top of page
Project | 03

Time Series Analysis on COVID-19 and Airline Industry

이미지 제공: Adam Nieścioruk
Business Question
 How the COVID-19 outbreak affected airline industries on many grounds?
​
Dataset
 Several datasets have been handled. S&P 500 stock data was retrieved from Yahoo Finance using Shiny App which scrapes all the stock prices with tickers and generates a dataframe. Hence, 'industrial ticker' datasets consists of 'symbol' of each company's ticker and corresponding close stock price for a certain period of time, 2019-12-05 through 2020-05-04.
data1.PNG
'mobility' dataset consists of 11 columns, but we have used only the dates and mean mobility values(represented by percent change in mobility in each sector) to compare with the stock prices joining by the dates.
data2.PNG
Just like S&P 500 Stock tickers dataset, airline companies stock prices datasets were handled, and there are more detailed prices such as adjusted return prices and volume other than open and close prices.
data3.PNG
'airfreight' dataset was retrieved from a Yahoo finance() ticker of Fedex, UPS, C.H. Robinson, Expeditors International of Washington, with S&P 500.
data4.PNG
Data Processing
 We used web scraping technique to retrieve the data as well as using Shiny App (http://shiny.stat.ucla.edu:3838/c183c283/) . The raw data retrieved from all over the web was at first very messy so we needed to handle with missing values, and date time components were not in a correct form so we needed to work on feature engineering. It was very important as our primary methodology was time series analysis on the stock price.
​
Exploratory Data Analysis

 Figure 1 demonstrates the number of US confirmed cases against the date in 2020. We log-transformed it to make the plot more come in better sense. Figure 2 shows the box plots of stock price return on different industries. We see that airline industry is the lowest among five sectors and biotechnology is the highest one.

image01.png
image02.png
 Then, we started to take a look with its stock data. We retrieved the daily stock price data of S&P 500 companies from Yahoo Finance and calculated how much return of investment each industry had on average. As we can see, the airline industries had a huge negative return of investment, while having a small variance compared to other industries for which we suspected to be correlated with physical distancing and reduced mobility. On the graph on the left hand side, we plotted how the mobility and stock prices had been changing in the last three months, and we figured that they had a very similar trend.
image03.PNG
Model Approach : Time Series Analysis on airline stock price, social distancing, and air-freight data
 Most of the airline companies are acting in the same way, so we picked United Airline which has the highest correlation with COVID impact. We predicted its stock price and revenue and observed unnatural depreciation as soon as it became 2020. Every prediction has been made using "Exponential Smoothing" technique. (Refer to the code below)
image04.PNG
 Google trend data revealed international travel has decreased for 70% and domestic flight has decreased for average 40%, obviously due to the social distancing and rapid worldwide spread of COVID-19. After investigating COVID impact on airline companies and LAX, we sought for alternative strategies on how airline companies make profit.
image05.PNG
 To clarify, we simplified the total passenger count of the left hand side. Middle plot illustrates the scaled count of cargo transported and aircraft operations in total. Due to the seasonality, cargo has its lowest points in February, but it bounced back to it's usual routine in March. From the effect of social distancing and decreased industrial activities, cargo also somewhat decreased compared to previous years. But for the aircraft operations, counts have drastically decreased ever since the pandemic arises. This made air-freight fees to be more expensive than usual. Oil prices are currently at their lowest point, so airline companies are using passenger aircraft as a cargo carrier to make some profit but this is still not enough for them.
image06.PNG
Conclusion
 Our analysis suggests that the biggest collapse in terms of stock return and revenue was airline industry 1 with showing the large gap between its actual and predicted values. The high correlation between workplace mobility and stock price of airline industries has led the further analysis of the airport passenger count and social distancing. A strong negative correlation between the airport passenger count and social distancing value at its peak point 1 proposes that travel count as explained by the airport passenger count 1 increases as the social distancing value shows its decreasing trend after its peak. The rise in the demand in cargo with the reduction in the flight operation and its passenger count implies the airline companies are replacing the passengers with the air-freights. 
​
​
Presentation Video
References
yfinance: https://pypi.org/project/yfinance/
covid19py: https://pypi.org/project/COVID19Py/
Google trend: https://trends.google.com/trends/?geo=US
Cargo & Aircraft Operations: https://www.lawa.org
Stock Data: http://shiny.stat.ucla.edu:3838/c183c283/
Jet Fuel Price: https://www.iata.org/en/publications/economics/fuel-monitor/
10-Q: https://www.sec.gov
Mobility: https://www.google.com/covid19/mobility/
COVID-19 data:https://coronavirus.jhu.edu/map.html
Figure2 shows side-by-side barplots for every categorical variable against Diabetic. Considering we have about 70% of factor "no" and 30% of factor "yes" for Diabetic in total, we can grasp a general idea of data from these plots.

