def birch(X, y, n):
"""
Birchによるクラスタリング
Parameters
----------
X : numpy array
データ
y : numpy array
正解ラベル
n : int
クラスタ数
Returns
-------
acc_br : float
正解率
time_br : float
実行時間
"""
br = Birch(n_clusters=2)
start_br = time.time()
y_br = br.fit_predict(X)
end_br = time.time()
y_br = np.reshape(y_br, (1, len(y[0])))
acc_br, _, _ = acc(y_br, y)
time_br = round(end_br - start_br, 2)
make_graph(X, y_br, n, "Birch")
return acc_br, time_br
def dbscan(X, y, eps, n):
"""
DBSCANを用いたクラスタリング
Parameters
----------
X : numpy array
データ
y : numpy array
正解ラベル
eps : float
二つのサンプルの最大距離
n : int
クラスタ数
Returns
-------
acc_dbs : float
正解率
time_dbs : float
実行時間
"""
dbs = DBSCAN(eps=eps, min_samples=2)
start_dbs = time.time()
y_dbs = dbs.fit_predict(X)
end_dbs = time.time()
y_dbs = np.reshape(y_dbs, (1, len(y[0])))
acc_dbs, _, _ = acc(y_dbs, y)
time_dbs = round(end_dbs - start_dbs, 2)
make_graph(X, y_dbs, n, "DBSCAN")
return acc_dbs, time_dbs
def agglomerativeclustering(X, y, n):
"""
Agglomerative Clusteringを用いたクラスタリング
Parameters
----------
X : numpy array
データ
y : numpy array
正解ラベル
n : int
クラスタ数
Returns
-------
acc_ac : float
正解率
time_ac : float
実行時間
-------
"""
ac = AgglomerativeClustering(n_clusters=n)
start_ac = time.time()
y_ac = ac.fit_predict(X)
end_ac = time.time()
y_ac = np.reshape(y_ac, (1, len(y[0])))
acc_ac, _, _ = acc(y_ac, y)
time_ac = round(end_ac - start_ac, 2)
make_graph(X, y_ac, n, "AgglomerativeClustering")
return acc_ac, time_ac
def kmeans(X, y, n):
"""
K Meansによるクラスタリング
Parameters
----------
X : numpy array
データ
y : numpy array
正解ラベル
n : int
クラスタ数
Returns
-------
acc_km : float
正解率
time_km : float
実行時間
"""
km = KMeans(n_clusters=n,
init="random",
n_init=10,
max_iter=300,
random_state=0)
start_km = time.time()
y_km = km.fit_predict(X)
end_km = time.time()
y_km = np.reshape(y_km, (1, len(y[0])))
acc_km, _, _ = acc(y_km, y)
time_km = round(end_km - start_km, 2)
make_graph(X, y_km, n, "KMeans")
return acc_km, time_km
def spectralclustering(X, y, n):
"""
Spectral Clusteringによるクラスタリング
Parameters
----------
X : numpy array
データ
y : numpy array
正解ラベル
n : int
クラスタ数
Returns
-------
acc_sc : float
正解率
time_sc : float
実行時間
"""
sc = SpectralClustering(n_clusters=2,
affinity="nearest_neighbors")
start_sc = time.time()
y_sc = sc.fit_predict(X)
end_sc = time.time()
y_sc = np.reshape(y_sc, (1, len(y[0])))
acc_sc, _, _ = acc(y_sc, y)
time_sc = round(end_sc - start_sc, 2)
make_graph(X, y_sc, n, "SpectralClustering")
return acc_sc, time_sc
def predict_folder(image_folder, ref_folder, output_prefix,
net_config_location, net_weights_location, alpha_channel,
use_gpu, win_size, crop_size):
images = listdir(image_folder)
if (ref_folder != ''):
pred_concat = np.array([])
ref_concat = np.array([])
refs = listdir(ref_folder)
#start counting time
start_time = time.time()
for i in range(0, len(images)):
image_location = path.join(image_folder, images[i])
pred = classify(net_config_location, net_weights_location,
image_location, alpha_channel, use_gpu, win_size,
crop_size)
out = Image.fromarray(pred * 40)
out.save(output_prefix + images[i])
if (ref_folder != ''):
ref_location = path.join(ref_folder, refs[i])
ref = np.array(Image.open(ref_location)) / 40
pred = pred.flatten()
ref = ref.flatten()
pred_concat = np.concatenate((pred_concat, pred))
