data = pd.read_csv('../dataset/combined/Combined_News_DJIA.csv')
train = data[data['Date'] < '2015-01-01']
test = data[data['Date'] > '2014-12-31']
example = train.iloc[3, 10]
print 'EXAMPLE 1 -- ', example
example2 = example.lower()
print'EXAMPLE 2 -- ', example2
example3 = TfidfVectorizer().build_tokenizer()(example2)
print 'EXAMPLE 3 -- ', example3
pd.DataFrame([[x,example3.count(x)] for x in set(example3)], columns = ['Word', 'Count'])
trainheadlines = []
for row in range(0,len(train.index)):
trainheadlines.append(' '.join(str(x) for x in train.iloc[row,2:27]))
basicvectorizer = TfidfVectorizer()
# 将trainheadlines转换为稀疏矩阵，表示每日的新闻里每个词出现的次数
basictrain = basicvectorizer.fit_transform(trainheadlines)
# basictrain is a sparse matrix
# (x,y),x组数据，y组特征
print 'The shape of the sparce matrix -- ',basictrain.shape
basicmodel = LogisticRegression() # 逻辑回归分类器
basicmodel = basicmodel.fit(basictrain, train["Label"]) # 输入数据，分类目标，开始训练