---

title: "COVID datafest"

author: "Park"

date: "5/4/2020"

output: pdf_document

---

```{r}

# for scraping

library(rvest)

# blanket import for core tidyverse packages

library(tidyverse)

# tidy financial analysis

library(tidyquant)

# tidy data cleaning functions

library(janitor)

library(ggplot2)

library(readxl)

library(dplyr)

library(readxl)

library(tseries)

```

```{r}

# save current system date to a variable

today <- Sys.Date()

# subtract 3 months from the current date

date = today %m+% months(-5)

print(date)

# pass SP500 ticker ^GSPC to tq_get function

one_ticker = tq_get("^GSPC", from = date)

one_ticker %>% head()

# save mobility data

mobility <- read.csv("https://www.gstatic.com/covid19/mobility/Global_Mobility_Report.csv?cachebust=a88b56a24e1a1e25", stringsAsFactors=FALSE, fileEncoding="latin1")

#write.csv(mobility,"mobility.csv")

co <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/CO_ca.csv", header = TRUE)

co$Date <- as.Date(co$Date, "%m/%d/%Y")

ozone <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/Ozone_ca.csv", header = TRUE)

ozone2 <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/ozone2019.csv", header = TRUE)

ozone3 <- rbind(ozone2, ozone)

ozone3$Date <- as.Date(ozone3$Date, "%m/%d/%Y")

pm25 <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/PM25_ca.csv", header = TRUE)

pm25_19 <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/pm25_2019.csv", header = TRUE)

pm25_total <- rbind(pm25_19, pm25)

pm25_total$Date <- as.Date(pm25_total$Date, "%m/%d/%Y")

# get Quarterly Report

air_revenue_quarterly <- read_excel("~/Dropbox/School/UCLA/Datafest 2020/Airline data/airline_revenue_quarterly.xlsx")

air_revenue_quarterly$Year <- as.Date(air_revenue_quarterly$Year, "%m/%d/%Y")

air_revenue_quarterly <- air_revenue_quarterly[,c(1:3)]

air_revenue_quarterly <- air_revenue_quarterly %>% group_by(Symbol) %>% mutate(scaled <- scale(Revenue))

colnames(air_revenue_quarterly)[4] <- "scaled"

t_air <- ts(air_revenue_quarterly %>% filter(Symbol == "DAL"), frequency = 4, start = c(2015,1))

# Mobility

mobility <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/mobility.csv")

us_mobility <- mobility %>% filter(country_region_code == "US")

us_mobility$date <- as.Date(us_mobility$date, "%Y-%m-%d")

us_mobility$retail_and_recreation_percent_change_from_baseline <- as.numeric(us_mobility$retail_and_recreation_percent_change_from_baseline)

us_mobility$sub_region_1 <- as.character(us_mobility$sub_region_1)

us_mobility$sub_region_2 <- as.character(us_mobility$sub_region_2)

us_mobility <- us_mobility %>% mutate_if(is.numeric, ~replace(., is.na(.), 0))

summary(us_mobility)

state_mobility <- us_mobility %>%

group_by(date) %>%

mutate(work = mean(workplaces_percent_change_from_baseline),

retail = mean(retail_and_recreation_percent_change_from_baseline),

transit = mean(transit_stations_percent_change_from_baseline)) %>%

select(c(date,work,retail,transit))

state_mobility <- state_mobility[1:76,]

#write.csv(state_mobility, "state_mobility.csv")

```

```{r, echo=FALSE}

#Load necessary packages.

x <- c("httr", "jsonlite", "tidyverse", "dplyr","tidyr", "stringr", "lubridate","data.table", "readr","tibble", "ggplot2", "scales","RCurl")

suppressPackageStartupMessages(lapply(x, library, character.only = TRUE))

