def capital_gains_lift(source):
"""
Computes capital gains lift in top income percentages over time chart
"""
dataset = gds.getDataSetUsingCSV(source)
columns = (
("Top 10% income share-including capital gains",
"Top 10% income share"),
("Top 5% income share-including capital gains", "Top 5% income share"),
("Top 1% income share-including capital gains", "Top 1% income share"),
("Top 0.5% income share-including capital gains",
"Top 0.5% income share"),
("Top 0.1% income share-including capital gains",
"Top 0.1% income share"),
("Top 0.05% income share-including capital gains",
"Top 0.05% income share"),
)
source = list(dataset)
series = [
md.delta(gds.timeseries(source, a), gds.timeseries(source, b))
for a, b in columns
]
return pld.linechart(series,
labels=list(col[1] for col in columns),
title="U.S. Capital Gains Income Lift",
ylabel="Percentage Difference")
def percent_income_share(source):
"""Create Income Share chart"""
columns = (
"Top 10% income share",
"Top 5% income share",
"Top 1% income share",
"Top 0.5% income share",
"Top 0.1% income share",
)
source = list(gds.getDataSetUsingCSV(source))
return pld.linechart([gds.timeseries(source, col) for col in columns],
labels=columns,
title="U.S. Percentage Income Share",
ylabel="Percentage")
def mean_normalized_percent_income_share(source):
dataset = gds.getDataSetUsingCSV(source)
columns = (
"Top 10% income share",
"Top 5% income share",
"Top 1% income share",
"Top 0.5% income share",
"Top 0.1% income share",
)
source = list(dataset)
return pld.linechart(
[md.normalize(gds.timeseries(source, col)) for col in columns],
labels=columns,
title="Mean Normalized U.S. Percentage Income Share",
ylabel="Percentage")
def generate_report_for_selected_countries(source):
# Select countries to include
include = ("United States", "France", "Italy", "Germany", "South Africa",
"New Zealand")
# Get dataset from CSV
data = list(gds.getDataSetUsingDictReader(source, include))
years = set(gds.extract_years(data))
# Generate context
context = {
'title':
"Average Income per Family, %i - %i" % (min(years), max(years)),
'years': json.dumps(list(years)),
'countries': [v[0] for v in data],
'series': json.dumps(list(gds.extract_series(data, years))),
}
# Write HTML with template
gr.write(context)
def average_incomes(source):
"""
Compares percentage average incomes
"""
dataset = gds.getDataSetUsingCSV(source)
columns = (
"Top 10% average income",
"Top 5% average income",
"Top 1% average income",
"Top 0.5% average income",
"Top 0.1% average income",
"Top 0.05% average income",
)
source = list(dataset)
return pld.linechart([gds.timeseries(source, col) for col in columns],
labels=columns,
title="U.S. Average Income",
ylabel="2008 US Dollars")
def income_composition(source):
"""
Compares income composition
"""
dataset = gds.getDataSetUsingCSV(source)
columns = (
"Top 10% income composition-Wages, salaries and pensions",
"Top 10% income composition-Dividends",
"Top 10% income composition-Interest Income",
"Top 10% income composition-Rents",
"Top 10% income composition-Entrepreneurial income",
)
source = list(dataset)
labels = ("Salary", "Dividends", "Interest", "Rent", "Business")
return pld.stackedarea([gds.timeseries(source, col) for col in columns],
labels=labels,
title="U.S. Top 10% Income Composition",
ylabel="Percentage")
def average_top_income_lift(source):
"""
Compares top percentage avg income over total avg
"""
dataset = gds.getDataSetUsingCSV(source)
columns = (
("Top 10% average income", "Top 0.1% average income"),
("Top 5% average income", "Top 0.1% average income"),
("Top 1% average income", "Top 0.1% average income"),
("Top 0.5% average income", "Top 0.1% average income"),
("Top 0.1% average income", "Top 0.1% average income"),
)
source = list(dataset)
series = [
md.delta(gds.timeseries(source, a), gds.timeseries(source, b))
for a, b in columns
]
return pld.linechart(series,
labels=list(col[0] for col in columns),
title="U.S. Income Disparity",
ylabel="2008 US Dollars")
def main(argv):
try:
if argv[0] == "--help":
help()
elif argv[0] == "--retrain":
model_type = argv[1]
pad = int(argv[2])
