import sys
sys.exit()
f_data = options.filename
eval = False
if 'test' in f_data:
eval = True
wesave = options.save
#--------------------------------------------------------#
# Prepare data
#--------------------------------------------------------#
data = pd.read_csv(f_data)
data, keeps = prep(data, eval)
import sys
sys.exit()
#--------------------------------------------------------#
# Choose a method
#--------------------------------------------------------#
# K-nn training and testing
if opt == 0:
from KNN import KNN_test_train
KNN_test_train(data[ keeps[:-1] ], data[ keeps[-1] ], wesave)
# BDT
if opt == 1:
from BDT import BDT_test_train
BDT_test_train(data[ keeps[:-1] ], data[ keeps[-1] ])
sys.exit()
# --------------------------------------------------------#
# Prepare data differently. Load in chuncks since it
# takes so much memory to process
# --------------------------------------------------------#
f_data = "data/test.csv"
data = pd.read_csv(f_data, chunksize=50000)
# Loop over each chunk and run options
probs = None
counter = 0
for chunk in data:
print "Working on ", counter, "..."
chunk, keeps = prep(chunk, True)
# Run KNN and get probs
if opt == 0:
if probs == None:
probs = KNN_evaluate(chunk[keeps])
else:
probs = np.concatenate([probs, KNN_evaluate(chunk[keeps])])
counter += 1
# Now we are done, write the output
outcat = np.array(np.arange(len(categories)), dtype="string")
for key in categories:
outcat[categories[key]] = key