```

```{r}

# get the URL for the wikipedia page with all SP500 symbols

url <- "https://en.wikipedia.org/wiki/List_of_S%26P_500_companies"

# use that URL to scrape the SP500 table using rvest

tickers <- url %>%

# read the HTML from the webpage

read_html() %>%

# one way to get table

#html_nodes(xpath='//*[@id="mw-content-text"]/div/table[1]') %>%

# easier way to get table

html_nodes(xpath = '//*[@id="constituents"]') %>%

html_table()

#create a vector of tickers

sp500tickers <- tickers[[1]]

sp500tickers = sp500tickers %>% mutate(Symbol = case_when(Symbol == "BRK.B" ~ "BRK-B",

Symbol == "BF.B" ~ "BF-B",

TRUE ~ as.character(Symbol)))

#html_nodes(xpath = '//*[@id="constituents"]'

sectors <- sp500tickers$`GICS Sector` %>% unique()

Industrials_sector <- sp500tickers %>% filter(`GICS Sector` == sectors[1]) %>% select("Symbol")

# get monthly stock price

monthly_air <- read_csv("~/Dropbox/School/UCLA/Datafest 2020/stockData.csv")

monthly_air <- monthly_air[,-1]

monthly_air[,c(2:6)] <- sapply(monthly_air[,c(2:6)], function(x) scale(x) )

# get the numbers from the report

report_air <- read_xlsx("~/Dropbox/School/UCLA/Datafest 2020/Airline data/airline_revenue.xlsx")

summary(report_air)

report_air$`Net income` <- gsub(",","",report_air$`Net income`)

report_air$Year <- gsub("^([0-9]{4})","\\1-12-31",report_air$Year)

report_air$Year <- as.Date(report_air$Year,"%Y-%m-%d")

revenue_list <- report_air[,c(1:3)]

revenue_list <- revenue_list %>% group_by(Ticker) %>% mutate(scaled <- scale(Revenue))

colnames(revenue_list)[4] <- "scaled"

# lax data

lax <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/Airline data/lax.csv")

lax$Date <- as.Date(lax$Date, "%Y-%m-%d")

lax <- lax %>% mutate(cargo.scaled <- scale(Cargo.Total), operation.scaled <- scale(Flight.Operations))

lax <- lax[,c(1,4,5)]

colnames(lax)[2:3] <- c("Cargo.Scaled", "Operation.Scaled")

lax[,c(2,3)] <- lax[,c(2,3)] %>% lapply(function(x) round(x,2))

#write.csv(lax, "lax_scaled.csv")

```

```{r}

# S&P 500 by each GICS sectors

industrial_tickers <- tq_get(Industrials_sector$Symbol, from = date)

industrial_tickers <- industrial_tickers[,c(1,2,6)]

industrial_scaled <- industrial_tickers %>% group_by(symbol) %>% mutate(scaled <- scale(close))

colnames(industrial_scaled)[4] <- "scaled"

#write.csv(industrial_tickers,"industrial_ticker.csv")

# Aerospace and defence stock scaled

aerospace_tickers <- tq_get(c("BA","GD","HWM","HII","LHX","LMT","NOC","RTX","TXT","TDG"), from = date)

#write.csv(aerospace_tickers,"aerospace_tickers.csv")

aerospace_scale <- aerospace_tickers %>% group_by(symbol) %>% mutate(scaled <- scale(close))

colnames(aerospace_scale)[9] <- "scaled"

# Airline stocks scaled

airlines_tickers <- tq_get(c("ALK","AAL","DAL","LUV","UAL"), from = date)

#write.csv(airlines_tickers,"airlines_tickers.csv")

airlines_scale <- airlines_tickers %>% group_by(symbol) %>% mutate(scaled <- scale(close))

colnames(airlines_scale)[9] <- "scaled"

# Horel Resort & Cruise Lines stocks scaled

hoteletc_tickers <- tq_get(c("CCL","HLT","MAR","NCLH","RCL"), from = date)

#write.csv(hoteletc_tickers,"hoteletc_tickers.csv")

hoteletc_scale <- hoteletc_tickers %>% group_by(symbol) %>% mutate(scaled <- scale(close))

colnames(hoteletc_scale)[9] <- "scaled"

monthly_hotel <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/hotel.csv")

monthly_hotel <- monthly_hotel[-78,-1]

monthly_hotel[,c(2:7)] <- sapply(monthly_hotel[,c(2:7)], function(x) scale(x) )

# Restaurants stocks scaled

restaurant_tickers <- tq_get(c("CMG","DRI","MCD","SBUX","YUM"), from = date)

#write.csv(restaurant_tickers,"restaurant_tickers.csv")

restaurant_scale <- restaurant_tickers %>% group_by(symbol) %>% mutate(scaled <- scale(close))

colnames(restaurant_scale)[9] <- "scaled"

monthly_rest <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/restaurant.csv")

monthly_rest <- monthly_rest[-126,-1]

monthly_rest[,c(2:7)] <- sapply(monthly_rest[,c(2:7)], function(x) scale(x) )

# Railroad stocks scaled

railroad_tickers <- tq_get(c("CSX","KSU","NSC","UNP"), from = date)

#write.csv(railroad_tickers,"railroad_tickers.csv")

railroad_scale <- railroad_tickers %>% group_by(symbol) %>% mutate(scaled <- scale(close))

colnames(railroad_scale)[9] <- "scaled"

# Air freight scaled

monthly_afreight <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/airfreight.csv")

monthly_afreight <- monthly_afreight[-126,-1]

monthly_afreight[,c(2:6)] <- sapply(monthly_afreight[,c(2:6)], function(x) scale(x) )

# Cosmetic stock scaled

options("getSymbols.warning4.0"=FALSE)

options("getSymbols.yahoo.warning"=FALSE)

# Downloading Apple price using quantmod

getSymbols("KAOOY", from = '2019-01-01',

to = "2020-05-05",warnings = FALSE,

auto.assign = TRUE)

chart_Series(KAOOY)

chart_Series(KAOOY['2020-02/2020-05'])

# Ozone data

la_ozone <- ozone3 %>% filter(COUNTY == "Los Angeles")

la_ozone <- la_ozone %>% group_by(Site.Name) %>% mutate(scaled <- scale(Daily.Max.8.hour.Ozone.Concentration))

colnames(la_ozone)[21] <- "scaled"

west_la_ozone <- la_ozone %>% filter(Site.Name == "West Los Angeles")

west_la_ozone <- west_la_ozone[,c(1,5)]

la_pm25 <- pm25_total %>% filter(COUNTY == "Los Angeles")

la_pm25 <- la_pm25 %>% group_by(Site.Name) %>% mutate(scaled <- scale(Daily.Mean.PM2.5.Concentration))

colnames(la_pm25)[21] <- "scaled"

sc_la_pm25 <- la_pm25 %>% filter(Site.Name == "Santa Clarita")

sc_la_pm25 <- sc_la_pm25[,c(1,5,8)]

