# Step 1: Load R Packages
### options(repos='http://cran.rstudio.com/')
#install.packages("astsa")
#install.packages('ggplot2')
#install.packages('forecast')
#install.packages('tseries')
#install.packages("data.table")
library(astsa)
library(forecast)
library(tseries)
library(zoo)
library(tseries)
library(data.table)
library(dplyr)
library(tidyr)
library(naniar)
library(stringr)
library(ggplot2)
library(DT)
library(lubridate)
library(ggpubr)
setwd("/scratch/user/cma16/Task4_Deliverable2/NCprocess4/AllCrash/FacilityBased/")
load("./multi-lane_undivided_NC_reduce_withCrash_no_intersection.rData")
mytype = 'RMU'
setwd(paste0("/scratch/user/cma16/Task4_Deliverable2/NCprocess4/AllCrash/FacilityBased/",mytype))
df_RMU <- N_mun_nomed
dim(df_RMU)
## [1] 3995136 30
### Calculating Speed
df_RMU$spd_av = 3600*df_RMU$TMC_length/df_RMU$Travel_TIME_ALL_VEHICLES/5280
df_RMU$spd_pv = 3600*df_RMU$TMC_length/df_RMU$Travel_TIME_PASSENGER_VEHICLES/5280
df_RMU$spd_ft = 3600*df_RMU$TMC_length/df_RMU$Travel_TIME_FREIGHT_TRUCKS/5280
### Month, Day
df_RMU$date <- as.character(df_RMU$DATE)
df_RMU$date <- str_pad(df_RMU$DATE, 8, pad = "0")
df_RMU$Month <- substr(df_RMU$date, start = 1, stop = 2)
df_RMU$Day <- substr(df_RMU$date, start = 3, stop = 4)
df_RMU$Year <- substr(df_RMU$date, start = 5, stop = 8)
df_RMU$MonthDay <- paste0(df_RMU$Month,"_", df_RMU$Day)
head(df_RMU)
## TimeStamp TMC DATE EPOCH15 Travel_TIME_ALL_VEHICLES
## 1: 110N17806_0701_0 110N17806 7012015 0 NA
## 2: 110N17806_0701_1 110N17806 7012015 1 NA
## 3: 110N17806_0701_10 110N17806 7012015 10 NA
## 4: 110N17806_0701_11 110N17806 7012015 11 NA
## 5: 110N17806_0701_12 110N17806 7012015 12 NA
## 6: 110N17806_0701_13 110N17806 7012015 13 NA
## Travel_TIME_PASSENGER_VEHICLES Travel_TIME_FREIGHT_TRUCKS TMC_length
## 1: NA NA 24722.67
## 2: NA NA 24722.67
## 3: NA NA 24722.67
## 4: NA NA 24722.67
## 5: NA NA 24722.67
## 6: NA NA 24722.67
## ave_aadt ave_wtdsgspd ave_medwid ave_peaklane ave_row ave_sur_wid
## 1: 5249.505 65 NA NA 71.37837 23.10593
## 2: 5249.505 65 NA NA 71.37837 23.10593
## 3: 5249.505 65 NA NA 71.37837 23.10593
## 4: 5249.505 65 NA NA 71.37837 23.10593
## 5: 5249.505 65 NA NA 71.37837 23.10593
## 6: 5249.505 65 NA NA 71.37837 23.10593
## ave_no_lanes ave_spd_limt ave_rodwycls ave_rshldwid FC TER ACC MED
## 1: 2.098732 40.52784 8.098732 6 6 2 F Cu
## 2: 2.098732 40.52784 8.098732 6 6 2 F Cu
## 3: 2.098732 40.52784 8.098732 6 6 2 F Cu
## 4: 2.098732 40.52784 8.098732 6 6 2 F Cu
## 5: 2.098732 40.52784 8.098732 6 6 2 F Cu
## 6: 2.098732 40.52784 8.098732 6 6 2 F Cu
## Total K A B C O DAYMTH Crash spd_av spd_pv spd_ft date Month Day
## 1: 0 0 0 0 0 0 0701 0 NA NA NA 