def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_inout_analize.txt')
self.__city_dict = {'广州': ['机场南', '广州东站'], }
self.__params = {}
self.__data = {}
self.name = "module_inout_analize"
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_jamanalize.txt')
self.__params = {}
self.__data = {}
self.name = "module_jamanalize"
self.__time_period = [(7, 9), (17, 19)] # 长度必须为2，代表早晚上班时段
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_ticketrate.txt')
self.__params = {}
self.name = "module_ticketrate"
def run(self, df, global_params=None):
# STATUS ==5 的是交易成功的
df['is_success'] = df['order_status'].apply(lambda x: 1 if x == 5 else 0)
status = df.groupby(['is_success']).ticket_num.count()
st_status = df.groupby(['entry_station']).apply(lambda df: np.mean(df['is_success'])).reset_index()
st_status.columns = ['entry_station', 'rate']
st = df.groupby(['entry_station'], as_index=False).ticket_num.count()
st_status = st_status.merge(st, on=['entry_station'], how='left')
st_status['success_ticket'] = st_status['ticket_num'] * st_status['rate']
st_status['success_ticket'] = st_status['success_ticket'].apply(lambda x: int(round(x)))
st_status['fail_ticket'] = st_status['ticket_num'] - st_status['success_ticket']
st_status = st_status.sort_values('rate', ascending=True)
# print(st_status.head())
self.__params['M5_total_rate'] = status[1] / status.sum()
self.__params['M5_tail_stations'] = st_status.entry_station[:5].tolist()
self.__params['M5_success_tk'] = st_status.success_ticket[:5].tolist()
self.__params['M5_fail_tk'] = st_status.fail_ticket[:5].tolist()
self.__params['M5_rate'] = st_status.rate[:5].tolist()
# print(self.__params)
params = {}
params['M5_total_rate'] = self.__params['M5_total_rate']
params['M5_tail_stations'] = self.__params['M5_tail_stations']
params['M5_success_tk'] = self.__params['M5_success_tk']
params['M5_fail_tk'] = self.__params['M5_fail_tk']
params['M5_rate'] = self.__params['M5_rate']
self.__data = params
global_params['M5_total_rate'] = '%.2f'%(params['M5_total_rate']*100)
global_params['M5_tail_stations'] = '、'.join(params['M5_tail_stations'][:3])
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_userstay.txt')
self.__params = {}
self.name = "module_userstay"
def run(self, df, global_params=None):
# STATUS ==5 的是交易成功的
df_suc = df[df['order_status'] == 5].copy()
single_ft = df_suc.groupby(['owner_id'
])['reg_date'].min().reset_index()
single_ft = single_ft.rename(index=str,
columns={'reg_date': 'first_time'})
df_suc = df_suc.merge(single_ft, on=['owner_id'], how='left')
# print(df_suc[['owner_id', 'first_time', 'entry_date', 'reg_date']].head())
df_suc['time'] = pd.to_datetime(df_suc['first_time'],
format='%Y-%m-%d %H:%M:%S')
df_suc['reg_date'] = df_suc['reg_date'].astype(str)
df_suc['day'] = df_suc['time'].apply(lambda x: x.dayofweek)
df_suc['is_weekend'] = df_suc['day'].apply(lambda x: 1
if x == 0 or x == 6 else 0)
us_date, us_num = day_actitve_num_print(
df_suc[df_suc['is_weekend'] == 1])
self.__params['US_date'] = us_date
self.__params['US_num'] = us_num
# print(self.__params)
params = {}
params['US_date'] = self.__params['US_date']
params['US_num'] = self.__params['US_num']
self.__data = params
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module0.txt')
self.__params = {}
for param in self.__templete.get_params():
self.__params[param] = ''
def run(self, df):
pass
def maketext(self):
return self.__templete.format_templet(self.__params)
def makedata(self): # js? db? whatever.