```

```{r}

# Scaled stock change in industrial sector

ggplot() +

geom_line(aes(x = date, y = scaled, col = symbol), data = industrial_scaled) +

ggtitle("Scaled Closed Stock Price of Industrial Sector")

# Scaled stock change in Aerospace

ggplot() +

geom_line(aes(x = date, y = scaled, col = symbol), data = aerospace_scale) +

ggtitle("Scaled Closed Stock Price of Aerospace companies")

# Scaled stock change in Airline Companies

ggplot() +

geom_line(aes(x = date, y = scaled, col = symbol), data = airlines_scale) +

ggtitle("Scaled Closed Stock Price of Airline Companies")

# Scaled stock change in Hotel resort & Curise Companies

ggplot() +

geom_line(aes(x = date, y = scaled, col = symbol), data = hoteletc_scale) +

ggtitle("Scaled Closed Stock Price of Hotel resort & Curise Companies")

# Scaled stock change in Restaurant Companies

ggplot() +

geom_line(aes(x = date, y = scaled, col = symbol), data = restaurant_scale) +

ggtitle("Scaled Closed Stock Price of Restaurant Companies")

# Scaled stock change in Railroad Companies

ggplot() +

geom_line(aes(x = date, y = scaled, col = symbol), data = railroad_scale) +

ggtitle("Scaled Closed Stock Price of Railroad Companies")

ggplot() +

geom_line(aes(x = Date, y = `^GSPC`), data = monthly_air) +

geom_line(aes(x = Date, y = DAL), data = monthly_air,color="blue") +

geom_line(aes(x = Date, y = AAL), data = monthly_air,color="red") +

geom_line(aes(x = Date, y = UAL), data = monthly_air,color="green") +

geom_line(aes(x = Date, y = LUV), data = monthly_air,color="purple")

# Air quality

ggplot() +

geom_line(aes(x = Year, y = scaled, col = Ticker), data = revenue_list)

ggplot() +

geom_line(aes(x = Date, y = scaled, col = Site.Name), data = la_ozone)

ggplot() +

geom_line(aes(x = Date, y = scaled, col = Site.Name), data = la_pm25)

ggplot() +

geom_line(aes(x = Date, y = Daily.Max.8.hour.CO.Concentration), data = co)

ggplot() +

geom_line(aes(x = Year, y = Revenue, col = Symbol), data = air_revenue_quarterly)

boxplot(t_air[,3] ~ cycle(t_air[,3]))

plot(decompose(t_air[,3], type = "mult"))

state_mobility %>% ggplot() +

geom_line(aes(x = date, y = transit),col = "green") +

geom_line(aes(x = date, y = retail), col = "orange") +

geom_line(aes(x = date, y = work), col = "blue")

