-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy path核概率密度plan2.py
191 lines (166 loc) · 6.5 KB
/
核概率密度plan2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
import gc
import math
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
import kalepy as kale
import pandas as pd
from kalepy.plot import nbshow
import numpy as np
import matplotlib.pyplot as plt
import kalepy as kale
from scipy.interpolate import interp1d
def makeSampleData(df):
samples = df['timeByMin'].values
# 创建三个版本的样本数组
samples_mirror1 = samples - 1440
# 截取大于-720的部分
samples_mirror1 = samples_mirror1[samples_mirror1 > -120]
samples_mirror2 = samples + 1440
# 截取小于720的部分
samples_mirror2 = samples_mirror2[samples_mirror2 < 1440 + 120]
# 将三个数组合并成一个
samples_all = np.concatenate([samples, samples_mirror1, samples_mirror2])
weights = np.ones(len(samples_all))
weights = weights / sum(weights)
return samples_all, weights
# 输出np数组的平均值、标准差
def printMeanAndStd(samples):
print("mean:" + str(np.mean(samples)))
print("std:" + str(np.std(samples)))
def makeDiySampleData(dfs, dfu):
num = len(dfs)
diyNum = len(dfu)
# 定义常数
a = 1
b = 0.004
e = np.exp(1) # 自然对数的底数
UW = (a + e ** (-b * diyNum)) ** 5 * 4
# print("UW"+str(UW))
# 使用merge函数找到两个DataFrame中相同的部分
common = dfs.merge(dfu, how='inner')
# 从dfs中删除在dfu中也存在的行
dfs = dfs[(~dfs.isin(common))].dropna()
dfs = dfs.sample(n=num - diyNum, random_state=2023)
weights = np.ones(num - diyNum)
df = pd.concat([dfs, dfu], ignore_index=True)
weightsD = np.ones(diyNum) * UW
samples = df['timeByMin'].values
# makeSampleData(samples)
# 创建三个版本的样本数组
samples_mirror1 = samples - 1440
# 截取大于-720的部分
samples_mirror1 = samples_mirror1[samples_mirror1 > -120]
samples_mirror2 = samples + 1440
# 截取小于720的部分
samples_mirror2 = samples_mirror2[samples_mirror2 < 1440 + 120]
# 将三个数组合并成一个
samples_all = np.concatenate([samples, samples_mirror1, samples_mirror2])
weightsss = np.concatenate([weights, weightsD, np.ones(len(samples_all) - num)])
weightsss = weightsss / sum(weightsss)
return samples_all, weightsss
# 计算
def getP(points, density, start_time):
interp_density = interp1d(points, density)
# 创建一个时间范围
# time_range = np.linspace(start_time, end_time, 60*timeLength) # 可以调整最后一个参数以改变时间范围的精度
# 计算每个时间点的密度
density_range = interp_density(start_time)
# print(density_range)
# 计算总概率
# total_probability = np.sum(density_range) * (end_time - start_time) / len(time_range)
# # 映射到全局概率
# total_probability = total_probability
# print(f"The probability of event A happening between {start_time} and {end_time} is {total_probability}")
return density_range
def getAKdePdf(df=pd.read_csv('./data/模型数据/df.csv'), isWorkDay=0, BD=1 / 20, show=False):
df = df.copy()
df = initPD(df)
df = df[df['is_workday_or_holiday'] == isWorkDay]
# 如果包含action_taken列，则删除
if 'action_taken' in df.columns:
df = df[df['action_taken'] == 1]
# 将小时、分钟和秒转换为以分钟为单位的时间
if len(df) > 30000:
df = df.sample(n=30000, random_state=2023)
df['timeByMin'] = df['hour'] * 60 + df['minute'] + df['second'] / 60
samples_all, weights = makeSampleData(df)
points, density = kale.density(samples_all, weights=weights, probability=True, bandwidth=BD, reflect=True)
if show:
# 绘制核密度估计
plt.figure(figsize=(10, 6))
plt.plot(points / 60, density * 1.1, 'k-', lw=2.0, alpha=0.8, label='KDE')
plt.ylim([0, 0.0015])
plt.xlim([0, 24])
plt.xlabel('timeByHour for ALL')
plt.ylabel('Density')
plt.legend()
plt.show()
return [points, density]
def getPfromPdf(pdf, startTime=10 * 60 + 2):
points = pdf[0]
density = pdf[1]
# 计算概率
density_range = getP(points, density, startTime)
return density_range
def getDKdePdf(df=pd.read_csv('./data/模型数据/df.csv'), isWorkDay=0, userID="", BD=1 / 20, show=False):
df = df.copy()
df = initPD(df)
df = df[df['is_workday_or_holiday'] == isWorkDay]
# 如果包含action_taken列，则删除
if 'action_taken' in df.columns:
df = df[df['action_taken'] == 1]
if len(df) > 30000:
dfs = df.sample(n=30000, random_state=2023)
dfu = df[df['user_id'] == userID]
samples_new, weights_new = makeDiySampleData(dfs, dfu)
Npoints, Ndensity = kale.density(samples_new, weights=weights_new, probability=True, bandwidth=BD, reflect=True)
if show:
# 绘制核密度估计
plt.figure(figsize=(10, 6))
plt.plot(Npoints / 60, Ndensity * 1.1, 'k-', lw=2.0, alpha=0.8, label='NKDE')
plt.ylim([0, 0.0015])
plt.xlim([0, 24])
plt.xlabel('timeByHour for ' + str(userID))
plt.ylabel('Density')
plt.legend()
plt.show()
return [Npoints, Ndensity]
def initPD(historyDf):
if "hour" not in historyDf.columns:
historyDf['timestamp'] = pd.to_datetime(historyDf['timestamp'])
historyDf['hour'] = historyDf['timestamp'].dt.hour
historyDf['minute'] = historyDf['timestamp'].dt.minute
historyDf['second'] = historyDf['timestamp'].dt.second
historyDf = historyDf.drop(['timestamp'], axis=1)
return historyDf
def getMultiplier(opdf=None, updf=None, time=21 * 60 + 2, show=False):
if opdf is None:
a = 1.0
else:
a = getPfromPdf(opdf, time)
b = getPfromPdf(updf, time)
c = (b / a) ** 2
if show:
print(c)
bouth(opdf, updf)
return c
def bouth(opdf=None, updf=None):
points = opdf[0]
density = opdf[1]
Npoints = updf[0]
Ndensity = updf[1]
plt.figure(figsize=(10, 6))
plt.plot(points / 60, density * 1.1, lw=2.0, alpha=0.8, label='ALL KDE', color="red")
plt.plot(Npoints / 60, Ndensity * 1.1, lw=2.0, alpha=0.8, label='USER KDE', color="deepskyblue")
plt.ylim([0, 0.0015])
plt.xlim([0, 24])
plt.xlabel('timeByHour')
plt.ylabel('Density')
plt.legend()
plt.show()
if __name__ == "__main__":
df = pd.read_csv('./data/模型数据/user_history_False_False_False.json')
apdf = getAKdePdf(df, 0, 1 / 20, True)
dpdf = getDKdePdf(df, 0, "0005b30eb7d0182124c853809d726912", 1 / 20, True)
getMultiplier(apdf, dpdf, 21 * 60 + 2, True)