07012015 07 01
## 2: 0 0 0 0 0 0 0701 0 NA NA NA 07012015 07 01
## 3: 0 0 0 0 0 0 0701 0 NA NA NA 07012015 07 01
## 4: 0 0 0 0 0 0 0701 0 NA NA NA 07012015 07 01
## 5: 0 0 0 0 0 0 0701 0 NA NA NA 07012015 07 01
## 6: 0 0 0 0 0 0 0701 0 NA NA NA 07012015 07 01
## Year MonthDay
## 1: 2015 07_01
## 2: 2015 07_01
## 3: 2015 07_01
## 4: 2015 07_01
## 5: 2015 07_01
## 6: 2015 07_01
day1<- df_RMU[,-c(1)] %>% group_by(MonthDay) %>% summarize(Speed_All_Mean=mean(spd_av, na.rm=TRUE))
day1
## # A tibble: 365 x 2
## MonthDay Speed_All_Mean
## <chr> <dbl>
## 1 01_01 33.8
## 2 01_02 33.7
## 3 01_03 33.0
## 4 01_04 34.1
## 5 01_05 34.7
## 6 01_06 34.2
## 7 01_07 34.3
## 8 01_08 34.7
## 9 01_09 33.8
## 10 01_10 34.5
## # ... with 355 more rows
# Step 2: Examine Data
speed_clean <- tsclean(ts(day1$Speed_All_Mean))
plot.ts(speed_clean)
# ggplot() + geom_line(data = Q1, aes(x = TimeStamp, y = speed_clean)) + ylab('Cleaned Speed Records')
day1$cnt_ma = ma(speed_clean, order=7) # using the clean count with no outliers
day1$cnt_ma30 = ma(speed_clean, order=30)
# Step 3: Decompose Your Data
count_ma = ts(na.omit(speed_clean), frequency=30)
decomp = stl(count_ma, s.window="periodic")
deseasonal_cnt <- seasadj(decomp)
plot(decomp)
# Step 4: Stationarity
# statinary test
adf.test(count_ma, alternative = "stationary")
##
## Augmented Dickey-Fuller Test
##
## data: count_ma
## Dickey-Fuller = -3.7146, Lag order = 7, p-value = 0.02354
## alternative hypothesis: stationary
adf.test(deseasonal_cnt, alternative = "stationary")
##
## Augmented Dickey-Fuller Test
##
## data: deseasonal_cnt
## Dickey-Fuller = -3.2578, Lag order = 7, p-value = 0.07838
## alternative hypothesis: stationary
d1 = diff(deseasonal_cnt)
adf.test(d1, alternative = "stationary")
## Warning in adf.test(d1, alternative = "stationary"): p-value smaller than
## printed p-value
##
## Augmented Dickey-Fuller Test
##
## data: d1
## Dickey-Fuller = -11.685, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
# Step 5: Autocorrelations and Choosing Model Order
# check ACF and PACF
acf2(count_ma)
## ACF PACF
## [1,] 0.45 0.45
## [2,] 0.34 0.16
## [3,] 0.32 0.15
## [4,] 0.33 0.15
## [5,] 0.23 0.00
## [6,] 0.39 0.28
## [7,] 0.49 0.29
## [8,] 0.38 0.04
## [9,] 0.30 0.00
## [10,] 0.30 0.00
## [11,] 0.26 0.00
## [12,] 0.21 -0.02
## [13,] 0.28 0.01
## [14,] 0.33 0.04
## [15,] 0.27 -0.02
## [16,] 0.21 -0.06
## [17,] 0.24 0.03
## [18,] 0.19 -0.03
## [19,] 0.16 -0.01
## [20,] 0.18 -0.02
## [21,] 0.35 0.21
## [22,] 0.25 0.01
## [23,] 0.17 -0.07
## [24,] 0.21 0.05
## [25,] 0.17 -0.03
## [26,] 0.11 0.00
## [27,] 0.18 0.02
## [28,] 0.28 0.05
## [29,] 0.22 0.01
## [30,] 0.19 0.01
## [31,] 0.20 0.01