return ''
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module0.txt')
self.__params = {}
self.__data = {}
self.name = "module_usertimes"
def run(self, df, global_params=None):
if global_params is None:
global_params = {}
user_counts = df[df.order_status == 5].groupby(
'owner_id').order_no.count().reset_index().rename(
columns={'order_no': 'user_counts'})
user_counts['user_counts_level'] = user_counts.user_counts.map(
lambda x: '1次' if x == 1 else '2-5次' if x <= 5 else '6-20次'
if x <= 20 else '20次以上')
user_counts = user_counts.groupby(
'user_counts_level').owner_id.count().to_dict()
# 填数据
self.__data['levels'] = list(user_counts.keys())
self.__data['user_times'] = user_counts
counts_sum = sum(user_counts.values())
self.__data['user_percent'] = {
each: user_counts[each] / counts_sum
for each in user_counts
}
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
import json
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module0.txt')
self.__params = {}
self.__data = {}
self.name = "module_newuseranalize"
def run(self, df, global_params=None):
if global_params is None:
global_params = {}
# 每个用户第一次使用作为新增
user_reg_date = df.groupby('owner_id').reg_date.min().reset_index().rename(
columns={'reg_date': 'user_reg_date'})
# 计数， 每天新增人数
user_reg_date['user_reg_day'] = user_reg_date.user_reg_date.map(lambda x:str(x)[:10])
new_user_day_count = user_reg_date['user_reg_day'].value_counts().sort_index()
# 填数据
self.__data['new_user_day_count'] = [new_user_day_count.index.tolist(), new_user_day_count.tolist()]
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
import json
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
# return ''
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module1.txt')
self.__params = {}
def run(self, df):
# STATUS ==5 的是交易成功的
df_suc = df[df['ORDER_STATUS'] == 5]
tmp = df_suc.groupby(['START_NAME', 'END_NAME'
]).SINGLE_TICKET_NUM.sum().reset_index()
starts = tmp.groupby('START_NAME').SINGLE_TICKET_NUM.sum().sort_values(
ascending=False)
ends = tmp.groupby('END_NAME').SINGLE_TICKET_NUM.sum().sort_values(
ascending=False)
self.__params['start_top10'] = starts[:10].index.tolist()
self.__params['end_top10'] = ends[:10].index.tolist()
self.__params['start_top10_percent'] = starts[:10].sum() / starts.sum(
) * 100
self.__params['end_top10_precent'] = ends[:10].sum() / ends.sum() * 100
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
return ''
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_hotstation.txt')
self.__params = {}
self.name = "module_hotstation"
def run(self, df, global_params=None):
if global_params is None:
global_params = {}
# STATUS ==5 的是交易成功的
df_suc = df[df['order_status'] == 5].copy()
df_suc['entry_date'] = df_suc['entry_date'].astype(str)
df_suc['date'] = df_suc['entry_date'].apply(lambda x: x[0:10])
# df_suc['time'] = pd.to_datetime(df_suc['entry_date'], format='%Y-%m-%d %H:%M:%S')
tmp = df_suc.groupby(['entry_station',
'exit_station']).ticket_num.sum().reset_index()
starts = tmp.groupby('entry_station').ticket_num.sum().sort_values(
ascending=False)
ends = tmp.groupby('exit_station').ticket_num.sum().sort_values(
ascending=False)
station = pd.DataFrame()
station['st_name'] = list(
set(list(df_suc['entry_station']) + list(df_suc['exit_station'])))
starts = pd.DataFrame(starts)
starts = starts.reset_index()
ends = pd.DataFrame(ends)
ends = ends.reset_index()
station = station.rename(index=str,
columns={'st_name': 'entry_station'})
# print(station.head())
station = station.merge(starts, on=['entry_station'], how='left')
station = station.rename(index=str,
columns={'entry_station': 'exit_station'})
station = station.merge(ends, on=['exit_station'], how='left')
station['total_ticket'] = station.ticket_num_x + station.ticket_num_y
station = station.sort_values(by=['total_ticket'], ascending=False)
# print(station.head())
trend = df_suc.groupby(['entry_station',
'date']).ticket_num.sum().reset_index()
# print(trend.head())
routes_groupby = df_suc.groupby([
'entry_station', 'exit_station'
]).ticket_num.sum().sort_values(ascending=False).index.tolist()[:10]
routes = reduce(operator.add, routes_groupby)
routes = sorted(dict(Counter(routes)).items(),
key=lambda x: x[1],
reverse=True)[:10]
self.__params['M2_hotstations'] = station[
station.total_ticket >
station.total_ticket.mean()].exit_station.tolist()
self.__params['M2_hotroutes'] = routes_groupby
self.__params['M2_hotroutes_topstations'] = [
route[0] for route in routes
][:5]
# print(self.__params)
params = {}