ref_concat = np.concatenate((ref_concat, ref))
#print elapsed time
print("--- %s seconds ---" % (time.time() - start_time))
if (ref_folder != ''):
accuracies = acc(pred_concat, ref_concat)
return accuracies
else:
return 0
def predict_folder(image_folder,ref_folder,output_prefix,
net_config_location,net_weights_location,
alpha_channel,use_gpu, win_size, crop_size):
images = listdir(image_folder)
if (ref_folder!=''):
pred_concat = np.array([])
ref_concat = np.array([])
refs = listdir(ref_folder)
#start counting time
start_time = time.time()
for i in range(0,len(images)):
image_location = path.join(image_folder,images[i])
pred = classify(net_config_location, net_weights_location, image_location, alpha_channel, use_gpu, win_size, crop_size)
out = Image.fromarray(pred*40)
out.save(output_prefix+images[i])
if (ref_folder!=''):
ref_location = path.join(ref_folder,refs[i])
ref = np.array(Image.open(ref_location))/40
pred=pred.flatten()
ref=ref.flatten()
pred_concat=np.concatenate((pred_concat,pred))
ref_concat=np.concatenate((ref_concat,ref))
#print elapsed time
print("--- %s seconds ---" % (time.time() - start_time))
if (ref_folder!=''):
accuracies = acc(pred_concat,ref_concat)
return accuracies
else:
return 0
def main(args):
names=[]
with open(args.inputfile) as af:
seq_list=read_fasta(af)
for e in seq_list:
names.append(e.name)
#kmer.py -f tab -l +1 -r 1 -k ',num2str(k),' TAIR10_DHSs.fas TAIR10_DHSs_reckmer_',num2str(k),'.txt DNA'
res_kmer = make_kmer_vector(k=2, alphabet=index_list.DNA, filename=args.inputfile, revcomp=True)
#acc.py -e user_indices.txt -f svm -l +1 -lag ',num2str(lag),' TAIR10_DHSs.fas TAIR10_DHSs_dac_',num2str(lag),'.txt DNA DAC'
if args.s==0:
model_file='pDHSdata_TAIR_model.txt'
lag=3
else:
model_file='pDHSdata_TIGR_model.txt'
lag=8
with open(args.inputfile) as f:
k = read_k('DNA', 'DAC', 0)
ind_list=[]
res_acc = acc(f, k, lag, ind_list, index_list.DNA, extra_index_file='user_indices.txt', all_prop=False, theta_type=1)
# features= revckmer+dac,formed by add each row
res=[]
for i in range(len(res_kmer)):
res.append(res_kmer[i]+res_acc[i])
featuresfile=args.inputfile+'_tmp_features.txt'
# Write correspond res file.
from util import write_libsvm
write_libsvm(res, ['+1'] * len(res), featuresfile)
#predict the result
tmp_predict_result_file=args.inputfile+'_tmp_result.txt'
if sys.platform == 'win32':
options='svm-predict -b 1 -q '+featuresfile+' '+model_file+' '+ tmp_predict_result_file
else:
options='./svm-predict -b 1 -q '+featuresfile+' '+model_file+' '+ tmp_predict_result_file
os.system(options)
pf=open(args.outputfile,'w')
with open(tmp_predict_result_file) as nf:
label, TrueProb, FalseProb= '', '',''
count = 0
while True:
line = nf.readline().strip()
if not line:
break
if count>len(names):
break
if 0==count:
pf.write('ID\t\tLabel\t\tProb\n')
count+=1
continue
label=int(line.split()[0])
TrueProb=line.split()[1]
FalseProb=line.split()[2]
if label==-1:
pf.write(names[count-1]+'\t\t'+'Non DHS'+'\t\t'+str(FalseProb)+'\n')
else:
pf.write(names[count-1]+'\t\t'+'DHS'+'\t\t'+str(TrueProb)+'\n')
count+=1
pf.close()
cwd = os.getcwd()
files = [x for x in os.listdir(os.getcwd()) if os.path.isfile(os.path.join(cwd,x))]
#print files
for file in files:
if -1 != file.find('tmp'):
os.remove(file)