csv_path = GetDataSet.get_csv("./dataset/image/data",
"./dataset/image/label", pad)
if model_type == "DecisionTree":
GetModel.decisiontree(csv_path)
elif model_type == "RandomForest":
GetModel.randomforest(csv_path)
elif argv[0] == "--evaluate":
model_path = argv[1]
GetModel.evaluate(model_path)
elif argv[0] == "--predict":
model_path = argv[1]
image_path = argv[2]
Predict.predict(image_path, model_path)
else:
help()
except getopt.GetoptError:
help()
def train():
x1 = tf.placeholder(tf.float32, [None, Inference.input1Node],
name='FP-input')
x2 = tf.placeholder(tf.float32, [None, Inference.input2Node],
name='Des-input')
keeprate = tf.placeholder(tf.float32, name='keeprate')
label = tf.placeholder(tf.float32, [None, Inference.outputNode],
name='Label-input')
one = tf.ones_like(label, name='ones')
zero = tf.zeros_like(label, name='zeros')
t = Inference.fakerinference(x1, x2, keeprate)
y = tf.where(t < 0.5, zero, one)
global_step = tf.Variable(0, False)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=t, labels=tf.argmax(label, 1))
cross_entropy_mean = tf.reduce_mean(cross_entropy, name='MeanCrossEntropy')
loss = cross_entropy_mean + tf.add_n(tf.get_collection('losses'),
name='Regularizer')
train_step = tf.train.AdamOptimizer().minimize(loss,
global_step,
name='TrainStep')
#
correct_prediction = tf.equal(tf.argmax(y, 1),
tf.argmax(label, 1),
name='CorrectNum')
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32),
name='Acc')
auc_value, auc_op = tf.metrics.auc(label, y, name="Auc")
traindata, trainlabel, traindesc = GetDataSet.getNextBatch(True, False)
testdata, testlabel, testdesc = GetDataSet.getNextBatch(False, False)
with tf.Session() as sess:
writer = tf.summary.FileWriter("logs/", sess.graph)
tf.global_variables_initializer().run()
tf.local_variables_initializer().run()
for i in range(5000):
x1s, labels, x2s = GetDataSet.getNextBatch()
sess.run(train_step,
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 0.75
})
if i % 10 == 0:
print(i)
a, b = sess.run([accuracy, loss],
feed_dict={
x1: traindata,
x2: traindesc,
label: trainlabel,
keeprate: 1
})
trainacc.append(a)
trainloss.append(b)
a, b, c = sess.run([accuracy, loss, auc_op],
feed_dict={
x1: testdata,
x2: testdesc,
label: testlabel,
keeprate: 1
})
testacc.append(a)
testloss.append(b)
auc.append(
sess.run(auc_value,
feed_dict={
x1: testdata,
x2: testdesc,
label: testlabel,
keeprate: 1
}))
print(auc[-1])
writer.close()
keeprate: 1
})
testacc.append(a)
testloss.append(b)
auc.append(
sess.run(auc_value,
feed_dict={
x1: testdata,
x2: testdesc,
label: testlabel,
keeprate: 1
}))
print(auc[-1])
writer.close()
trainloss = []
trainacc = []
testloss = []
testacc = []
auc = []
projectDir = r'C:\Users\lenovo\Desktop\毕业论文\result\des\projects\project0-0\FP'
GetDataSet.getDataSet(projectDir)
train()
projectDir = r'C:\Users\lenovo\Desktop\毕业论文\result\des\projects'
GetDataSet.save(trainacc, projectDir + '\\trainacc.csv', 1)
GetDataSet.save(trainloss, projectDir + '\\trainloss.csv', 1)
GetDataSet.save(testacc, projectDir + '\\testacc.csv', 1)
GetDataSet.save(testloss, projectDir + '\\testloss.csv', 1)
GetDataSet.save(auc, projectDir + '\\auc.csv', 1)
def train():
x1 = tf.placeholder(tf.float32, [None, Inference.input1Node],
name='x1-input')
x2 = tf.placeholder(tf.float32, [None, Inference.input2Node],
name='x2-input')
keeprate = tf.placeholder(tf.float32)
label = tf.placeholder(tf.float32, [None, Inference.outputNode],
name='label-input')
one = tf.ones_like(label)
zero = tf.zeros_like(label)
t = Inference.inference(x1, x2, keeprate)
y = tf.where(t < 0.5, zero, one)
global_step = tf.Variable(0, False)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=t, labels=tf.argmax(label, 1))
cross_entropy_mean = tf.reduce_mean(cross_entropy)
loss = cross_entropy_mean # + tf.add_n(tf.get_collection('losses'))
learning_rate = tf.train.exponential_decay(LearningRateBase, global_step,
30, LearningRateDecay)
train_step = tf.train.AdamOptimizer().minimize(loss, global_step)
# train_step = tf.train.GradientDescentOptimizer(0.9).minimize(loss,global_step)
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(label, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
