.Rhistory

#work with the number of sales from dataset
adj <- select(Sale_of_fish, -c("year", "month", "end.of.period", "value..E.", "unsoldes..Kg."))
install.packages("dplyr")
select
select
library(dplyr)
#import dataset
Sale_of_fish <- read.csv("Fish dataset.csv",header=T) #data source=FranceAgriMer/VISIOMer
#import dataset
Sale_of_fish <- read.csv("Fish dataset.csv",header=T) #data source=FranceAgriMer/VISIOMer
Fish
Fish
Fish
adj$sales <- as.numeric(gsub(' ', '', adj$sales))
#Analysis of Fish sales dataset
#Import important library for time series analysis in R
library(tidyverse)
library(fpp)
library(forecast)
library(backtest)
library(quantmod)
library(lubridate)
library(dplyr)
#import dataset
Sale_of_fish <- read.csv("Fish dataset.csv",header=T) #data source=FranceAgriMer/VISIOMer
head(Sale_of_fish) #See head of the dataset
tail(Sale_of_fish) #see tail of the dataset
#work with the number of sales from dataset
adj <- select(Sale_of_fish, -c("year", "month", "end.of.period", "value..E.", "unsoldes..Kg."))
adj$sales <- as.numeric(gsub(' ', '', adj$sales))
adj$start <- dmy(adj$start)##This tells you that the data series is in a time series format
adj$logr <- log(lag(adj$sales)) - log(adj$sales)
head(adj)
narm <- function (x) {
x[is.na(x)] <- 1
return(x)
}
adj$logr <- narm(adj$logr)
head(adj)
setwd("C:/Users/miche/OneDrive - HKUST Connect/Data Science/Data-Science-Private-Consulting/FR-P20201130-Time_Series_Forecast")
#Analysis of Fish sales dataset
#Import important library for time series analysis in R
library(tidyverse)
library(fpp)
library(forecast)
library(backtest)
library(quantmod)
library(lubridate)
library(dplyr)
#import dataset
Sale_of_fish <- read.csv("Fish dataset.csv",header=T) #data source=FranceAgriMer/VISIOMer
head(Sale_of_fish) #See head of the dataset
tail(Sale_of_fish) #see tail of the dataset
#work with the number of sales from dataset
adj <- select(Sale_of_fish, -c("year", "month", "end.of.period", "value..E.", "unsoldes..Kg."))
adj$sales <- as.numeric(gsub(' ', '', adj$sales))
adj$start <- dmy(adj$start)##This tells you that the data series is in a time series format
adj$logr <- log(lag(adj$sales)) - log(adj$sales)
head(adj)
narm <- function (x) {
x[is.na(x)] <- 1
return(x)
}
adj$logr <- narm(adj$logr)
head(adj)
max_Date <- max(adj$start)
min_Date <- min(adj$start)
test_ts <- ts(adj$logr, end=c(year(max_Date), month(max_Date)),start=c(year(min_Date), month(min_Date)),frequency=12)#freq 12 => Monthly data
logr <- adj
logr
dev.off() #to face margin problem error
dev.off() #to face margin problem error
plot(test_ts)
print(adf.test(test_ts)) # p-value < 0.05 indicates the TS is stationary
summary(m1)
summary(m1)
acf = acf(test_ts, main='ACF Plot', lag.max=100) # autocorrelation
pacf.logr = pacf(test_ts, main='PACF Plot', lag.max=100) # partial autocorrelation
accuracy(forecast(m1))
m1 <- auto.arima(test_ts, seasonal = TRUE)
summary(m1)
accuracy(forecast(m1))
source("C:/Users/miche/OneDrive - HKUST Connect/Data Science/Data-Science-Private-Consulting/FR-P20201130-Time_Series_Forecast/ts.R")
#Analysis for Yahoo finance data
#Import important library
library(quantmod)
library(fpp)
library(backtest)
#read yahoo finance datasets for last one year CA.PA
data <- getSymbols("CA.PA", from="2019-11-25",to = "2020-11-24" ,src="yahoo", auto.assign=F)
head(data)#see the head of dataset
tail(data) #see the tail of dataset
adj <- data[,5]
adj
dev.off() #to face margin problem error
plot(data[,6], main = "Adhusted Closing Price")
log <- periodReturn(adj,period = "daily", type = "log", leading = TRUE)
log <- 1+log
head(logr)
#differentiate
diff(log)
par(mfrow = c(2,1))
# both acf() and pacf() generates plots by default
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
#Estimate the coefficients Using Seasonal Arima model
m1 <- auto.arima(logr, seasonal = FALSE)
summary(m1)
#studying the residues
