def GenerateOutcomes(x, z, num_cont, num_bin):
"""
Following the generating procedure defined by Madras in Algorithm 2
"""
# As defined by Madras
num_z = z.shape[1]
w = -11
beta_a = 6
# Algorithm 2
# horizontal concatenation
xz = np.concatenate((x, z), 1)
W = np.ones(xz.shape[1])*.5
# lists to store generated values
y_t0_a0, y_t1_a0, y_t0_a1, y_t1_a1 = list(), list(), list(), list()
mu_t0_a0, mu_t1_a0, mu_t0_a1, mu_t1_a1 = list(), list(), list(), list()
# loop over observations because all need individual beta sample
for obs in xz:
# sample new beta
beta_cont = choice([0, .1, .2, .3, .4], num_cont, p=[.5, .125, .125, .125, .125])
beta_bin = choice([0, .1, .2, .3, .4], num_bin, p=[.6, .1, .1, .1, .1])
beta_z = choice([.4, .6], num_z, p=[.5, .5])
# in x, continuous variables come first
beta = np.concatenate((beta_cont, beta_bin, beta_z), 0)
# calculate y dist
mu1 = np.matmul(np.exp(obs + W), beta)
mu_t0_a0.append(mu1)
mu2 = np.matmul(obs, beta)-w
mu_t1_a0.append(mu2)
mu3 = np.matmul(np.exp(obs + W), beta) + beta_a
mu_t0_a1.append(mu3)
mu4 = np.matmul(obs, beta) - w + beta_a
mu_t1_a1.append(mu4)
# sample new y
y_t0_a0.append(np.random.normal(mu1, 1, 1)[0])
y_t1_a0.append(np.random.normal(mu2, 1, 1)[0])
y_t0_a1.append(np.random.normal(mu3, 1, 1)[0])
y_t1_a1.append(np.random.normal(mu4, 1, 1)[0])
plt_entries = {'y_t0_a0': y_t0_a0, 'y_t1_a0': y_t1_a0, 'y_t0_a1': y_t0_a1, 'y_t1_a1': y_t1_a1}
plt.figure()
plt.title('Generated data')
for label, entry in plt_entries.items():
plt.hist(entry, label=label, alpha=0.5, bins=20)
plt.legend()
plt.show()
y_all = np.transpose(np.vstack((y_t0_a0, y_t1_a0, y_t0_a1, y_t1_a1)))
mu_all = np.transpose(np.vstack((mu_t0_a0, mu_t1_a0, mu_t0_a1, mu_t1_a1)))
# column names should be consistent with above vstack
y_column = 'y_t0_a0, y_t1_a0, y_t0_a1, y_t1_a1'
mu_column = 'mu_t0_a0, mu_t1_a0, mu_t0_a1, mu_t1_a1'
return y_all, mu_all, y_column, mu_column
def draw_hist(heights):
#创建直方图
#第一个参数为待绘制的定量数据，不同于定性数据，这里并没有事先进行频数统计
#第二个参数为划分的区间个数
plt.hist(heights, 100)
plt.xlabel('Heights')
plt.ylabel('Frequency')
plt.title('Heights Of Male Students')
plt.show()
def draw_cumulative_hist(heights):
#创建累积曲线
#第一个参数为待绘制的定量数据
#第二个参数为划分的区间个数
#normed参数为是否无量纲化
#histtype参数为'step'，绘制阶梯状的曲线
#cumulative参数为是否累积
plt.hist(heights, 20, normed=True, histtype='step', cumulative=True)
plt.xlabel('Heights')
plt.ylabel('Frequency')
plt.title('Heights Of Male Students')
plt.show()
'''
import json
from textblob import TextBlob
from wordcloud import WordCloud
import matplotlib.pylot as plt
# Get the JSON data
tweetFile = open("tweets.json", "r")
tweetData = json.load(tweetFile)
tweetFile.close()
polarity_values = []
for tweet in tweetData:
tweets.append(tweet["text"])
giant_string = " ".join(tweets)
tb = TextBlob(tweet_text)
print("{}: {}".format(tweet_text, tb.polarity))
polarity_values.append(tb.polarity)
# bins = [-1, -0.5, 0, 0.5, 1]
plt.hist(polarity_values, bins)
plt.title("tweet polarity")
plt.ylabel("Count of tweets")
plt.xlabel("Polarity")
plt.show()
X_add, y_add = mydat['features'], mydat['labels']
with open('./mydata/train.p', mode='rb') as f:
mytrain = pickle.load(f)
X_mytrain, y_mytrain = mytrain['features'], mytrain['labels']
with open('./mydata/test.p', mode='rb') as f:
mytest = pickle.load(f)
X_mytest, y_mytest = mytest['features'], mytest['labels']
X_train = np.append(X_train_, X_mytrain, axis = 0)
y_train = np.append(y_train_, y_mytrain)
X_test = np.append(X_test_, X_mytest, axis = 0)
y_test = np.append(y_test_, y_mytest)
plt.hist(y_train, bins=50, color='#FF69B4')
#==============================================================================
# Failed attempt to train the model by incrementally increase the number of
# near-zero training examples.
#==============================================================================
#def limit(X, y, s = 700):
# bad = [k for k,v in enumerate(y) if v in [0, -.25, .25]]
# good = list(set(range(0, len(y)))-set(bad))
# new = good + [bad[i] for i in np.random.randint(0,len(bad),s)]
# X,y = X[new,], y[new]
# return X, y
#