```

# See if there was no COVID19

```{r}

# Airline Industry

without_covid_stock <- monthly_air[1:120,]

training <- ts(without_covid_stock[,-1], start = c(2010,1), end = c(2019, 12), freq = 12)

test = ts(monthly_air[121:125,-1], start = c(2020), end = c(2020,5), freq = 12)

GSPC = training[,1]

DAL = training[,2]

AAL = training[,3]

UAL = training[,4]

LUV = training[,5]

# Hotel Industry

without_covid_stock <- monthly_hotel[1:72,]

training_h <- ts(without_covid_stock[,-1], start = c(2014,1), end = c(2019, 12), freq = 12)

test_h = ts(monthly_hotel[73:77,-1], start = c(2020), end = c(2020,5), freq = 12)

CCL = training_h[,2]

HLT = training_h[,3]

MAT = training_h[,4]

NCLH = training_h[,5]

RCL = training_h[,6]

# Restaurant Industry

without_covid_stock <- monthly_rest[1:120,]

training_r <- ts(without_covid_stock[,-1], start = c(2010,1), end = c(2019, 12), freq = 12)

test_r = ts(monthly_rest[121:125,-1], start = c(2020), end = c(2020,5), freq = 12)

CMG = training_r[,2]

DRI = training_r[,3]

MCD = training_r[,4]

SBUX = training_r[,5]

YUM = training_r[,6]

# Air Freight Industry

without_covid_stock <- monthly_afreight[1:120,]

training_f <- ts(without_covid_stock[,-1], start = c(2010,1), end = c(2019, 12), freq = 12)

test_f = ts(monthly_afreight[121:125,-1], start = c(2020,1), end = c(2020,5), freq = 12)

CHRW = training_f[,2]

EXPD = training_f[,3]

FDX = training_f[,4]

UPS = training_f[,5]

# Airline revenue quartely

DAL_rev_train <- air_revenue_quarterly %>% filter(Symbol == "DAL")

DAL_rev_test <- DAL_rev_train[21,]

DAL_rev_train <- DAL_rev_train[-21,]

DAL_rev_train <- ts(DAL_rev_train[,4], start = c(2015,1), end = c(2019,4), freq = 4)

DAL_rev_test <- ts(DAL_rev_test[,4], start = c(2020,1), freq = 4)

AAL_rev_train <- air_revenue_quarterly %>% filter(Symbol == "AAL")

AAL_rev_test <- AAL_rev_train[21,]

AAL_rev_train <- AAL_rev_train[-21,]

AAL_rev_train <- ts(AAL_rev_train[,4], start = c(2015,1), end = c(2019,4), freq = 4)

AAL_rev_test <- ts(AAL_rev_test[,4], start = c(2020,1), freq = 4)

UAL_rev_train <- air_revenue_quarterly %>% filter(Symbol == "UAL")

UAL_rev_test <- UAL_rev_train[21,]

UAL_rev_train <- UAL_rev_train[-21,]

UAL_rev_train <- ts(UAL_rev_train[,4], start = c(2015,1), end = c(2019,4), freq = 4)

UAL_rev_test <- ts(UAL_rev_test[,4], start = c(2020,1), freq = 4)

LUV_rev_train <- air_revenue_quarterly %>% filter(Symbol == "LUV")

LUV_rev_test <- LUV_rev_train[21,]

LUV_rev_train <- LUV_rev_train[-21,]

LUV_rev_train <- ts(LUV_rev_train[,4], start = c(2015,1), end = c(2019,4), freq = 4)

LUV_rev_test <- ts(LUV_rev_test[,4], start = c(2020,1), freq = 4)