params['M2_hotstations'] = self.__params['M2_hotstations']
params['M2_hotstations_ticketnum'] = station[
station.total_ticket >
station.total_ticket.mean()].total_ticket.tolist()
params['M2_hotstations_trend1_time'] = trend[
trend.entry_station == self.__params['M2_hotstations']
[0]].date.tolist()
params['M2_hotstations_trend1'] = trend[trend.entry_station ==
self.__params['M2_hotstations']
[0]].ticket_num.tolist()
params['M2_hotstations_trend2_time'] = trend[
trend.entry_station == self.__params['M2_hotstations']
[1]].date.tolist()
params['M2_hotstations_trend2'] = trend[trend.entry_station ==
self.__params['M2_hotstations']
[1]].ticket_num.tolist()
params['M2_hotstations_trend3_time'] = trend[
trend.entry_station == self.__params['M2_hotstations']
[2]].date.tolist()
params['M2_hotstations_trend3'] = trend[trend.entry_station ==
self.__params['M2_hotstations']
[2]].ticket_num.tolist()
params['M2_hotroutes'] = routes_groupby
params['M2_hotroutes_ticketnum'] = df_suc.groupby([
'entry_station', 'exit_station'
]).ticket_num.sum().sort_values(ascending=False).tolist()[:10]
self.__data = params
global_params['M2_hotstations'] = '、'.join(params['M2_hotstations'])
for i in range(3):
global_params['M2_top%d' % (i + 1)] = params['M2_hotstations'][i]
for i in range(3):
global_params['M2_hotroutes'] = '、'.join(
'-'.join(each) for each in params['M2_hotroutes'][:2])
global_params['M2_hotroutes_topstations'] = '、'.join(
self.__params['M2_hotroutes_topstations'][:3])
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_inout_analize.txt')
self.__city_dict = {'广州': ['机场南', '广州东站'], }
self.__params = {}
self.__data = {}
self.name = "module_inout_analize"
def run(self, df, global_params=None):
if global_params is None:
global_params = {}
df_suc = df[df.order_status == 5][['ticket_num', 'entry_date', 'entry_station', 'exit_station']]
df_suc['weekday'] = df_suc.entry_date.map(lambda x: x.weekday())
df_suc['date'] = df_suc.entry_date.map(lambda x: str(x)[:10])
stations = self.__city_dict[global_params.get('city', '广州')]
# 日进出站人数
entry_day_nums = {}
exit_day_nums = {}
# # 日进出站排名
# entry_ranks = {}
# exit_ranks = {}
# 周末去向分布
weekend_exits = {}
for station in stations:
# 日进出站人数
entry_day_count = df_suc[df_suc.entry_station == station].groupby('date').ticket_num.sum().sort_index()
exit_day_count = df_suc[df_suc.exit_station == station].groupby('date').ticket_num.sum().sort_index()
# # 日进出站排名
# entry_rank = df_suc[df_suc.entry_station == station].groupby('entry_station').ticket_num.sum().sort_values(
# ascending=False).head(20)
# exit_rank = df_suc[df_suc.exit_station == station].groupby('exit_station').ticket_num.sum().sort_values(
# ascending=False).head(20)
# 各站点客流去向,放进字典
entry_tmp = df_suc[df_suc.entry_station == station].groupby('exit_station').ticket_num.sum().sort_values(
ascending=False).head(20)
entry_day_nums[station] = entry_day_count
exit_day_nums[station] = exit_day_count
# entry_ranks[station] = entry_rank
# exit_ranks[station] = exit_rank
weekend_exits[station] = entry_tmp
# 填参数
self.__params['stations'] = stations
self.__params['stations_all'] = '，'.join(stations)
for i in range(len(stations)):
self.__params['st_%d_wk_top3' % i] = '、'.join(weekend_exits[stations[i]].head(3).index.tolist())
# 填数据
self.__data['stations'] = stations
self.__data['entry_nums'] = {station: [entry_day_nums[station].index.tolist(), entry_day_nums[station].tolist()]
for station in stations}
self.__data['exit_nums'] = {station: [exit_day_nums[station].index.tolist(), exit_day_nums[station].tolist()]
for station in stations}
# self.__data['entry_ranks'] = {station:[entry_ranks[station].index.tolist(), entry_ranks[station].tolist()] for station in stations}
# self.__data['exit_ranks'] =
self.__data['weekend_exits'] = {
station: [weekend_exits[station].index.tolist(), weekend_exits[station].tolist()] for station in stations}
global_params['stations_all'] = '、'.join(stations)
for i in range(len(stations)):
global_params['stations[%d]'%i] = stations[i]
global_params['st_%d_wk_top3'%i] = '、'.join(self.__data['weekend_exits'][stations[i]][0][:3])
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
import json
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_peopleflow.txt')
self.__params = {}
self.name = "module_peopleflow"
def run(self, df, global_params=None):
# STATUS ==5 的是交易成功的
df_suc = df[df['order_status'] == 5].copy()
df_suc['time'] = pd.to_datetime(df_suc['entry_date'],
format='%Y-%m-%d %H:%M:%S')
