def train():
posi_result = {}
train_feature, test_feature, train_id_list, test_id_list, train_tar_list = merge_feature(feature_str)
tmp1 = [m < 32 for m in trainTarList]
tmp1 = np.array(tmp1)
# train_feature = train_feature[tmp1]
target_list = np.array(trainTarList)
target_list = target_list[tmp1]
# train_id_list = np.array(train_id_list)
# train_id_list = train_id_list[tmp1]
c_feature = trainFeature.columns[:]
clf1 = RandomForestClassifier(n_estimators=200, min_samples_split=17)
clf1.fit(trainFeature[c_feature], target_list)
# rf_preds = clf1.predict(test_feature)
rf_prob = clf1.predict_proba(test_feature)
gbdt1 = GradientBoostingClassifier(n_estimators=150, min_samples_split=17)
gbdt1.fit(trainFeature[c_feature], target_list)
# gbdt_preds = gbdt1.predict(test_feature)
gbdt_prob = gbdt1.predict_proba(test_feature)
all_prob = rf_prob + gbdt_prob
all_preds = []
print all_prob.shape
for k in range(all_prob.shape[0]):
prob1 = list(allProb[k, :])
ind1 = prob.index(max(prob1))
allPreds.append(ind1)
for j in range(len(all_preds)):
all_pre_name = dl.get_num_position(all_preds[j])
posi_result[test_id_list[j]] = all_pre_name
return posi_result
def train():
posi_result = {}
train_feature, test_feature, train_id_list, test_id_list, train_tar_list = merge_feature(
feature_str)
tmp1 = [m < 32 for m in trainTarList]
tmp1 = np.array(tmp1)
# train_feature = train_feature[tmp1]
target_list = np.array(trainTarList)
target_list = target_list[tmp1]
# train_id_list = np.array(train_id_list)
# train_id_list = train_id_list[tmp1]
c_feature = trainFeature.columns[:]
clf1 = RandomForestClassifier(n_estimators=200, min_samples_split=17)
clf1.fit(trainFeature[c_feature], target_list)
# rf_preds = clf1.predict(test_feature)
rf_prob = clf1.predict_proba(test_feature)
gbdt1 = GradientBoostingClassifier(n_estimators=150, min_samples_split=17)
gbdt1.fit(trainFeature[c_feature], target_list)
# gbdt_preds = gbdt1.predict(test_feature)
gbdt_prob = gbdt1.predict_proba(test_feature)
all_prob = rf_prob + gbdt_prob
all_preds = []
print all_prob.shape
for k in range(all_prob.shape[0]):
prob1 = list(allProb[k, :])
ind1 = prob.index(max(prob1))
allPreds.append(ind1)
for j in range(len(all_preds)):
all_pre_name = dl.get_num_position(all_preds[j])
posi_result[test_id_list[j]] = all_pre_name
return posi_result
trainIdList = trainIdList[tmp]
cFeature = trainFeature.columns[:]
clf = RandomForestClassifier(n_estimators=200, min_samples_split=17)
clf.fit(trainFeature[cFeature], target)
rfPreds = clf.predict(testFeature)
rfProb = clf.predict_proba(testFeature)
gbdt = GradientBoostingClassifier(n_estimators=100, min_samples_split=17)
gbdt.fit(trainFeature[cFeature], target)
gbdtPreds = gbdt.predict(testFeature)
gbdtProb = gbdt.predict_proba(testFeature)
allProb = rfProb + gbdtProb
allPreds = []
print allProb.shape
for tt in range(allProb.shape[0]):
prob = list(allProb[tt, :])
ind = prob.index(max(prob))
allPreds.append(ind)
rfRight, gbRight, allRight = 0, 0, 0
for i in range(len(rfPreds)):
rfPreName = dl.get_num_position(rfPreds[i])
gbdtPreName = dl.get_num_position(rfPreds[i])
allPreName = dl.get_num_position(allPreds[i])
# if rfPreName == real:rfRight+=1.0
# if rfPreName == real:gbRight+=1.0
# if allPreName == real:allRight+=1.0
# outFile1.write(testIdList[i]+'\t'+rfPreName+'\t'+real+'\n')
# outFile2.write(testIdList[i]+'\t'+gbdtPreName+'\t'+real+'\n')
# outFile3.write(testIdList[i]+'\t'+allPreName+'\t'+real+'\n')
print 'rf:' + str(rfRight / 20000) + '\n gbdt:' + str(gbRight / 20000) + '\n all:' + str(allRight / 20000)
outFile1.write(str(rfRight / 20000))
idList = np.array(idList)
tt = []
for i in range(20000):
tt.append(i % 15)
tFeature['is_train'] = tt
rightAll = 0
for i in range(15):
print i
train, test = tFeature[tFeature['is_train'] != i], tFeature[
tFeature['is_train'] == i]
tmp1 = np.array([t != i for t in tFeature['is_train']])
tmp2 = np.array([t == i for t in tFeature['is_train']])
trainTar, testTar = target[tmp1], target[tmp2]
testId = idList[tmp2]
clf = RandomForestClassifier(
n_estimators=200,
min_samples_split=13) # ,max_depth=35,max_features=0.4)
features = tFeature.columns[:-1]
clf.fit(train[features], trainTar)
preds = clf.predict(test[features])
right = 0
for n in range(len(preds)):
if preds[n] == testTar[n]:
right += 1.0
rightAll += 1.0
outFiles.write(testId[n] + '\t' + dl.get_num_position(preds[n]) +
'\t' + dl.get_num_position(testTar[n]) + '\n')
print right / len(preds)
outFiles.close()
print rightAll / 20000
tFeature = get_feature(Tlines1, Tlines2, TrfProb, TgbdtProb)
