try:
aux_file = sys.argv[1]
except:
print "Argument expected: weka-csv (inst#,actual,predicted,error,prediction) (with header) file with labels"
sys.exit(-1)
try:
label_no = int(sys.argv[2])
except:
print "Argument expected: label to be overwritten no."
sys.exit(-1)
print "Loading labels' (weka-csv) file:", aux_file
#inst#,actual,predicted,error,prediction
lines = open(aux_file).readlines()
labels2bool = list( line.split(",")[2].find("present")>=0 for line in lines[1:] if len(line.strip())>0 )
print "",len(labels2bool),"rows loaded"
#print labels2bool
labels = []
for ocur in labels2bool:
if ocur:
labels.append([label_no])
else:
labels.append([])
aux_file = aux_file + "_jrs.txt"
print "Writing to file:", aux_file
jrs_io.store_labels(open(aux_file,"w"), labels)
print "Argument expected: path to a labels' file."
sys.exit(-1)
try:
labels2_path = sys.argv[2]
except:
print "Argument expected: path to a second labels' file."
sys.exit(-1)
try:
out_path = sys.argv[3]
except:
print "Argument expected: output labels' file."
sys.exit(-1)
print "Loading labels' 1 file:", labels1_path
labels1 = jrs_io.load_labels(open(labels1_path))
try:
print "Loading labels' 2 file:", labels2_path
labels2 = jrs_io.load_labels(open(labels2_path))
except:
print "Failed loading file", labels2_path
print "Using empty file"
labels2 = [[] for i in xrange(len(labels1))]
labels12 = []
for i in xrange(len(labels1)):
labels12.append(sorted(set(labels1[i] + labels2[i])))
print "Writing to 1+2 file:", out_path
jrs_io.store_labels(open(out_path, "w"), labels12)
sys.exit(-1)
print "The program shuffles distances and trainingLabels."
print "Loading labels' file:", labels_path
labels = jrs_io.load_labels(open(labels_path))
n = len(labels)
print "",n," labels' sets loaded."
order = range(n)
random.shuffle(order)
print "Random order:", order[:30],"..."
print "Shuffling labels..."
labels_shuffled = [labels[ix] for ix in order]
jrs_io.store_labels(open(labels_path+"_shuffled","w"), labels_shuffled)
print "Loading distances' file:", distance_matrix_path
distances = jrs_io.load_data(open(distance_matrix_path), lambda x: x)
try: print "",len(distances), "x",len(distances[0])
except: pass
print "Extending order..."
order = order + range(n, len(distances))
print "Extended order:", order
print "Shuffling columns"
distances_tmp = []
for row in distances:
new_row = [ row[ix] for ix in order ]
distances_tmp.append(new_row)
sys.exit(-1)
print "The program shuffles distances and trainingLabels."
print "Loading labels' file:", labels_path
labels = jrs_io.load_labels(open(labels_path))
n = len(labels)
print "", n, " labels' sets loaded."
order = range(n)
random.shuffle(order)
print "Random order:", order[:30], "..."
print "Shuffling labels..."
labels_shuffled = [labels[ix] for ix in order]
jrs_io.store_labels(open(labels_path + "_shuffled", "w"), labels_shuffled)
print "Loading distances' file:", distance_matrix_path
distances = jrs_io.load_data(open(distance_matrix_path), lambda x: x)
try:
print "", len(distances), "x", len(distances[0])
except:
pass
print "Extending order..."
order = order + range(n, len(distances))
print "Extended order:", order
print "Shuffling columns"
distances_tmp = []
for row in distances:
print "Argument expected: path to a second labels' file."
sys.exit(-1)
try:
out_path = sys.argv[3]
except:
print "Argument expected: output labels' file."
sys.exit(-1)
print "Loading labels' 1 file:", labels1_path
labels1 = jrs_io.load_labels(open(labels1_path))
try:
print "Loading labels' 2 file:", labels2_path
labels2 = jrs_io.load_labels(open(labels2_path))
except:
print "Failed loading file", labels2_path
print "Using empty file"
labels2 = [[] for i in xrange(len(labels1))]
labels12 = []
for i in xrange(len(labels1)):
labels12.append( sorted(set(labels1[i]+labels2[i])) )
print "Writing to 1+2 file:", out_path
jrs_io.store_labels(open(out_path,"w"), labels12)
#subs = max(votes[0] - diff,0.0)
#print "","subs=",subs
#votes_for_yes = [v-subs for v in votes_for_yes]
#votes_for_no = [v-subs for v in votes_for_no]
#print "","sample_no:",sample_no," label:",label,"votes_for_yes:",votes_for_yes,"votes_for_no:",votes_for_no
try:
#yes = float(sum(votes_for_yes))/len(votes_for_yes)
yes = sum(votes_for_yes)
except:
yes = 0.0
try:
#no = float(sum(votes_for_no))/len(votes_for_no)
no = sum(votes_for_no)
except:
no = 0.0
if yes > no:
selected_sample_labels.append(label)
predicted_labels.append(sorted(selected_sample_labels))
print "", "all labels:", all_sample_labels, " -> sel:", sorted(
selected_sample_labels), "\n"
print "STORING TO FILE:", out_path
import sys
sys.path.append(r'../')
sys.path.append(r'../../')
import jrs_io
jrs_io.store_labels(open(out_path, "w"), predicted_labels)
if FINALTEST:
start = time.clock()
print "Final predicting..."
