-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path2-BSTS_Causal_Impact.R
254 lines (236 loc) · 9.37 KB
/
2-BSTS_Causal_Impact.R
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
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
# need to install package first: install.packages("CausalImpact")
library(CausalImpact)
library(foreach)
library(doParallel)
library(mgcv)
library(car)
library(ggplot2)
library(lattice)
library(spdep)
library(sf)
library(tmap)
library(scales)
library(reshape2)
library(forecast)
dat <-
read.csv('C:\\Users\\hsonghua\\Desktop\\CausualImpact\\Daily_Lstaion_Final.csv') # Daily_Lstaion_Final_0806.csv
dat$date <- as.Date(dat$date)
AllCounty <- unique(dat$station_id)
# Setup parallel backend
cores <- detectCores()
cl <- makeCluster(cores[1] - 1)
registerDoParallel(cl)
finalMatrix <- data.frame()
# length(AllCounty)-1800
finalMatrix <-
foreach(
ccount = 1:(length(AllCounty)),
.combine = rbind,
.packages = c("CausalImpact", "reshape2", "lattice", "ggplot2", "forecast")
) %dopar%
{
eachstation <- AllCounty[ccount]
# eachstation <- 40440
# eachstation <- 40090
print(ccount)
dat_Each <- dat[dat$station_id == eachstation,]
rownames(dat_Each) <- NULL
first_enforce_day <- as.numeric(rownames(dat_Each[dat_Each$date == as.Date('2020-03-02'),]))
pre.period <- c(1, first_enforce_day - 1)
post.period <- c(first_enforce_day, nrow(dat_Each))
# We keep a copy of the actual observed response in "post.period.response
post.period.response <- dat_Each$rides[post.period[1]:post.period[2]]
dat_Each$rides[post.period[1]:post.period[2]] <- NA
response <- zoo(dat_Each$rides, dat_Each$date)
#plot(response)
# drop outliers
response_cl <- tsclean(response, replace.missing = TRUE, lambda = NULL)
response_cl[post.period[1]:post.period[2]] <- NA
response <- zoo(response_cl, dat_Each$date)
#plot(response1)
# Build a bsts model
ss <- AddSemilocalLinearTrend(list(), response)
ss <- AddSeasonal(ss, response, nseasons = 7)
ss <- AddMonthlyAnnualCycle(ss, response)
bsts.model1 <- bsts(
response ~ PRCP + TMAX + IsWeekend + Holidays,
state.specification = ss, niter = 2000, data = dat_Each, expected.model.size = 2)
#plot(bsts.model1)
#plot(bsts.model1, "components") + scale_x_date(date_breaks = "1 month", labels = date_format("%b-%Y"), limits = as.Date(c('2019-01-01', '2020-05-01')))
#plot(bsts.model1, "coef")
# Estiamting counterfactual and compare with actual post period response
impact <- CausalImpact(
bsts.model = bsts.model1,
post.period.response = post.period.response)
#impact.plot <- plot(impact) +
# theme_bw(base_size = 20) +
# scale_x_date(date_breaks = "1 month", labels = date_format("%b-%Y"), limits = as.Date(c('2019-01-01', '2020-05-01')))
#plot(impact.plot)
### Get the number of burn-ins to discard
burn <- SuggestBurn(0.1, bsts.model1)
### Get the components
components.withreg <- cbind.data.frame(
colMeans(bsts.model1$state.contributions[-(1:burn), "trend",]),
colMeans(bsts.model1$state.contributions[-(1:burn), "seasonal.7.1",]),
colMeans(bsts.model1$state.contributions[-(1:burn), "Monthly",]),
colMeans(bsts.model1$state.contributions[-(1:burn), "regression",]),
as.Date((dat_Each$date)))
names(components.withreg) <- c("Trend", "Seasonality", "Monthly", "Regression", "Date")
components.withreg$Predict <- impact$series$point.pred
components.withreg$Response <- impact$series$response
components.withreg$Predict_Lower <- impact$series$point.pred.lower
components.withreg$Predict_Upper <- impact$series$point.pred.upper
components.withreg$Raw_Response <- dat_Each$rides
#plot(impact$series$response)
#plot(components.withreg$Raw_Response)
components.withreg <- melt(components.withreg, id.vars = "Date")
names(components.withreg) <- c("Date", "Component", "Value")
# Plot different components
#ggplot(data = components.withreg, aes(x = Date, y = Value)) +
# geom_line() +
# theme_bw() +
# theme(legend.title = element_blank()) +
# ylab("") +
# xlab("") +
# facet_grid(Component ~ ., scales = "free") +
# guides(colour = FALSE) +
# scale_x_date(date_breaks = "12 month", labels = date_format("%b-%Y")) +
# theme(axis.text.x = element_text(angle = -30, hjust = 0))
#, limits = as.Date(c('2010-01-01', '2020-05-01'))
# Coefficient
#colMeans(bsts.model1$coefficients)
#plot(bsts.model1, "coef")
components.withreg$CTNAME <- eachstation
components.withreg
#write.csv(components.withreg, 'finalMatrix_Transit_temp.csv')
}
stopCluster(cl)
write.csv(finalMatrix, 'finalMatrix_Transit_0810.csv')
GetInclusionProbabilities <- function(bsts.object) {
# Pulls code from
# - BoomSpikeSlab::PlotMarginalInclusionProbabilities
# - bsts::PlotBstsCoefficients
burn <- SuggestBurn(0.1, bsts.object)
beta <- bsts.object$coefficients
beta <- beta[-(1:burn), , drop = FALSE]
inclusion.prob <- colMeans(beta != 0)
index <- order(inclusion.prob)
inclusion.prob <- inclusion.prob[index]