checkresiduals(m1) # p-value over .05 confirms no autocorrelations
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
summary(m1)
#Analysis for Yahoo finance data
#Import important library
library(quantmod)
library(fpp)
library(backtest)
#read yahoo finance datasets for last one year CA.PA
data <- getSymbols("CA.PA", from="2019-11-25",to = "2020-11-24" ,src="yahoo", auto.assign=F)
head(data)#see the head of dataset
tail(data) #see the tail of dataset
adj <- data[,5]
adj
dev.off() #to face margin problem error
plot(data[,6], main = "Adhusted Closing Price")
log <- periodReturn(adj,period = "daily", type = "log", leading = TRUE)
log <- 1+log
head(logr)
#differentiate
diff(log)
par(mfrow = c(2,1))
# both acf() and pacf() generates plots by default
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
#Estimate the coefficients Using Seasonal Arima model
m1 <- auto.arima(logr, seasonal = FALSE)
summary(m1)
#studying the residues
checkresiduals(m1) # p-value over .05 confirms no autocorrelations
#Analysis of Fish sales dataset
#Import important library for time series analysis in R
library(tidyverse)
library(fpp)
library(forecast)
library(backtest)
library(quantmod)
library(lubridate)
library(dplyr)
#import dataset
Sale_of_fish <- read.csv("Fish dataset.csv",header=T) #data source=FranceAgriMer/VISIOMer
head(Sale_of_fish) #See head of the dataset
tail(Sale_of_fish) #see tail of the dataset
#work with the number of sales from dataset
adj <- select(Sale_of_fish, -c("year", "month", "end.of.period", "value..E.", "unsoldes..Kg."))
adj$sales <- as.numeric(gsub(' ', '', adj$sales))
adj$start <- dmy(adj$start)##This tells you that the data series is in a time series format
adj$logr <- log(lag(adj$sales)) - log(adj$sales)
head(adj)
narm <- function (x) {
x[is.na(x)] <- 1
return(x)
}
adj$logr <- narm(adj$logr)
head(adj)
max_Date <- max(adj$start)
min_Date <- min(adj$start)
test_ts <- ts(adj$logr, end=c(year(max_Date), month(max_Date)),start=c(year(min_Date), month(min_Date)),frequency=12)#freq 12 => Monthly data
logr <- adj
logr
dev.off() #to face margin problem error
plot(test_ts)
plot(stl(test_ts,s.window = "periodic"))
# both acf() and pacf() generates plots by default
acf = acf(test_ts, main='ACF Plot', lag.max=100) # autocorrelation
pacf.logr = pacf(test_ts, main='PACF Plot', lag.max=100) # partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(test_ts)) # p-value < 0.05 indicates the TS is stationary
#Estimate the coefficients Using Seasonal Arima model
m1 <- auto.arima(test_ts, seasonal = TRUE)
summary(m1)
#Check Accuracy
accuracy(forecast(m1))
#studying the residues
checkresiduals(m1) # p-value over .05 confirms no autocorrelations
#Analysis for Yahoo finance data
#Import important library
library(quantmod)
library(fpp)
library(backtest)
#read yahoo finance datasets for last one year CA.PA
data <- getSymbols("CA.PA", from="2019-11-25",to = "2020-11-24" ,src="yahoo", auto.assign=F)
head(data)#see the head of dataset
tail(data) #see the tail of dataset
adj <- data[,5]
adj
dev.off() #to face margin problem error
plot(data[,6], main = "Adhusted Closing Price")
log <- periodReturn(adj,period = "daily", type = "log", leading = TRUE)
log <- 1+log
head(logr)
#differentiate
diff(log)
par(mfrow = c(2,1))
# both acf() and pacf() generates plots by default
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
#Estimate the coefficients Using Seasonal Arima model
m1 <- auto.arima(logr, seasonal = FALSE)
summary(m1)
#studying the residues
checkresiduals(m1) # p-value over .05 confirms no autocorrelations
View(acf.logr)
View(adj)
View(log)
View(narm)
Analysis for Yahoo finance data
#Import important library
library(quantmod)
library(fpp)
library(backtest)
#read yahoo finance datasets for last one year CA.PA
data <- getSymbols("CA.PA", from="2019-11-25",to = "2020-11-24" ,src="yahoo", auto.assign=F)
head(data)#see the head of dataset
tail(data) #see the tail of dataset