print " loading from file:", distance_matrix_path, " in range", FINALTEST_START, "-", FINALTEST_END
predicted_labels = []
for i, line in enumerate(open(distance_matrix_path).xreadlines()):
if i % 1000 == 0: print "", i, "rows processed..."
if i >= FINALTEST_END: break
if i < FINALTEST_START: continue
row = [cast_method(x) for x in line.split()]
final2training_distances = row[training_range[0]:training_range[1]]
predicted_labels.append(
multilabel_classifier(final2training_distances,
training_labels))
jrs_io.store_labels(open(FINAL_RESULT_PATH, "w"), predicted_labels)
lcount = [len(ll) for ll in predicted_labels]
print " avg labels in predicted:", float(sum(lcount)) / (len(lcount))
print " done in", (time.clock() - start), "sec..."
print "------------------------------------------"
else:
def eval():
start = time.clock()
print "Calculating predictions of ", len(
testing_labels), " labels' sets..."
predicted_labels = jrs_multilabel_classifier.classify_multilabel(
testing2training_distances, training_labels,
multilabel_classifier)
accuracy, precision, recall, hammingloss, subset01loss, fmeasure = jrs_evaluation.jrs_evaluate(
testing_labels, predicted_labels)
print "------------------------------------------"
print "Classifying file:", features_matrix_path
f = open(features_matrix_path)
predicted_labels = []
for i,line in enumerate(f.xreadlines()):
if i%1000==0: print "",i,"..."
row = [int(x) for x in line.split()]
ll = []
for label,feature_ixs in label2feature_ixs.iteritems():
says_yes = sum(row[ix]>0 for ix in feature_ixs)
if says_yes >= len(feature_ixs)*MIN_FRACTION_OF_VOTES:
ll.append(label)
ll = sorted(ll)
predicted_labels.append(ll)
print "",i," oracle",labels[i]," pred",ll
print "","len=",len(labels[:(i+1)]), len(predicted_labels)
accuracy, precision, recall, hammingloss, subset01loss, fmeasure = jrs_evaluation.jrs_evaluate(labels[:(i+1)], predicted_labels)
print "\t\t\t\t\t","%.2f" %precision,"%.2f" %recall,"%.2f" %fmeasure
print "------------------------------------------"
accuracy, precision, recall, hammingloss, subset01loss, fmeasure = jrs_evaluation.jrs_evaluate(labels, predicted_labels)
print "\t\t\t\t\t","%.2f" %precision,"%.2f" %recall,"%.2f" %fmeasure
print "Wrining results to", out_path
jrs_io.store_labels(open(out_path,"w"), predicted_labels)
#############################################################################################################################
#############################################################################################################################
if FINALTEST:
start = time.clock()
print "Final predicting..."
print " loading from file:",distance_matrix_path," in range",FINALTEST_START,"-",FINALTEST_END
predicted_labels = []
for i,line in enumerate(open(distance_matrix_path).xreadlines()):
if i%1000 == 0: print "",i,"rows processed..."
if i>=FINALTEST_END: break
if i<FINALTEST_START: continue
row = [cast_method(x) for x in line.split()]
final2training_distances = row[training_range[0]:training_range[1]]
predicted_labels.append(multilabel_classifier(final2training_distances, training_labels))
jrs_io.store_labels(open(FINAL_RESULT_PATH,"w"), predicted_labels)
lcount = [len(ll) for ll in predicted_labels]
print " avg labels in predicted:", float(sum(lcount))/(len(lcount))
print " done in", (time.clock() - start), "sec..."
print "------------------------------------------"
else:
def eval():
start = time.clock()
print "Calculating predictions of ",len(testing_labels)," labels' sets..."
predicted_labels = jrs_multilabel_classifier.classify_multilabel(testing2training_distances, training_labels, multilabel_classifier)
try:
print testing_labels[:10],"\n", predicted_labels[:10]
accuracy, precision, recall, hammingloss, subset01loss, fmeasure = jrs_evaluation.jrs_evaluate(testing_labels, predicted_labels)
except:
print "[knn] Error in jrs_evaluation.jrs_evaluate(testing_labels, predicted_labels):",testing_labels, predicted_labels
print " accuracy:", accuracy,"\n precision:", precision,"\n recall:", recall,"\n fmeasure:", fmeasure
try:
aux_file = sys.argv[1]
except:
print "Argument expected: weka-csv (inst#,actual,predicted,error,prediction) (with header) file with labels"
sys.exit(-1)
try:
label_no = int(sys.argv[2])
except:
print "Argument expected: label to be overwritten no."
sys.exit(-1)
print "Loading labels' (weka-csv) file:", aux_file
#inst#,actual,predicted,error,prediction
lines = open(aux_file).readlines()
labels2bool = list(
line.split(",")[2].find("present") >= 0 for line in lines[1:]
if len(line.strip()) > 0)
print "", len(labels2bool), "rows loaded"
#print labels2bool
labels = []
for ocur in labels2bool:
if ocur:
labels.append([label_no])
else:
labels.append([])
aux_file = aux_file + "_jrs.txt"
print "Writing to file:", aux_file
jrs_io.store_labels(open(aux_file, "w"), labels)