auc_value, auc_op = tf.metrics.auc(tf.argmax(label, 1), tf.argmax(y, 1))
trainfakername = [0 for i in range(len(GetDataSet.trainDataSet))]
testfakername = [0 for i in range(len(GetDataSet.testDataSet))]
with tf.Session() as sess:
tf.global_variables_initializer().run()
tf.local_variables_initializer().run()
for i in range(5000):
x1s, labels, x2s = GetDataSet.getNextBatch()
sess.run(train_step,
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 0.9
})
if i % 200 == 0:
x1s, labels, x2s = GetDataSet.getNextBatch()
a, b = sess.run([accuracy, loss],
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 1
})
print("train:%g\ntrainloss:%g" % (a, b))
x1s, labels, x2s = GetDataSet.getNextBatch(False, False)
a, b, c = sess.run([accuracy, loss, auc_op],
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 1
})
print("test:%g\ntestloss:%g" % (a, b))
print("auc:%g" % (sess.run(auc_value,
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 1
})))
#修改前0.6674
if i > 4000 and i % 10 == 0:
x1s, labels, x2s = GetDataSet.getNextBatch(False, False)
predy = sess.run(correct_prediction,
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 1
})
for index2 in range(len(predy)):
if not predy[index2]:
testfakername[index2] += 1
#print(testfakername)
#print('end')
for i in range(len(testfakername)):
if testfakername[i] > 98:
print(GetDataSet.testName[i])
import tensorflow as tf
import GetDataSet
import Inference
result = []
projectDir = r'C:\Users\lenovo\Desktop\毕业论文\result\des\projects\temp'
#temp'
#project0-0\FP'
GetDataSet.getDataSet(projectDir)
def train():
x1 = tf.placeholder(tf.float32, [None, Inference.input1Node],
name='x1-input')
x2 = tf.placeholder(tf.float32, [None, Inference.input2Node],
name='x2-input')
keeprate = tf.placeholder(tf.float32)
label = tf.placeholder(tf.float32, [None, Inference.outputNode],
name='label-input')
one = tf.ones_like(label)
zero = tf.zeros_like(label)
t = Inference.inference(x1, x2, keeprate)
y = tf.where(t < 0.5, zero, one)
global_step = tf.Variable(0, False)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=t, labels=tf.argmax(label, 1))
cross_entropy_mean = tf.reduce_mean(cross_entropy)
loss = cross_entropy_mean # + tf.add_n(tf.get_collection('losses'))
learning_rate = tf.train.exponential_decay(LearningRateBase, global_step,
30, LearningRateDecay)
def train():
x1 = tf.placeholder(tf.float32, [None, Inference.input1Node],
name='x1-input')
x2 = tf.placeholder(tf.float32, [None, Inference.input2Node],
name='x2-input')
keeprate = tf.placeholder(tf.float32)
label = tf.placeholder(tf.float32, [None, Inference.outputNode],
name='label-input')
one = tf.ones_like(label)
zero = tf.zeros_like(label)
t = Inference.inference(x1, x2, keeprate)
y = tf.where(t < 0.5, zero, one)
global_step = tf.Variable(0, False)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=t, labels=tf.argmax(label, 1))
cross_entropy_mean = tf.reduce_mean(cross_entropy)
loss = cross_entropy_mean # + tf.add_n(tf.get_collection('losses'))
train_step = tf.train.AdamOptimizer().minimize(loss, global_step)
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(label, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
auc_value, auc_op = tf.metrics.auc(tf.argmax(label, 1), tf.argmax(y, 1))
traindata, trainlabel, traindesc = GetDataSet.getNextBatch(True, False)
testdata, testlabel, testdesc = GetDataSet.getNextBatch(False, False)
with tf.Session() as sess:
tf.global_variables_initializer().run()
tf.local_variables_initializer().run()
for i in range(6000):
x1s, labels, x2s = GetDataSet.getNextBatch()
sess.run(train_step,
feed_dict={
x1: x1s,
x2: x2s,
label: labels,
keeprate: 0.75
})
if i % 10 == 0:
print(i)
a, b = sess.run([accuracy, loss],
feed_dict={
x1: traindata,
x2: traindesc,
label: trainlabel,
keeprate: 1
})
trainacc.append(a)
trainloss.append(b)
a, b, c = sess.run([accuracy, loss, auc_op],
feed_dict={
x1: testdata,
x2: testdesc,
label: testlabel,
keeprate: 1
})
testacc.append(a)
testloss.append(b)
auc.append(
sess.run(auc_value,
feed_dict={
x1: testdata,
x2: testdesc,
label: testlabel,
keeprate: 1
}))