def xandy(alldata, traindatestart, traindateend, testdateend):
# 需要测试的数据
teststartdate = traindateend
traindata = onetimeintveraldata(alldata, traindatestart, traindateend)
predictdata = onetimeintveraldata(alldata, teststartdate, testdateend)
groupresult = getusernumberfeature(traindata)
# a = predictdata.groupby("uid").size()
# print a.shape, groupresult.shape
times = pd.DatetimeIndex(predictdata["time"])
dayofweek = times.dayofweek
dayofweek = pd.DataFrame(
dayofweek, index=predictdata.index, columns=["week"])
face = countdiffface(predictdata)
content = predictdata["content"]
strfeature = pd.concat([
content.str.count("\?", ),
content.str.count("【"),
content.str.count("http"),
content.str.count("@"),
content.str.count("#"),
content.str.count("《"),
content.str.count("分享"),
content.str.count("~"),
content.str.count("\[", ),
content.str.count("发起了一个话题"),
content.str.count("转载博"),
content.str.count("（"),
content.str.count("!"),
content.str.len(),
dayofweek,
predictdata[
["uid", "mid", "forward_count", "comment_count", "like_count"]]
# face,
], axis=1)
# 下面的这种merge会是上个月未出现的用户的数据都是为0
train = pd.merge(groupresult, strfeature, left_index=True,
right_on="uid", how="inner")
don_not_have_before = predictdata[~predictdata["mid"].isin(train["mid"])]
# print mids
train.fillna(0)
# print train_x[:, :-4], train_y
# print train_x.values[:, :-1], train_y.values
return train, don_not_have_before
def predictresult(models, traindatestart, traindateend, testdateend):
data = readdata()
don, need_train_data = usercluseter(data)
train_data, don_not_have_before = xandy(
need_train_data, traindatestart, traindateend, testdateend)
pymid_and_ys = testp(models, train_data)
ty = onetimeintveraldata(data, traindateend, testdateend)
a = pd.merge(pymid_and_ys[0], pymid_and_ys[1],
left_index=True, right_index=True, how="outer")
b = pd.merge(
a, pymid_and_ys[2], left_index=True, right_index=True, how="outer")
py = b.fillna(0)
pyy = ty.copy(deep=True)
print id(pyy), id(ty)
# print py.ix[:, :]
# sys, exit()
mindex = pyy["mid"].isin(py.index)
pyy.loc[
mindex, ["forward_count", "comment_count", "like_count"]] = py.values[3:6]
# print pyy.loc[~mindex, ["forward_count", "comment_count", "like_count"]]
pyy.loc[~mindex, ["forward_count", "comment_count", "like_count"]] = 0
f = pyy.values[:, 3:6]
t = ty.values[:, 3:6]
print f, t
print scores(f, t)