```

```{r}

#Regular differencing

reg.diff=diff(UAL, lag=1,diff=1)

#Seasonal differencing

seas.diff=diff(UAL, lag=12,diff=1)

#Regular seasonal differencing

seas.reg.diff=diff(reg.diff, lag=12,diff=1)

acf(reg.diff,lag=25, main="reg only diff")

acf(seas.diff,lag=25,main="seas only diff")

acf(seas.reg.diff,lag=25,main="reg and seas diff")

U.diff <- reg.diff

#Regular differencing

reg.diff=diff(DAL, lag=1,diff=1)

#Seasonal differencing

seas.diff=diff(DAL, lag=12,diff=1)

#Regular seasonal differencing

seas.reg.diff=diff(reg.diff, lag=12,diff=1)

acf(reg.diff,lag=25, main="reg only diff")

acf(seas.diff,lag=25,main="seas only diff")

acf(seas.reg.diff,lag=25,main="reg and seas diff")

D.diff <- reg.diff # MA1

#Regular differencing

reg.diff=diff(LUV, lag=1,diff=1)

#Seasonal differencing

seas.diff=diff(LUV, lag=12,diff=1)

#Regular seasonal differencing

seas.reg.diff=diff(reg.diff, lag=12,diff=1)

acf(reg.diff,lag=25, main="reg only diff")

acf(seas.diff,lag=25,main="seas only diff")

acf(seas.reg.diff,lag=25,main="reg and seas diff")

L.diff <- reg.diff

#Regular differencing

reg.diff=diff(AAL, lag=1,diff=1)

#Seasonal differencing

seas.diff=diff(AAL, lag=12,diff=1)

#Regular seasonal differencing

seas.reg.diff=diff(reg.diff, lag=12,diff=1)

acf(reg.diff,lag=25, main="reg only diff")

acf(seas.diff,lag=25,main="seas only diff")

acf(seas.reg.diff,lag=25,main="reg and seas diff")

A.diff <- reg.diff

```

# VAR Model

```{r, eval=FALSE}

airline.var=VAR(cbind(DAL,AAL,UAL,LUV),p=3, type="trend")

coef(airline.var)

acf(resid(airline.var))

acf(resid(airline.var)^2)

airline.predict = predict(airline.var, n.ahead = 5)

predict.val = ts(airline.predict$fcst$DAL[,1], start = c(2020,1), end = c(2020,5), freq = 12)

ci.low = ts(airline.predict$fcst$DAL[,2], start = c(2020,1), end = c(2020,5), freq = 12)

ci.high = ts(airline.predict$fcst$DAL[,3], start = c(2020,1), end = c(2020,5), freq = 12)

ts.plot(cbind(DAL, predict.val,ci.low, ci.high, test[,2]),

lty=c(1,1,3,3,1), col=c("black", "red","blue","blue","black"),

main="Predicted Monthly Stock Close Value of Delta Airline",

ylab="Stockprice scaled")

legend("topleft", legend=c("Stock Price","Predicted","Actual","CI"), lty=c(1,1,2,2), col=c(1,2,1,4))