df_suc['hour'] = df_suc['time'].apply(lambda x: x.hour)
tmp = df_suc.groupby(['entry_station',
'hour']).ticket_num.sum().reset_index()
starts = df_suc.groupby('entry_station').ticket_num.sum().sort_values(
ascending=False)
hour = tmp[tmp.entry_station == starts.index[0]][tmp.ticket_num > (
tmp[tmp.entry_station == starts.index[0]].ticket_num.mean()) * 1.3]
# print(hour.head())
self.__params['M3_stations'] = starts[:3].index.tolist()
self.__params['M3_station0_t1'] = hour.hour.min()
self.__params['M3_station0_t2'] = hour.hour.max()
# print(self.__params)
params = {}
params['M3_stations'] = self.__params['M3_stations']
params['M3_stations_trend1_time'] = tmp[tmp.entry_station ==
self.__params['M3_stations']
[0]].hour.tolist()
params['M3_stations_trend1'] = tmp[tmp.entry_station ==
self.__params['M3_stations']
[0]].ticket_num.tolist()
params['M3_stations_trend2_time'] = tmp[tmp.entry_station ==
self.__params['M3_stations']
[1]].hour.tolist()
params['M3_stations_trend2'] = tmp[tmp.entry_station ==
self.__params['M3_stations']
[1]].ticket_num.tolist()
params['M3_stations_trend3_time'] = tmp[tmp.entry_station ==
self.__params['M3_stations']
[2]].hour.tolist()
params['M3_stations_trend3'] = tmp[tmp.entry_station ==
self.__params['M3_stations']
[2]].ticket_num.tolist()
self.__data = params
global_params['M3_station0_t1'] = self.__params['M3_station0_t1']
global_params['M3_station0_t2'] = self.__params['M3_station0_t2']
for i in range(3):
global_params['M3_stations[%d]' %
i] = self.__params['M3_stations'][i]
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_ticketway.txt')
self.__params = {}
self.name = "module_ticketway"
def run(self, df, global_params=None):
# STATUS ==5 的是交易成功的
# payment_type source
payment_type_dict = {
0: '支付宝',
1: '中移动',
2: '支付宝网上购票',
3: '微信支付',
4: '微信扫码支付',
5: '翼支付',
6: '支付宝网页支付',
7: '微信公众号支付',
8: '首信易支付',
9: '中移动WAP支付',
10: '银联支付',
11: '银联支付',
12: '微信小程序支付'
}
source_dict = {
1: '盘缠ios',
2: '盘缠android',
3: '插件ios',
4: '插件android',
5: 'h5公众号或扫码支付',
6: '非闪客蜂公众号',
7: '咖啡',
8: '长沙ios',
9: '长沙android'
}
df_suc = df[df['order_status'] == 5].copy()
tmp = df_suc.groupby(['entry_station',
'exit_station']).ticket_num.sum().reset_index()
starts = df_suc.groupby('entry_station').ticket_num.sum().sort_values(
ascending=False)
df_suc.payment_type = df_suc.payment_type.fillna('0')
df_suc.payment_type = df_suc.payment_type.astype(int)
df_suc.source = df_suc.source.fillna('5')
df_suc.source = df_suc.source.astype(int)
df_suc.payment_type = df_suc.payment_type.map(payment_type_dict)
df_suc.source = df_suc.source.map(source_dict)
df_suc = df_suc[df_suc.entry_station.isin(starts[:10].index.tolist())]
total = df_suc.shape[0]
type = df_suc.groupby(
['payment_type']).ticket_num.count().sort_values(ascending=False)
source = df_suc.groupby(
['source']).ticket_num.count().sort_values(ascending=False)
type = type / total
source = source / total
# print(type.head())
# print(source.head())
self.__params['M4_stations'] = starts[:10].index.tolist()
self.__params['M4_top_methods'] = source.index.tolist()
self.__params['M4_top_source'] = type.index.tolist()
self.__params['M4_top_perc'] = source.tolist()
self.__params['M4_top_perc_source'] = type.tolist()
# print(self.__params)
params = {}
params['M4_top_methods'] = self.__params['M4_top_methods']
params['M4_top_perc'] = self.__params['M4_top_perc']
params['M4_top_source'] = self.__params['M4_top_source']
params['M4_top_perc_source'] = self.__params['M4_top_perc_source']
self.__data = params
global_params['M4_stations'] = self.__params['M4_stations']
global_params['M4_top_methods[0]'] = self.__params['M4_top_methods'][0]
global_params['M4_top_perc[0]'] = '%.2f' % (
self.__params['M4_top_perc'][0] * 100)
if len(self.__params['M4_top_methods']) > 1:
global_params['M4_top_methods[1]'] = self.__params[
'M4_top_methods'][1]
global_params['M4_top_perc[1]'] = '%.2f' % (
self.__params['M4_top_perc'][1] * 100)
global_params['M4_top_source[0]'] = self.__params['M4_top_source'][0]
global_params['M4_top_perc_source[0]'] = '%.2f' % (
self.__params['M4_top_perc_source'][0] * 100)
if len(self.__params['M4_top_source']) > 1:
global_params['M4_top_source[1]'] = self.__params['M4_top_source'][
1]
global_params['M4_top_perc_source[1]'] = '%.2f' % (
self.__params['M4_top_perc_source'][1] * 100)
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_details.txt')
self.__params = {}
self.name = "module_details"
def run(self, df, global_params=None):
# STATUS ==5 的是交易成功的
df_suc = df[df['order_status'] == 5].copy()
