plt.rc('font', family='sans-serif')
rcParams['figure.figsize'] = 15, 2
all_metrics = []
bdf = pd.DataFrame.from_csv('baseline.csv', index_col=False, header=None)
results_dir = 'final_results_wl_2_to_20'
idx = 0
a = 10
print a
for folder in get_folders('NewlyAddedDatasets/'):
metrics = {}
best = 0
# for a in range(a, a + 1):
for a in range(3, 21):
try:
clsssifiedf = join(folder, results_dir, 'alphabet_%s' % a, 'classified.csv')
df = pd.DataFrame.from_csv(clsssifiedf, index_col=False)
y_true, y_pred = df['label'].values.tolist(), df['predicted'].values.tolist()
accuracy = accuracy_score(y_true, y_pred)
if accuracy > best:
best = accuracy
except Exception as e:
print e
def build_all_corpora(path):
for folder in get_folders(path):
build_corpus_for_folder(folder)
return
from build_corpus import build_all_corpora
from dsco_classification import process_results
from os.path import join
import pandas as pd
if __name__ == '__main__':
print 'Converting .mat to .csv and Symbolizing real-valued data to strings'
saxify_all('NewlyAddedDatasets')
print 'Done converting datasets'
print '-' * 80
print 'Extracting unigrams and bigrams'
build_all_corpora('NewlyAddedDatasets')
print 'Done extracting unigrams and bigrams'
print '-' * 80
for folder in get_folders('NewlyAddedDatasets/'):
print 'Processing %s' % folder
results = process_results(folder, resultsdir='results')
df = pd.DataFrame(results, columns=['WL', 'A', 'Acc'])
retf = join(folder, 'results.csv')
df.to_csv(retf, index=False)
print '-' * 80
print "DSCo-NG's suggestion:"
i = df['Acc'].argmax()
print 'w: %d\ta: %d\taccuracy: %f' % (df.iloc[i, 0], df.iloc[i, 1],
df.iloc[i, 2])
print 'Results are saved to %s' % retf
print '-' * 80
def build_all_corpora(path):
for folder in get_folders(path):
build_corpus_for_folder(folder)
return