```

# Exponential smoothing

```{r}

# Airline Stock Price

par(mfrow = c(1,1))

es = HoltWinters(DAL)

fitted.DAL = fitted(es)

DAL_forecast= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(DAL,fitted.DAL[,1]), DAL_forecast, test[,2],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Delta Airline", ylab = "Scaled Stock Price", xlim = c(2019,2020.3))

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(AAL)

fitted.AAL = fitted(es)

AAL_forecast= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(AAL,fitted.AAL[,1]), AAL_forecast, test[,3],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of American Airline", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(UAL)

fitted.UAL = fitted(es)

UAL_forecast= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(UAL,fitted.UAL[,1]), UAL_forecast, test[,4],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of United Airline", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(LUV)

fitted.LUV = fitted(es)

LUV_forecast= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(LUV,fitted.LUV[,1]), LUV_forecast, test[,5],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of SouthWest Airline", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

#write.csv(data.frame(rbind(training[,-1],test[,-1])), "air_stock.csv")

#write.csv(rbind(cbind(fitted.DAL[,1],fitted.AAL[,1],fitted.UAL[,1],fitted.LUV[,1]),cbind(DAL_forecast[,1],AAL_forecast[,1],UAL_forecast[,1],LUV_forecast[,1])),"predict_stock.csv")

# Airline Revenue

es = HoltWinters(DAL_rev_train)

DAL.rev.fit = fitted(es)

DAL.rev.forecast= predict(es, n.ahead = 1, prediction.interval = TRUE)

ts.plot(cbind(DAL_rev_train,DAL.rev.fit[,1]),DAL.rev.forecast,DAL_rev_test,lty=c(1,1), col=c(1,2), main="Predicted Quarterly Revenue of Delta Airline", ylab = "Quarter Revenue")

points(DAL.rev.forecast, col = "red")

points(DAL_rev_test)

legend("topleft", legend=c("Predicted","Actual"), lty=c(1,1), col=c(2,1))

es = HoltWinters(AAL_rev_train)

AAL.rev.fit = fitted(es)

AAL.rev.forecast= predict(es, n.ahead = 1, prediction.interval = TRUE)

ts.plot(cbind(AAL_rev_train,AAL.rev.fit[,1]),AAL.rev.forecast,AAL_rev_test,lty=c(1,1), col=c(1,2), main="Predicted Quarterly Revenue of American Airline", ylab = "Quarter Revenue")

points(AAL.rev.forecast, col = "red")

points(AAL_rev_test)

legend("topleft", legend=c("Predicted","Actual"), lty=c(1,1), col=c(2,1))

es = HoltWinters(UAL_rev_train)

UAL.rev.fit = fitted(es)

UAL.rev.forecast= predict(es, n.ahead = 1, prediction.interval = TRUE)

ts.plot(cbind(UAL_rev_train,UAL.rev.fit[,1]),UAL.rev.forecast,UAL_rev_test,lty=c(1,1), col=c(1,2), main="Predicted Quarterly Revenue of United Airline", ylab = "Quarter Revenue")

points(UAL.rev.forecast, col = "red")

points(UAL_rev_test)

legend("topleft", legend=c("Predicted","Actual"), lty=c(1,1), col=c(2,1))

es = HoltWinters(LUV_rev_train)

LUV.rev.fit = fitted(es)

LUV.rev.forecast= predict(es, n.ahead = 1, prediction.interval = TRUE)

ts.plot(cbind(LUV_rev_train,LUV.rev.fit[,1]),LUV.rev.forecast,LUV_rev_test,lty=c(1,1), col=c(1,2), main="Predicted Quarterly Revenue of SouthWest Airline", ylab = "Quarter Revenue")

points(LUV.rev.forecast, col = "red")

points(LUV_rev_test)

legend("topleft", legend=c("Predicted","Actual"), lty=c(1,1), col=c(2,1))

#write.csv(data.frame(rbind(cbind(LUV_rev_train,UAL_rev_train,AAL_rev_train,DAL_rev_train),cbind(LUV_rev_test,UAL_rev_test,AAL_rev_test,DAL_rev_test))),"air_revenue.csv")

#write.csv(rbind(cbind(DAL.rev.fit[,1],AAL.rev.fit[,1],UAL.rev.fit[,1],LUV.rev.fit[,1]),cbind(DAL.rev.forecast[,1],AAL.rev.forecast[,1],UAL.rev.forecast[,1],LUV.rev.forecast[,1])),"rev_fit.csv")

```

```{r}

# Hotel

es = HoltWinters(CCL)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(CCL,fitted.es[,1]), LT.forecast.es, test_h[,2],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Carnival Corp", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(HLT)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(HLT,fitted.es[,1]), LT.forecast.es, test_h[,3],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Hilton Worldwide Holdings Inc", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(MAT)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(MAT,fitted.es[,1]), LT.forecast.es, test_h[,4],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Marriott Int'l", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(NCLH)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(NCLH,fitted.es[,1]), LT.forecast.es, test_h[,5],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Norwegian Cruise Line Holdings", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(RCL)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(RCL,fitted.es[,1]), LT.forecast.es, test_h[,6],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Royal Caribbean Cruises Ltd", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

```

```{r}

# Air Freight

es = HoltWinters(CHRW)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(CHRW,fitted.es[,1]), LT.forecast.es, test_f[,2],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of C. H. Robinson Worldwide", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(EXPD)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(EXPD,fitted.es[,1]), LT.forecast.es, test_f[,3],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of Expeditors", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(FDX)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(FDX,fitted.es[,1]), LT.forecast.es, test_f[,4],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of FedEx Corporation", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

es = HoltWinters(UPS)

fitted.es = fitted(es)

LT.forecast.es= predict(es, n.ahead = 5, prediction.interval = TRUE)

ts.plot(cbind(UPS,fitted.es[,1]), LT.forecast.es, test_f[,5],lty=c(1,1,1,2,2,1), col=c(1,2,2,4,4,1), main="Predicted Monthly Stock Close Value of United Parcel Service", ylab = "Scaled Stock Price")

legend("topleft", legend=c("Predicted","Actual","CI"), lty=c(1,1,2), col=c(2,1,4))