# End from BoomSpikeSlab/bsts.
return(data.frame(predictor = names(inclusion.prob),
inclusion.prob = inclusion.prob))
}
# Coeffi
# Setup parallel backend
cores <- detectCores()
cl <- makeCluster(cores[1] - 1)
registerDoParallel(cl)
finalMatrix <- data.frame()
# length(AllCounty)-1800
finalMatrix <-
foreach(
ccount = 1:(length(AllCounty)),
.combine = rbind,
.packages = c("CausalImpact", "reshape2", "lattice", "ggplot2", "forecast")
) %dopar%
{
eachstation <- AllCounty[ccount]
# eachstation <- 40440
# eachstation <- 40090
print(ccount)
dat_Each <- dat[dat$station_id == eachstation,]
rownames(dat_Each) <- NULL
first_enforce_day <- as.numeric(rownames(dat_Each[dat_Each$date == as.Date('2020-03-13'),]))
pre.period <- c(1, first_enforce_day - 1)
post.period <- c(first_enforce_day, nrow(dat_Each))
# We keep a copy of the actual observed response in "post.period.response
post.period.response <- dat_Each$rides[post.period[1]:post.period[2]]
dat_Each$rides[post.period[1]:post.period[2]] <- NA
response <- zoo(dat_Each$rides, dat_Each$date)
#plot(response)
# drop outliers
response_cl <- tsclean(response, replace.missing = TRUE, lambda = NULL)
response_cl[post.period[1]:post.period[2]] <- NA
response <- zoo(response_cl, dat_Each$date)
#plot(response1)
# Build a bsts model
ss <- AddSemilocalLinearTrend(list(), response)
ss <- AddSeasonal(ss, response, nseasons = 7)
ss <- AddMonthlyAnnualCycle(ss, response)
bsts.model1 <- bsts(
response ~ PRCP + TMAX + Holidays + IsWeekend,
state.specification = ss, niter = 2000, data = dat_Each)
#plot(bsts.model1, "coef")
# Estiamting counterfactual and compare with actual post period response
impact <- CausalImpact(
bsts.model = bsts.model1,
post.period.response = post.period.response)
#impact.plot <- plot(impact) +
# theme_bw(base_size = 20) +
# scale_x_date(date_breaks = "1 month", labels = date_format("%b-%Y"), limits = as.Date(c('2019-01-01', '2020-05-01')))
#GetInclusionProbabilities(bsts.model1)
components.withreg <- as.data.frame(bsts.model1$coefficients)
# Coefficient
#colMeans(bsts.model1$coefficients)
#plot(bsts.model1, "coef")
components.withreg$CTNAME <- eachstation
components.withreg
#write.csv(components.withreg, 'finalMatrix_Transit_temp.csv')
}
stopCluster(cl)
write.csv(finalMatrix, 'finalCoeff_Transit_0810.csv')
# Causal impact
# Setup parallel backend
cores <- detectCores()
cl <- makeCluster(cores[1] - 1)
registerDoParallel(cl)
finalMatrix <- data.frame()
# length(AllCounty)-1800
finalMatrix <-
foreach(
ccount = 1:(length(AllCounty)),
.combine = rbind,
.packages = c("CausalImpact", "reshape2", "lattice", "ggplot2", "forecast")
) %dopar%
{
eachstation <- AllCounty[ccount]
# eachstation <- 40440
# eachstation <- 40090
print(ccount)
dat_Each <- dat[dat$station_id == eachstation,]
rownames(dat_Each) <- NULL
first_enforce_day <- as.numeric(rownames(dat_Each[dat_Each$date == as.Date('2020-03-13'),]))
pre.period <- c(1, first_enforce_day - 1)
post.period <- c(first_enforce_day, nrow(dat_Each))
# We keep a copy of the actual observed response in "post.period.response
post.period.response <- dat_Each$rides[post.period[1]:post.period[2]]
dat_Each$rides[post.period[1]:post.period[2]] <- NA
response <- zoo(dat_Each$rides, dat_Each$date)
#plot(response)
# drop outliers
response_cl <- tsclean(response, replace.missing = TRUE, lambda = NULL)
response_cl[post.period[1]:post.period[2]] <- NA
response <- zoo(response_cl, dat_Each$date)
#plot(response1)
# Build a bsts model
ss <- AddSemilocalLinearTrend(list(), response)
ss <- AddSeasonal(ss, response, nseasons = 7)
ss <- AddMonthlyAnnualCycle(ss, response)
bsts.model1 <- bsts(
response ~ PRCP + TMAX + Holidays + IsWeekend,
state.specification = ss, niter = 2000, data = dat_Each)
#plot(bsts.model1, "coef")
# Estiamting counterfactual and compare with actual post period response
impact <- CausalImpact(
bsts.model = bsts.model1,
post.period.response = post.period.response)
#summary(impact)
# plot(impact)
#summary(impact, "report")
components.withreg <- as.data.frame(impact$summary)
components.withreg$CTNAME <- eachstation
components.withreg
#write.csv(components.withreg, 'finalMatrix_Transit_temp.csv')
}
stopCluster(cl)
write.csv(finalMatrix, 'finalImpact_Transit_0810_old.csv')