def killAll():
'''结束所有记录的进程
'''
from util import Process
for app in App._totalapps:
Process(app.ProcessId).terminate()
del App._totalapps[:]
def add_past_purchase_key(_df):
"""
Add purchased couponkey.
- purchased couponkey is purchases of coupons with the same couponkey by the user.
- exclude the information in the period where the coupon is.
"""
key1 = Process.to_key1(_df)
idx1 = zip(_df.USER_ID, key1)
idx2 = zip(_df.USER_ID, key1, _df.period)
pivot = past_buy_key
pivot_period = past_buy_key_period
v1 = pivot.reindex(idx1).values
v1 = np.where(np.isnan(v1), 0, v1)
v2 = pivot_period.reindex(idx2).values
v2 = np.where(np.isnan(v2), 0, v2)
values = v1 - v2
_df2 = pd.DataFrame(values, index=_df.index, columns=["past_key"])
return pd.concat([_df, _df2], axis=1)
def add_past_purchase_key(_df):
"""
Add purchased couponkey.
- purchased couponkey is purchases of coupons with the same couponkey by the user.
- exclude the information in the period where the coupon is.
"""
key1 = Process.to_key1(_df)
idx1 = zip(_df.USER_ID, key1)
idx2 = zip(_df.USER_ID, key1, _df.period)
pivot = past_buy_key
pivot_period = past_buy_key_period
v1 = pivot.reindex(idx1).values
v1 = np.where(np.isnan(v1), 0, v1)
v2 = pivot_period.reindex(idx2).values
v2 = np.where(np.isnan(v2), 0, v2)
values = v1 - v2
_df2 = pd.DataFrame(values, index=_df.index, columns=["past_key"])
return pd.concat([_df, _df2], axis=1)
# load files
detail_tr = pd.read_pickle("../model/detail_tr.pkl")
coupons = pd.read_pickle("../model/coupons.pkl")
coupon_tr = pd.read_pickle("../model/coupon_tr.pkl")
users = pd.read_pickle("../model/users.pkl")
# count purchases for each coupon ---------------------------
# exclude duplication
detail_tr = detail_tr.groupby(["COUPON_ID","USER_ID"]).size().reset_index().drop(0, axis=1)
detail_cp = detail_tr.groupby("COUPON_ID").size().reset_index().rename(columns={0:"count"})
detail_cp = detail_cp.merge(coupon_tr, on="COUPON_ID")
detail_cp = detail_cp.sort("count")
detail_cp["key1"] = Process.to_key1(detail_cp)
# calculate couponkey and genre popularility ----------------
# couponkey popularity, calculated as mean
popular_key = detail_cp.groupby("key1")["count"].agg([np.mean, np.size])
popular_key = popular_key.rename(columns={"mean":"key_mean", "size":"key_size"})
# genre popularity, calculated as mean
popular_genre = detail_cp.groupby("genre")["count"].agg([np.mean, np.size])
popular_genre = popular_genre.rename(columns={"mean":"key_mean", "size":"key_size"})
# couponkey popularity, exclude couponkey of only one sample
popular_key_train = popular_key[popular_key["key_size"] > 1]
# set popularility for each coupon ----------------------
from util import Grouping, Process
from util_logger import get_logger
LOG = get_logger()
LOG.info("start b14")
# load files
users = pd.read_pickle("../model/users.pkl")
coupons = pd.read_pickle("../model/coupons.pkl")
coupon_tr = pd.read_pickle("../model/coupon_tr.pkl")
detail_tr = pd.read_pickle("../model/detail_tr.pkl")
# count purchases per (user, couponkey)
bought = detail_tr.copy()
bought = Grouping.to_group(bought, ["USER_ID", "COUPON_ID"], False)
bought = bought.merge(coupon_tr, on="COUPON_ID", how='left')
bought["key1"] = Process.to_key1(bought)
# remove duplication
bought = Grouping.to_group(bought, ["USER_ID", "period", "key1"], False)
# count purchases per (user, couponkey)
past_buy_key = Grouping.to_group(bought, ["USER_ID", "key1"], True)
past_buy_key = past_buy_key.set_index(["USER_ID", "key1"])
past_buy_key_period = Grouping.to_group(bought, ["USER_ID", "key1", "period"], True)
past_buy_key_period = past_buy_key_period.set_index(["USER_ID", "key1", "period"])
# write
past_buy_key.to_pickle("../model/past_buy_key.pkl")
past_buy_key_period.to_pickle("../model/past_buy_key_period.pkl")