eFeature = get_feature(Tlines1, Tlines2, TrfProb, TgbdtProb)
target = np.array(tarList)
idList = np.array(idList)
tt = []
for i in range(20000):
tt.append(i % 15)
tFeature['is_train'] = tt
rightAll = 0
for i in range(15):
print i
train, test = tFeature[tFeature['is_train'] != i], tFeature[tFeature['is_train'] == i]
tmp1 = np.array([t != i for t in tFeature['is_train']])
tmp2 = np.array([t == i for t in tFeature['is_train']])
trainTar, testTar = target[tmp1], target[tmp2]
testId = idList[tmp2]
clf = RandomForestClassifier(n_estimators=200, min_samples_split=13) # ,max_depth=35,max_features=0.4)
features = tFeature.columns[:-1]
clf.fit(train[features], trainTar)
preds = clf.predict(test[features])
right = 0
for n in range(len(preds)):
if preds[n] == testTar[n]:
right += 1.0
rightAll += 1.0
outFiles.write(
testId[n] + '\t' + dl.get_num_position(preds[n]) + '\t' + dl.get_num_position(testTar[n]) + '\n')
print right / len(preds)
outFiles.close()
print rightAll / 20000
cFeature = trainFeature.columns[:]
clf = RandomForestClassifier(n_estimators=200, min_samples_split=17)
clf.fit(trainFeature[cFeature], target)
rfPreds = clf.predict(testFeature)
rfProb = clf.predict_proba(testFeature)
gbdt = GradientBoostingClassifier(n_estimators=100, min_samples_split=17)
gbdt.fit(trainFeature[cFeature], target)
gbdtPreds = gbdt.predict(testFeature)
gbdtProb = gbdt.predict_proba(testFeature)
allProb = rfProb + gbdtProb
allPreds = []
print allProb.shape
for tt in range(allProb.shape[0]):
prob = list(allProb[tt, :])
ind = prob.index(max(prob))
allPreds.append(ind)
rfRight, gbRight, allRight = 0, 0, 0
for i in range(len(rfPreds)):
rfPreName = dl.get_num_position(rfPreds[i])
gbdtPreName = dl.get_num_position(rfPreds[i])
allPreName = dl.get_num_position(allPreds[i])
# if rfPreName == real:rfRight+=1.0
# if rfPreName == real:gbRight+=1.0
# if allPreName == real:allRight+=1.0
# outFile1.write(testIdList[i]+'\t'+rfPreName+'\t'+real+'\n')
# outFile2.write(testIdList[i]+'\t'+gbdtPreName+'\t'+real+'\n')
# outFile3.write(testIdList[i]+'\t'+allPreName+'\t'+real+'\n')
print 'rf:' + str(rfRight / 20000) + '\n gbdt:' + str(
gbRight / 20000) + '\n all:' + str(allRight / 20000)
outFile1.write(str(rfRight / 20000))
for i in range(len(trainFeatureR)):
tt.append(i % 5)
i = 4
tmp1 = np.array([t != i for t in tt])
tmp2 = np.array([t == i for t in tt])
trainFeature, testFeature = trainFeatureR[tmp1], trainFeatureR[tmp2]
trainTar, testTar = targetR[tmp1], targetR[tmp2]
trainId, testId = trainIdListR[tmp1], trainIdListR[tmp2]
clf = RandomForestClassifier(n_estimators=200, min_samples_split=17)
clf.fit(trainFeature[Cfeature], trainTar)
preds = clf.predict(testFeature)
predPro = clf.predict_proba(testFeature)
rfPro = predPro
right = 0
for n in range(len(preds)):
preName = dl.get_num_position(preds[n])
real = dl.get_num_position(testTar[n])
if preName == real:
right += 1.0
outFile1.write(str(testId[n]) + '\t' + preName + '\t' + real + '\n')
print right / (len(trainFeatureR) / 5.0)
pickle.dump(rfPro, outPkl1)
i = 4
print i
tmp1 = np.array([t != i for t in tt])
tmp2 = np.array([t == i for t in tt])
trainFeature, testFeature = trainFeatureR[tmp1], trainFeatureR[tmp2]
trainTar, testTar = targetR[tmp1], targetR[tmp2]
trainId, testId = trainIdListR[tmp1], trainIdListR[tmp2]
clf = GradientBoostingClassifier(n_estimators=6, min_samples_split=17)
for i in range(len(trainFeatureR)):
tt.append(i % 5)
i = 4
tmp1 = np.array([t != i for t in tt])
tmp2 = np.array([t == i for t in tt])
trainFeature, testFeature = trainFeatureR[tmp1], trainFeatureR[tmp2]
trainTar, testTar = targetR[tmp1], targetR[tmp2]
trainId, testId = trainIdListR[tmp1], trainIdListR[tmp2]
clf = RandomForestClassifier(n_estimators=200, min_samples_split=17)
clf.fit(trainFeature[Cfeature], trainTar)
preds = clf.predict(testFeature)
predPro = clf.predict_proba(testFeature)
rfPro = predPro
right = 0
for n in range(len(preds)):
preName = dl.get_num_position(preds[n])
real = dl.get_num_position(testTar[n])
if preName == real:
right += 1.0
outFile1.write(str(testId[n]) + '\t' + preName + '\t' + real + '\n')
print right / (len(trainFeatureR) / 5.0)
pickle.dump(rfPro, outPkl1)
i = 4
print i
tmp1 = np.array([t != i for t in tt])
tmp2 = np.array([t == i for t in tt])
trainFeature, testFeature = trainFeatureR[tmp1], trainFeatureR[tmp2]
trainTar, testTar = targetR[tmp1], targetR[tmp2]
trainId, testId = trainIdListR[tmp1], trainIdListR[tmp2]
clf = GradientBoostingClassifier(n_estimators=6, min_samples_split=17)