adj <- data[,5]
adj
dev.off() #to face margin problem error
plot(data[,6], main = "Adhusted Closing Price")
log <- periodReturn(adj,period = "daily", type = "log", leading = TRUE)
log <- 1+log
head(logr)
head(log)
diff(log)
par(mfrow = c(2,1))
acf.log = acf(log, main='ACF Plot', lag.max=100)# autocorrelation
adj <- data[,5]
adj
adj$logr <- log(lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
adj$logr <- log(lag(adj$Volume)) - log(adj$Volume)
adj[1]
adj$logr <- log(lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
log(lag(adj$CA.PA.Volume))
log(adj$CA.PA.Volume)
log(lag(adj$CA.PA.Volume))
adj$CA.PA.Volume
lag(adj$CA.PA.Volume)
adj$logr <- log(stats::lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
head(adj)
narm <- function (x) {
x[is.na(x)] <- 1
return(x)
}
both acf() and pacf() generates plots by default
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
m1 <- auto.arima(logr, seasonal = FALSE)
summary(m1)
logr <- adj
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
logr
head(adj)
narm <- function (x) {
x[is.na(x)] <- 1
return(x)
}
adj
logr <- adj
logr
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
logr
adj$logr <- log(stats::lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
head(adj)
narm <- function (x) {
x[is.na(x)] <- 1
return(x)
}
logr <- adj
logr.head()
head(logr)
adj$logr <- log(stats::lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
adj$logr <- narm(adj$logr)
logr <- adj
head(logr)
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
adj$logr
adj$logr <- log(stats::lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
head(adj)
narf <- function (x) {
x[is.infinite(x)] <- 1
return(x)
}
adj$logr <- narf(adj$logr)
logr <- adj
head(logr)
# both acf() and pacf() generates plots by default
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(logr)) # p-v
logr
adj$logr <- marm(narf(adj$logr))
adj$logr <- narm(narf(adj$logr))
logr <- adj
head(logr)
# both acf() and pacf() generates plots by default
acf.logr = acf(logr, main='ACF Plot', lag.max=100)# autocorrelation
pacf.logr = pacf(logr, main='PACF Plot', lag.max=100)# partial autocorrelation
#Augmented Dickey-Fuller(ADF) Test
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
#Estimate the coefficients Using Seasonal Arima model
m1 <- auto.arima(logr, seasonal = FALSE)
summary(m1)
logr
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
log <- periodReturn(adj,period = "daily", type = "log", leading = TRUE)
log <- 1+log
head(log)
adj$logr <- log(stats::lag(adj$CA.PA.Volume)) - log(adj$CA.PA.Volume)
diff(log)
par(mfrow = c(2,1))
narf <- function (x) {
x[is.infinite(x)] <- 1
return(x)
}
adj$logr <- narm(narf(adj$logr))
adj
logr <- adj
head(logr)
adj
logr <- adj
head(logr)
acf = acf(test_ts, main='ACF Plot', lag.max=100) # autocorrelation
pacf.logr = pacf(test_ts, main='PACF Plot', lag.max=100) # partial au
test_ts <- ts(adj$logr, end=c(year(max_Date), month(max_Date)),start=c(year(min_Date), month(min_Date)),frequency=12)#freq 12 => Monthly data
acf = acf(test_ts, main='ACF Plot', lag.max=100) # autocorrelation
pacf.logr = pacf(test_ts, main='PACF Plot', lag.max=100) # partial autocorrelation
# both acf() and pacf() generates plots by default
acf = acf(test_ts, main='ACF Plot', lag.max=100) # autocorrelation
pacf.logr = pacf(test_ts, main='PACF Plot', lag.max=100) # partial autocorrelation
print(adf.test(logr)) # p-value < 0.05 indicates the TS is stationary
print(adf.test(test_ts)) # p-value < 0.05 indicates the TS is stationary
test_ts <- ts(adj$logr, end=c(year(max_Date), month(max_Date)),start=c(year(min_Date), month(min_Date)),frequency=12)#freq 12 => Monthly data
print(adf.test(test_ts)) # p-value < 0.05 indicates the TS is stationary
m1 <- auto.arima(logr, seasonal = TRUE)
m1 <- auto.arima(test_ts, seasonal = TRUE)
summary(m1)
checkresiduals(m1) # p-value over .05 confirms no autocorrelations
summary(m1)