df_suc = df_suc[~pd.isnull(df_suc.reg_date)]
df_suc.reg_date = df_suc.reg_date.fillna(df_suc.reg_date.min())
df_suc['datetime'] = pd.to_datetime(df_suc['reg_date'],
format='%Y-%m-%d %H:%M:%S')
df_suc['year'] = df_suc['reg_date'].map(lambda x: x.year)
df_suc['month'] = df_suc['reg_date'].map(lambda x: x.month)
df_suc['day'] = df_suc['reg_date'].map(lambda x: x.day)
df_suc['hour'] = df_suc['reg_date'].map(lambda x: x.hour)
df_suc['reg_date'] = df_suc['reg_date'].astype(str)
df_suc['reg_month'] = df_suc['reg_date'].map(lambda x: x[0:7])
df_suc['reg_day'] = df_suc['reg_date'].map(lambda x: int(x[8:10]))
df_suc['dayofweek'] = df_suc['datetime'].apply(lambda x: x.dayofweek)
single_ft = df_suc.groupby(['owner_id'
])['reg_date'].min().reset_index()
single_ft = single_ft.rename(index=str,
columns={'reg_date': 'first_time'})
df_suc = df_suc.merge(single_ft, on=['owner_id'], how='left')
df_suc = df_suc[~pd.isnull(df_suc.first_time)]
df_suc['first_time'] = df_suc['first_time'].astype(str)
single_ft = df_suc.groupby('owner_id')['reg_date'].max().reset_index()
single_ft = single_ft.rename(index=str,
columns={'reg_date': 'last_time'})
df_suc = df_suc.merge(single_ft, on=['owner_id'], how='left')
df_suc = df_suc[~pd.isnull(df_suc.last_time)]
df_suc['last_time'] = df_suc['last_time'].astype(str)
df_suc['first_month'] = df_suc['first_time'].map(lambda x: x[0:7])
df_suc['first_date_obj'] = df_suc['first_time'].map(
lambda x: datetime.strptime(x, '%Y-%m-%d %H:%M:%S').date())
df_suc['reg_date_obj'] = df_suc['reg_date'].map(
lambda x: datetime.strptime(x, '%Y-%m-%d %H:%M:%S').date())
df_suc['tmp'] = df_suc['reg_date_obj'] - df_suc['first_date_obj']
df_suc['day_interval'] = df_suc['tmp'].map(lambda x: x.days)
df_suc.drop(['first_date_obj', 'reg_date_obj', 'tmp'],
axis=1,
inplace=True)
params = {}
params['D_new_cont_month_fm'] = {}
mon_count = used_month_count(df_suc)
fm = df_suc.reg_date.min()[0:7]
if fm <= df_suc.reg_date.max()[0:7]:
D_new_cont_month = []
D_new_cont_month_people = []
total_people = 0
# print(next_month(fm))
for i in range(month_diff(fm) - 1):
D_new_cont_month.append(i + 1)
people, total_people = used_next_month_count(
df_suc, fm, i + 1, has_used_in_all_next_months)
D_new_cont_month_people.append(people)
D_new_cont_month_ratio = list(
np.array(D_new_cont_month_people) / (total_people + 1))
params['D_new_cont_month_fm'][fm] = dict(
zip(D_new_cont_month, D_new_cont_month_ratio))
# print(params)
D_total_month = list(range(1, month_diff(fm)))
tmp_df = df_suc.groupby(['owner_id']).reg_month.unique().reset_index()
tmp_df['unique_month'] = tmp_df['reg_month'].apply(lambda x: len(x))
D_total_month_people = tmp_df.groupby(
['unique_month']).owner_id.count().reset_index().owner_id.tolist()
D_total_month = D_total_month[:len(D_total_month_people)]
D_total_month_ratio = list(
np.array(D_total_month_people) / sum(D_total_month_people))
# user_used_days_qr: 用户所有使用过的天次，第一次使用为第0天
# user_continue_day_count_qr: 用户最长连续使用天数
# 可以由 user_used_days_qr -> (用户最后一次使用距第一次使用相隔多少天；用户使用频率，即使用的天数/相隔的天数)
user_used_days_qr, user_continue_day_count_qr = longest_used_days(
df_suc[['owner_id', 'day_interval']])
user_continue_day_count_num_qr = dict() # 最长连续使用天数的用户数量
for k in user_continue_day_count_qr.keys():
v = user_continue_day_count_qr[k]
user_continue_day_count_num_qr[
v] = user_continue_day_count_num_qr.get(v, 0) + 1
continue_day_count_df_qr = pd.DataFrame()
count_list = []
num_list = []
for k in user_continue_day_count_num_qr.keys():
count_list.append(k)
num_list.append(user_continue_day_count_num_qr[k])
continue_day_count_df_qr = continue_day_count_df_qr.from_dict({
'continue_day_count':
count_list,
'num':
num_list
})
D_cont_day_people = []
D_cont_day_people.append(
day_ratio(continue_day_count_df_qr, 1, 1, 'continue_day_count'))
D_cont_day_people.append(
day_ratio(continue_day_count_df_qr, 2, 3, 'continue_day_count'))
D_cont_day_people.append(
day_ratio(continue_day_count_df_qr, 4, 7, 'continue_day_count'))
D_cont_day_people.append(
day_ratio(continue_day_count_df_qr, 8, 20, 'continue_day_count'))
D_cont_day_people.append(
day_ratio(continue_day_count_df_qr, 21, 999, 'continue_day_count'))
D_cont_day_ratio = list(
np.array(D_cont_day_people) / len(df_suc.owner_id.unique()))
user_day_count_num_qr = dict() # 总使用天数的用户数量 （最后一次-第一次）
for k in user_used_days_qr.keys():
v = user_used_days_qr[k][-1] + 1
user_day_count_num_qr[v] = user_day_count_num_qr.get(v, 0) + 1
day_count_df_qr = pd.DataFrame()
count_list = []