## [32,] 0.20 0.06
## [33,] 0.15 0.02
## [34,] 0.18 -0.02
## [35,] 0.28 0.09
## [36,] 0.16 -0.12
## [37,] 0.15 0.00
## [38,] 0.18 -0.02
## [39,] 0.21 0.05
## [40,] 0.17 0.05
## [41,] 0.18 -0.02
## [42,] 0.25 0.05
## [43,] 0.14 -0.10
## [44,] 0.10 -0.03
## [45,] 0.12 -0.04
## [46,] 0.15 0.01
## [47,] 0.16 0.09
## [48,] 0.14 -0.06
## [49,] 0.23 0.08
## [50,] 0.14 -0.03
## [51,] 0.03 -0.13
## [52,] 0.10 0.06
## [53,] 0.11 -0.02
## [54,] 0.11 0.03
## [55,] 0.13 0.05
## [56,] 0.25 0.07
## [57,] 0.07 -0.10
## [58,] 0.00 -0.07
## [59,] 0.04 -0.02
## [60,] 0.07 0.01
## [61,] 0.03 -0.05
## [62,] 0.09 -0.03
## [63,] 0.19 0.06
## [64,] 0.06 -0.05
## [65,] 0.00 0.01
## [66,] 0.02 -0.02
## [67,] 0.05 0.04
## [68,] 0.04 0.05
## [69,] 0.06 -0.06
## [70,] 0.16 0.08
## [71,] 0.07 -0.05
## [72,] 0.02 0.04
## [73,] 0.01 -0.02
## [74,] 0.05 0.00
## [75,] 0.02 0.03
## [76,] 0.07 -0.01
## [77,] 0.13 -0.04
## [78,] 0.07 0.02
## [79,] 0.04 0.04
## [80,] 0.06 0.02
## [81,] 0.01 -0.08
## [82,] -0.01 -0.04
## [83,] 0.04 0.04
## [84,] 0.16 0.05
## [85,] 0.07 0.00
## [86,] 0.01 -0.06
## [87,] 0.02 -0.05
## [88,] -0.01 -0.02
## [89,] 0.02 0.06
## [90,] 0.03 0.02
## [91,] 0.13 0.01
## [92,] 0.03 -0.04
## [93,] 0.00 0.01
## [94,] -0.02 -0.04
## [95,] 0.02 0.03
## [96,] 0.02 0.04
## [97,] 0.07 0.06
## [98,] 0.07 -0.03
## [99,] 0.04 0.03
## [100,] 0.06 0.11
## [101,] 0.02 -0.06
## [102,] 0.02 -0.03
## [103,] 0.03 0.01
## [104,] 0.05 0.02
## [105,] 0.09 0.00
## [106,] 0.06 -0.02
## [107,] -0.03 -0.12
## [108,] 0.01 0.02
## [109,] 0.03 0.01
## [110,] 0.05 0.03
## [111,] 0.03 0.00
## [112,] 0.14 0.03
## [113,] -0.01 -0.08
## [114,] -0.06 -0.06
## [115,] -0.04 0.02
## [116,] -0.01 -0.01
## [117,] -0.04 -0.06
## [118,] 0.05 0.08
## [119,] 0.08 -0.04
## [120,] -0.04 -0.03
## ACF PACF
## [1,] 0.47 0.47
## [2,] 0.36 0.18
## [3,] 0.34 0.15
## [4,] 0.35 0.16
## [5,] 0.27 0.01
## [6,] 0.42 0.28
## [7,] 0.52 0.30
## [8,] 0.42 0.06
## [9,] 0.32 -0.02
## [10,] 0.29 -0.03
## [11,] 0.28 0.00
## [12,] 0.24 -0.02
## [13,] 0.30 0.01
## [14,] 0.37 0.05
## [15,] 0.31 -0.02
## [16,] 0.23 -0.05
## [17,] 0.26 0.05
## [18,] 0.21 -0.03
## [19,] 0.17 -0.05
## [20,] 0.18 -0.06
## [21,] 0.37 0.22
## [22,] 0.27 0.01
## [23,] 0.18 -0.08
## [24,] 0.23 0.06
## [25,] 0.19 -0.03
## [26,] 0.12 0.00
## [27,] 0.19 0.05
## [28,] 0.30 0.06
## [29,] 0.23 -0.02
## [30,] 0.15 -0.07
## [31,] 0.20 0.05
## [32,] 0.21 0.09
## [33,] 0.15 0.01
## [34,] 0.19 0.00
## [35,] 0.31 0.10
## [36,] 0.17 -0.11
## [37,] 0.16 0.03
## [38,] 0.20 0.01
## [39,] 0.22 0.03
## [40,] 0.16 0.01
## [41,] 0.19 0.00
## [42,] 0.27 0.06
## [43,] 0.15 -0.13
## [44,] 0.11 -0.02
## [45,] 0.15 -0.04
## [46,] 0.17 0.00
