overall[tag] += alpha[metapathnum]*(prob) if rank_num >= 10: break metapathnum += 1 sorted_x = sorted(overall.items(), key=lambda x:x[1], reverse= True) sorted_x = sorted_x[:20] sum_ = 0.0 for item in sorted_x: sum_ += item[1] matchedTags = [] for item in sorted_x: matchedTags.append((item[0],item[1]/sum_)) predictedTags = [item[0] for item in matchedTags] prec = calcPrecision_2(expectedTags, predictedTags) rec = calcRecall_2(expectedTags, predictedTags) cumPre[0] += prec[0] cumPre[1] += prec[1] cumPre[2] += prec[2] cumRec[0] += rec[0] cumRec[1] += rec[1] cumRec[2] += rec[2] out = str(alpha[0]) + ',' + str(alpha[1]) + ',' + str(alpha[2]) + ',' out += str(cumPre[0]/totcnt) + ',' + str(cumPre[1]/totcnt) + ',' + str(cumPre[2]/totcnt) + ',' + str(cumRec[0]/totcnt) + ',' + str(cumRec[1]/totcnt) + ',' + str(cumRec[2]/totcnt) + ',' out += '\n' wfile.write(out) wfile.flush() wfile.close()
# predictedTags = [item[0] for item in post['predicted_by_Doc2Vec_Matching']] # predictedTags = [item[0] for item in post['predicted_by_StringAndFIC_Matching']] # predictedTags = [item[0] for item in post['predicted_by_StringAndWVTool_Matching']] # predictedTags = [item[0] for item in post['final_Recommended_Tags']] # predictedTags = [item[0] for item in post['spread_final_Recommended_Tags']] #for baseline checking # predictedTags = [item[0] for item in post['predicted']] # predictedTags = [item[0] for item in post['predicted_By_EnTagRec']] if qcount%1000 == 0: print qcount sys.stdout.flush() if len(predictedTags) <= 0: continue post['precision'] = calcPrecision_2(expectedTags, predictedTags) post['recall'] = calcRecall_2(expectedTags, predictedTags) cumPre[0] += post['precision'][0] cumPre[1] += post['precision'][1] cumPre[2] += post['precision'][2] cumRec[0] += post['recall'][0] cumRec[1] += post['recall'][1] cumRec[2] += post['recall'][2] # wfile.write(str(post) + '\n') # wfile.flush() # wfile.close() print 'Num Docs', qcount print 'Precision: ', print cumPre[0]/qcount,
expectedTags = FIC_tags[i]['expected']
tags = {}
for tag,tag_prob in LDA_tags[i]['predicted']:
if tag not in tags:
tags[tag] = 0.0
tags[tag] += alpha*tag_prob
for tag,tag_prob in FIC_tags[i]['predicted']:
if tag not in tags:
tags[tag] = 0.0
tags[tag] += beta*tag_prob
sorted_x = sorted(tags.items(), key=lambda x:x[1], reverse= True)
predicted = [i[0] for i in sorted_x]
recall += calcRecall_2(expectedTags, predicted)[-1]
if recall > recall_max:
recall_max = recall
alpha_max = alpha
beta_max = beta
print alpha_max, beta_max, recall_max/((end-start))
# wfile = open('../data/EnTagRecWeights_'+str(part)+'_.txt','w')
# wfile.write(str(alpha_max) + ' ' + str(beta_max) + ' ' + str(recall_max/(num-(end-start))) + '\n' )
# wfile.flush()
# wfile.close()
temp = {}
for tag,prob in predicted_by_String_Matching:
temp[tag] = alpha*prob
# for tag,prob in predicted_By_FIC:
for tag,prob in predicted_by_Doc2Vec_Matching:
if tag not in temp.keys():
temp[tag] = (1.0-alpha)*prob
else:
temp[tag] += (1.0-alpha)*prob
sorted_tags = sorted(temp.items(), key=operator.itemgetter(1), reverse=True)
sorted_tags = sorted_tags[:topK]
predicted = [it[0] for it in sorted_tags]
expected = post['expected']
precision = calcPrecision_2(expected, predicted)
recall = calcRecall_2(expected, predicted)
cum_pre[0] += precision[0]
cum_pre[1] += precision[1]
cum_pre[2] += precision[2]
cum_rec[0] += recall[0]
cum_rec[1] += recall[1]
cum_rec[2] += recall[2]
cnt += 1.0
# matchedTags = []
# sum_ = 0.0
# for sorted_tag in sorted_tags:
# sum_ += sorted_tag[1]
# for sorted_tag in sorted_tags:
# matchedTags.append((sorted_tag[0],sorted_tag[1]/sum_))