num_list = []
for k in user_day_count_num_qr.keys():
count_list.append(k)
num_list.append(user_day_count_num_qr[k])
day_count_df_qr = day_count_df_qr.from_dict({
'day_count': count_list,
'num': num_list
})
D_total_day_people = []
D_total_day_people.append(day_ratio(day_count_df_qr, 1, 1,
'day_count'))
D_total_day_people.append(day_ratio(day_count_df_qr, 2, 5,
'day_count'))
D_total_day_people.append(
day_ratio(day_count_df_qr, 6, 20, 'day_count'))
D_total_day_people.append(
day_ratio(day_count_df_qr, 21, 30, 'day_count'))
D_total_day_people.append(
day_ratio(day_count_df_qr, 31, 60, 'day_count'))
D_total_day_people.append(
day_ratio(day_count_df_qr, 61, 90, 'day_count'))
D_total_day_people.append(
day_ratio(day_count_df_qr, 91, 120, 'day_count'))
D_total_day_people.append(
day_ratio(day_count_df_qr, 120, 999, 'day_count'))
D_total_day_ratio = list(
np.array(D_total_day_people) / len(df_suc.owner_id.unique()))
user_ratio_num_qr = {
10: 0,
20: 0,
30: 0,
40: 0,
50: 0,
60: 0,
70: 0,
80: 0,
90: 0,
100: 0,
101: 0
}
for k in user_used_days_qr.keys():
n = len(user_used_days_qr[k])
last_day = user_used_days_qr[k][-1] + 1
ratio = n / last_day
if ratio <= 0.1:
user_ratio_num_qr[10] = user_ratio_num_qr[10] + 1
elif ratio <= 0.2:
user_ratio_num_qr[20] = user_ratio_num_qr[20] + 1
elif ratio <= 0.3:
user_ratio_num_qr[30] = user_ratio_num_qr[30] + 1
elif ratio <= 0.4:
user_ratio_num_qr[40] = user_ratio_num_qr[40] + 1
elif ratio <= 0.5:
user_ratio_num_qr[50] = user_ratio_num_qr[50] + 1
elif ratio <= 0.6:
user_ratio_num_qr[60] = user_ratio_num_qr[60] + 1
elif ratio <= 0.7:
user_ratio_num_qr[70] = user_ratio_num_qr[70] + 1
elif ratio <= 0.8:
user_ratio_num_qr[80] = user_ratio_num_qr[80] + 1
elif ratio <= 0.9:
user_ratio_num_qr[90] = user_ratio_num_qr[90] + 1
elif n > 1:
user_ratio_num_qr[100] = user_ratio_num_qr[100] + 1
else:
user_ratio_num_qr[101] = user_ratio_num_qr[101] + 1
tmp_list = list(user_ratio_num_qr.values())
D_user_ratio_people = [tmp_list[-1]] + tmp_list[:-1]
D_user_ratio_ratio = list(
np.array(D_user_ratio_people) / len(df_suc.owner_id.unique()))
user_used_month, user_used_mcount = user_months_and_count(df_suc)
user_mw_dict = user_month_week_count(df_suc)
user_ft_lt_dict = user_ft_lt(df_suc)
user_ww_dict = user_work_week_num(df_suc)
u_ids = []
u_mds = []
for user_id in df_suc['owner_id'].unique():
u_ids.append(user_id)
u_mds.append(
user_model(df_suc, user_id, user_used_month, user_used_mcount,
user_mw_dict, user_ft_lt_dict, user_ww_dict))
user_model_df = pd.DataFrame()
user_model_df = user_model_df.from_dict({
'owner_id': u_ids,
"model": u_mds
})
D_model_people, D_model_ratio = user_model_count(user_model_df)
self.__params['D_new_cont_month'] = D_new_cont_month
self.__params['D_new_cont_month_people'] = D_new_cont_month_people
self.__params['D_new_cont_month_ratio'] = D_new_cont_month_ratio
self.__params['D_total_month'] = D_total_month
self.__params['D_total_month_people'] = D_total_month_people
self.__params['D_total_month_ratio'] = D_total_month_ratio
self.__params['D_cont_day'] = [
'连续使用1天', '连续使用2~3天', '连续使用4~7天', '连续使用8~20天', '连续使用21天以上'
]
self.__params['D_cont_day_people'] = D_cont_day_people
self.__params['D_cont_day_ratio'] = D_cont_day_ratio
self.__params['D_total_day'] = [
'总使用1天', '总使用2~5天', '总使用6~20天', '总使用21~30天', '总使用31~60天',
'总使用61~90天', '总使用91~120天', '总使用121天以上'
]
self.__params['D_total_day_people'] = D_total_day_people
self.__params['D_total_day_ratio'] = D_total_day_ratio
self.__params['D_user_pre'] = [
'只使用1次', '0~0.1', '0.1~0.2', '0.2~0.3', '0.3~0.4', '0.4~0.5',
'0.5~0.6', '0.6~0.7', '0.7~0.8', '0.8~0.9', '0.9~1'
]
self.__params['D_user_pre_people'] = D_user_ratio_people
self.__params['D_user_pre_ratio'] = D_user_ratio_ratio
self.__params['D_model'] = [
'稳定高频周末型', '稳定高频工作型', '稳定高频常用型', '稳定低频周末型', '稳定低频工作型', '稳定低频常用型',
'一段时间之后不用型', '突发周末型', '突发工作型', '突发常用型', '本月新用户周末型', '本月新用户工作型',
'本月新用户常用型'
]
self.__params['D_model_people'] = D_model_people
self.__params['D_model_ratio'] = D_model_ratio
# print(self.__params)
for k, v in self.__params.items():
if 'people' not in k and 'ratio' not in k:
params[k] = {}
for i, vv in enumerate(v):
params[k][vv] = self.__params[k + '_ratio'][i]
params['D_model_people'] = dict(
zip(self.__params['D_model'], self.__params['D_model_people']))
self.__data = params
global_params['day_num'] = 30
global_params['month_num'] = 3
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module1.txt')
self.__params = {}
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module0.txt')