## [47,] 0.17 0.11
## [48,] 0.16 -0.06
## [49,] 0.23 0.03
## [50,] 0.14 -0.05
## [51,] 0.04 -0.09
## [52,] 0.12 0.09
## [53,] 0.13 -0.04
## [54,] 0.12 0.03
## [55,] 0.14 0.04
## [56,] 0.26 0.05
## [57,] 0.07 -0.08
## [58,] 0.01 -0.07
## [59,] 0.04 -0.05
## [60,] 0.05 -0.06
## [61,] 0.03 -0.02
## [62,] 0.10 0.02
## [63,] 0.20 0.04
## [64,] 0.07 -0.03
## [65,] 0.01 0.03
## [66,] 0.02 -0.01
## [67,] 0.05 0.07
## [68,] 0.06 0.07
## [69,] 0.06 -0.08
## [70,] 0.15 0.04
## [71,] 0.08 -0.03
## [72,] 0.02 0.06
## [73,] 0.01 -0.05
## [74,] 0.06 -0.02
## [75,] 0.04 0.04
## [76,] 0.08 -0.01
## [77,] 0.14 -0.01
## [78,] 0.09 0.02
## [79,] 0.04 0.01
## [80,] 0.04 -0.01
## [81,] 0.01 -0.07
## [82,] 0.00 -0.01
## [83,] 0.04 0.02
## [84,] 0.17 0.05
## [85,] 0.08 0.00
## [86,] 0.00 -0.06
## [87,] 0.02 -0.02
## [88,] 0.00 -0.01
## [89,] 0.02 0.03
## [90,] 0.00 -0.03
## [91,] 0.13 0.03
## [92,] 0.03 -0.02
## [93,] 0.00 0.01
## [94,] -0.01 -0.02
## [95,] 0.04 0.06
## [96,] 0.02 0.05
## [97,] 0.07 0.08
## [98,] 0.09 -0.01
## [99,] 0.05 0.03
## [100,] 0.05 0.09
## [101,] 0.02 -0.08
## [102,] 0.03 -0.02
## [103,] 0.03 -0.02
## [104,] 0.06 0.03
## [105,] 0.11 0.01
## [106,] 0.07 -0.03
## [107,] -0.03 -0.10
## [108,] 0.02 -0.01
## [109,] 0.02 0.00
## [110,] 0.04 0.03
## [111,] 0.02 -0.02
## [112,] 0.14 0.04
## [113,] 0.00 -0.10
## [114,] -0.06 -0.06
## [115,] -0.03 0.04
## [116,] 0.00 -0.01
## [117,] -0.03 -0.06
## [118,] 0.07 0.09
## [119,] 0.08 -0.06
## [120,] -0.06 -0.09
Seasonility Not in Consideration
# Step 6: Fitting an ARIMA model
auto.arima(deseasonal_cnt, seasonal=FALSE)
## Series: deseasonal_cnt
## ARIMA(1,1,1)
##
## Coefficients:
## ar1 ma1
## 0.1402 -0.9258
## s.e. 0.0566 0.0204
##
## sigma^2 estimated as 0.2174: log likelihood=-238.63
## AIC=483.25 AICc=483.32 BIC=494.94
# Step 7: Evaluate and Iterate
# (try different model)
fit<-auto.arima(deseasonal_cnt, seasonal=FALSE)
tsdisplay(residuals(fit), lag.max=45, main='Model Residuals [Seasonality not considered]')
# step 8 forcast
fcast <- forecast(fit, h=30)
plot(fcast)
Seasonility in Consideration
# Step 6: Fitting an ARIMA model
auto.arima(deseasonal_cnt, seasonal=TRUE)
## Series: deseasonal_cnt
## ARIMA(0,1,1)(2,0,0)[30] with drift
##
## Coefficients:
## ma1 sar1 sar2 drift
## -0.9227 -0.0910 -0.0804 -0.0037
## s.e. 0.0246 0.0566 0.0607 0.0017
##
## sigma^2 estimated as 0.2176: log likelihood=-238.19
## AIC=486.38 AICc=486.55 BIC=505.86
# Step 7: Evaluate and Iterate
# (try different model)
fit<-auto.arima(deseasonal_cnt, seasonal=TRUE)
tsdisplay(residuals(fit), lag.max=45, main='Model Residuals [Seasonality considered]')
# step 8 forcast
fcast <- forecast(fit, h=30)
plot(fcast)
