5]) + '/actual/wine_top_3_prc.pickle', 'rb') as handle:
actual_results = pickle.load(handle)
print("==========================================")
print("Top 1 ACTUAL model: " + actual_results[0])
print("Top 2 ACTUAL model: " + actual_results[1])
print("Top 3 ACTUAL model: " + actual_results[2])
for metafeature in wine_meta_features.keys():
if metafeature in metadataset_feature_selected.columns:
pass
else:
metafeatures_to_be_removed.append(metafeature)
[wine_meta_features.pop(key) for key in metafeatures_to_be_removed]
times = []
for i in range(3, 203):
start_time = time.time()
temp_metadaset = metadataset_feature_selected.iloc[:i, :]
top1, top2, top3 = KNN_ranking(temp_metadaset, wine_meta_features,
nested_results_roc)
times.append((time.time() - start_time))
training_datasets = list(range(3, 203))
plt.figure(dpi=1200)
plt.plot(training_datasets,
times,
color='blue',
label='Cylindrical bands dataset')
plt.legend(loc="upper left", fontsize=12)
plt.xlabel('Number of training datasets', fontsize=14)
plt.ylabel('Execution time of meta-learning [s]', fontsize=14)
plt.xticks(fontsize=12)
plt.yticks(fontsize=12)
plt.show()