self.__params = {}
self.__data = {}
self.name = "module_newuseranalize"
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module0.txt')
self.__params = {}
for param in self.__templete.get_params():
self.__params[param] = ''
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_workholi_cmp.txt')
self.__city_dict = {
'广州': ['机场南', '广州东站'],
}
self.name = "module_workholi_cmp"
self.__params = {}
self.__data = {}
def run(self, df, global_params=None):
if global_params is None:
global_params = {}
df_suc = df[df.order_status == 5][[
'ticket_num', 'entry_date', 'entry_station', 'exit_station'
]]
df_suc['weekday'] = df_suc.entry_date.map(lambda x: x.weekday())
df_suc['hour'] = df_suc.entry_date.map(lambda x: x.hour)
days = df_suc.entry_date.max() - df_suc.entry_date.min()
days = days.round('d').days
workday_seq = (df_suc[df_suc.weekday < 5].groupby(
'hour').ticket_num.sum().sort_index() /
days).map(lambda x: round(x, 3))
holiday_seq = (df_suc[df_suc.weekday >= 5].groupby(
'hour').ticket_num.sum().sort_index() /
days).map(lambda x: round(x, 3))
self.__params['workday_seq_maxhour'] = workday_seq.argmax()
self.__params['workday_seq_maxfluency'] = int(workday_seq.max())
self.__params['holiday_seq_maxhour'] = holiday_seq.argmax()
self.__params['holiday_seq_maxfluency'] = int(holiday_seq.max())
workday_seqs = {}
holiday_seqs = {}
for station in self.__city_dict[global_params.get('city', '广州')]:
workday_seqs[station] = (
df_suc[(df_suc.entry_station == station)
& (df_suc.weekday < 5)].groupby(
'hour').ticket_num.sum().sort_index() /
days).map(lambda x: round(x, 3))
holiday_seqs[station] = (
df_suc[(df_suc.entry_station == station)
& (df_suc.weekday >= 5)].groupby(
'hour').ticket_num.sum().sort_index() /
days).map(lambda x: round(x, 3))
# 填数据
self.__data['workday_full_seq'] = [
workday_seq.index.tolist(),
workday_seq.tolist()
]
self.__data['holiday_full_seq'] = [
holiday_seq.index.tolist(),
holiday_seq.tolist()
]
self.__data['stations'] = self.__city_dict[global_params.get(
'city', '广州')]
self.__data['station_workday_seqs'] = {
station: [
workday_seqs[station].index.tolist(),
workday_seqs[station].tolist()
]
for station in self.__data['stations']
}
self.__data['station_holiday_seqs'] = {
station: [
holiday_seqs[station].index.tolist(),
holiday_seqs[station].tolist()
]
for station in self.__data['stations']
}
#
global_params['workday_seq_maxhour'] = self.__params[
'workday_seq_maxhour']
global_params['workday_seq_maxfluency'] = self.__params[
'workday_seq_maxfluency']
global_params['holiday_seq_maxhour'] = self.__params[
'holiday_seq_maxhour']
global_params['holiday_seq_maxfluency'] = self.__params[
'holiday_seq_maxfluency']
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
import json
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_peopleflow.txt')
self.__params = {}
self.name = "module_peopleflow"
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_jamanalize.txt')
self.__params = {}
self.__data = {}
self.name = "module_jamanalize"
self.__time_period = [(7, 9), (17, 19)] # 长度必须为2，代表早晚上班时段
def run(self, df, global_params=None):
import pickle as pk
import pandas as pd
import os
if global_params is None:
global_params = {}
# 用最短路径作为预测
city = global_params.get('city', '广州')
if not os.path.exists('routes/%s_route.pk' % city):
return
routes = pk.load(open('routes/%s_route.pk' % city, 'rb'))
# 生成(工作日、周末) x (上班时段、下班时段) 路径人数并排序
df_suc = df[df.order_status == 5]
df_suc['weekday'] = df_suc.entry_date.map(lambda x: x.weekday())
df_suc['hour'] = df_suc.entry_date.map(lambda x: x.hour)
morning = self.__time_period[0]
evening = self.__time_period[1]
df_workday_morning = df_suc[
(df_suc.weekday < 5) & (df_suc.hour >= morning[0]) & (df_suc.hour <= morning[1])].groupby(
['entry_station', 'exit_station']).ticket_num.sum().reset_index().values
df_workday_evening = df_suc[
(df_suc.weekday < 5) & (df_suc.hour >= evening[0]) & (df_suc.hour <= evening[1])].groupby(
['entry_station', 'exit_station']).ticket_num.sum().reset_index().values
df_holiday_morning = df_suc[
(df_suc.weekday >= 5) & (df_suc.hour >= morning[0]) & (df_suc.hour <= morning[1])].groupby(
['entry_station', 'exit_station']).ticket_num.sum().reset_index().values
df_holiday_evening = df_suc[
(df_suc.weekday >= 5) & (df_suc.hour >= evening[0]) & (df_suc.hour <= evening[1])].groupby(
['entry_station', 'exit_station']).ticket_num.sum().reset_index().values
error_counts = 0
try:
for d in ['workday', 'holiday']:
for t in ['morning', 'evening']:
locals()['%s_%s_res' % (d, t)] = {}
for row in locals()['df_%s_%s' % (d, t)]:
if row[0] == row[1]:
continue
for path in routes[(row[0], row[1])]:
locals()['%s_%s_res' % (d, t)][path] = locals()['%s_%s_res' % (d, t)].get(path, 0) + row[2]
del locals()['df_%s_%s' % (d, t)]
tmp = locals()['%s_%s_res' % (d, t)]
locals()['%s_%s_res' % (d, t)] = pd.DataFrame(
[{'start': each[0], 'end': each[1], 'line':each[2],'fluency': tmp[each]} for each in tmp]).sort_values('fluency',ascending=False)
locals()['%s_%s_res'% (d, t)]['level'] = pd.cut(locals()['%s_%s_res'%(d, t)]['fluency'], 5, labels=False)
self.__params['%s_%s_jam_routes'% (d, t)] = '\n'.join([','.join(each) for each in locals()['%s_%s_res'% (d, t)][locals()['%s_%s_res'% (d, t)].level==4][['start','end','line','fluency']].astype(str).values])
del tmp
for jam_lv in range(0,5):
if jam_lv == 0:
self.__data['%s_%s_jam_routes'%(d, t)] = {}
self.__data['%s_%s_jam_routes' % (d, t)][jam_lv] = locals()['%s_%s_res'% (d, t)][locals()['%s_%s_res'% (d, t)].level==jam_lv][['start', 'end', 'level']].as_matrix().tolist()
global_params['M7_AM_busy_routes'] = '、'.join('-'.join([str(x) for x in each[:2]]) for each in self.__data['workday_morning_jam_routes'][4][:3])
except:
error_counts += 1
print('----Jam Analize Finished. %d error meet.'%error_counts)
global_params['M7_AM_lv4_count'] = len(self.__data['workday_morning_jam_routes'][4])
global_params['M7_AM_lv4_rate'] = '%.2f'%(global_params['M7_AM_lv4_count'] / len(routes) * 100)
global_params['M7_PM_busy_routes'] = '、'.join('-'.join([str(x) for x in each[:2]]) for each in self.__data['workday_evening_jam_routes'][4][:3])
global_params['M7_PM_lv4_count'] = len(self.__data['workday_evening_jam_routes'][4])
global_params['M7_PM_lv4_rate'] = '%.2f'%(global_params['M7_PM_lv4_count'] / len(routes) * 100)
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
import json
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
# return ''
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_ticketrate.txt')
self.__params = {}
self.name = "module_ticketrate"
class Module(object):
def __init__(self):
from common.TempletLoader import TempletLoader
self.__templete = TempletLoader('templets/module_dataanalize.txt')
self.name = "module_dataanalize"
self.__params = {}
self.__data = {}
def run(self, df, global_params=None):
if global_params is None:
global_params = {}
# 订单状态分布
order_status = df.order_status.value_counts().to_dict()
order_status_tk = df.groupby('order_status').ticket_num.sum().to_dict()
# print(order_status)
# 订单总数
order_nums = df.shape[0]
# 出票订单数、购票但未取票订单数、失效订单数
order_get_tk = order_status.get(5, 0)
order_notpay = order_status.get(1, 0)
order_pay_notuse = order_status.get(2, 0)
order_canceled = order_status.get(3, 0) + order_status.get(
6, 0) + order_status.get(7, 0)
# 出票数、订单总票数
tk_get_num = order_status_tk.get(5, 0)
tk_get_sum = sum(order_status_tk.values())
# 站点名称数、站点数量、编号数量
# 有过购票行为的站点数，站点编码数
st_num = len(
set(df.entry_station.unique()) | set(df.exit_station.unique()))
st_code_num = len(
set(df.entry_station_code.unique())
| set(df.exit_station_code.unique()))
# 填充至变量中!
# order_num, order_use_num, order_pay_not_use_num, order_fail_num, ticket_num, ticket_use_num
self.__params['order_num'] = order_nums
self.__params['order_use_num'] = order_get_tk
self.__params['order_notpay'] = order_notpay
self.__params['order_pay_not_use_num'] = order_pay_notuse
self.__params['order_fail_num'] = order_canceled
self.__params['ticket_num'] = tk_get_sum
self.__params['ticket_use_num'] = tk_get_num
self.__data['order_num'] = order_nums
self.__data['order_use_num'] = order_get_tk
self.__data['order_notpay'] = order_notpay
self.__data['order_pay_not_use_num'] = order_pay_notuse
self.__data['order_fail_num'] = order_canceled
self.__data['ticket_num'] = tk_get_sum
self.__data['ticket_use_num'] = tk_get_num
def maketext(self, global_params=None):
# 允许传入全局变量， 但局部变量的优先级更高
if global_params and type(global_params) == dict:
for param in global_params:
if param not in self.__params:
self.__params[param] = global_params[param]
# 如果有缺失的变量， 填空字符串
for param in self.__templete.get_params():
if param not in self.__params:
self.__params[param] = ''
# 返回format结果
return self.__templete.format_templet(self.__params)
def makedata(self):
# for each in self.__data:
# print(each, self.__data[each], self.__data[each].dtype)
import json
from common.MyEncoder import MyEncoder
return json.dumps(dict(self.__data), ensure_ascii=False, cls=MyEncoder)