```

```{r}

ar <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/air_revenue.csv")

as <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/air_stock.csv")

ps <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/predict_stock.csv")

pr <- read.csv("~/Dropbox/School/UCLA/Datafest 2020/rev_fit.csv")

ar$Date <- as.Date(ar$Date, "%m/%d/%Y")

as$Date <- as.Date(as$Date, "%m/%d/%Y")

ps$Date <- as.Date(ps$Date, "%m/%d/%Y")

pr$Date <- as.Date(pr$Date, "%m/%d/%Y")

ggplot() +

geom_line(aes(x = Date, y = DAL, color = "Actual"), data = as %>% filter(Date >= "2019-01-01")) +

geom_line(aes(x = Date, y = DAL, color = "Predicted"), data = ps %>% filter(Date >= "2019-01-01")) +

ylab("Scaled Stock Close Price") +

ggtitle("Delta Airline Difference Between Actual and Predicted Stock Price")

ggplot() +

geom_line(aes(x = Date, y = DAL_rev, color = "Actual"), data = ar %>% filter(Date > "2019-01-01")) +

geom_line(aes(x = Date, y = DAL, color = "Predicted"), data = pr %>% filter(Date > "2019-01-01")) +

ylab("Scaled Revenue") +

ggtitle("Delta Airline Difference Between Actual and Predicted Revenue")

ggplot() +

geom_line(aes(x = Date, y = AAL, color = "Actual"), data = as %>% filter(Date >= "2019-01-01")) +

geom_line(aes(x = Date, y = AAL, color = "Predicted"), data = ps %>% filter(Date >= "2019-01-01")) +

ylab("Scaled Stock Close Price") +

ggtitle("American Airline Difference Between Actual and Predicted Stock Price")

ggplot() +

geom_line(aes(x = Date, y = AAL_rev, color = "Actual"), data = ar %>% filter(Date > "2019-01-01")) +

geom_line(aes(x = Date, y = AAL, color = "Predicted"), data = pr %>% filter(Date > "2019-01-01")) +

ylab("Scaled Revenue") +

ggtitle("American Airline Difference Between Actual and Predicted Revenue")

ggplot() +

geom_line(aes(x = Date, y = UAL, color = "Actual"), data = as %>% filter(Date >= "2019-01-01")) +

geom_line(aes(x = Date, y = UAL, color = "Predicted"), data = ps %>% filter(Date >= "2019-01-01")) +

ylab("Scaled Stock Close Price") +

ggtitle("United Airline Difference Between Actual and Predicted Stock Price")

ggplot() +

geom_line(aes(x = Date, y = UAL_rev, color = "Actual"), data = ar %>% filter(Date > "2019-01-01")) +

geom_line(aes(x = Date, y = UAL, color = "Predicted"), data = pr %>% filter(Date > "2019-01-01")) +

ylab("Scaled Revenue") +

ggtitle("United Airline Difference Between Actual and Predicted Revenue")

ggplot() +

geom_line(aes(x = Date, y = LUV, color = "Actual"), data = as %>% filter(Date >= "2019-01-01")) +

geom_line(aes(x = Date, y = LUV, color = "Predicted"), data = ps %>% filter(Date >= "2019-01-01")) +

ylab("Scaled Stock Close Price") +

ggtitle("SouthWest Airline Difference Between Actual and Predicted Stock Price")

ggplot() +

geom_line(aes(x = Date, y = LUV_rev, color = "Actual"), data = ar %>% filter(Date > "2019-01-01")) +

geom_line(aes(x = Date, y = LUV, color = "Predicted"), data = pr %>% filter(Date > "2019-01-01")) +

ylab("Scaled Revenue") +

ggtitle("SouthWest Airline Difference Between Actual and Predicted Revenue")

```

​

---

title: "Oil & Number of flights & COVIDconfirmed"

output: html_notebook

---

```{r}

library(tidyverse)

library(ggplot2)

library(lubridate)

library(hrbrthemes)

```

Plot showing corrleation between oil price, # of flights and # of confirmed COVID patients

```{r}

number_of_flights <-read.csv("C:/Users/codud/Desktop/CL/DATAFEST/total-number-of-flights.csv",

fileEncoding="UTF-8-BOM",

check.names=FALSE)

oil_price <- read.csv("C:/Users/codud/Desktop/CL/DATAFEST/wti-crude-oil-prices-10-year-daily-chart.csv")

oil_price <- oil_price[2435:2518,]

corona <- read.csv("C:/Users/codud/Desktop/CL/DATAFEST/covid_US.txt")

corona_bydate <- corona %>%

group_by(date) %>%

summarise(confirmed = sum(confirmed), deaths = sum(deaths))

number_of_flights$`Number of flights` <- scale(number_of_flights$`Number of flights`)

number_of_flights <- number_of_flights %>%

select(Date, `Number of flights`)

number_of_flights$Date <- as.Date(number_of_flights$Date)

number_of_flights <- cbind(number_of_flights, as.data.frame(rep("Numberofflights", 90)))

colnames(number_of_flights) <- c("Date", "Value", "Name")

oil_price$value <- scale(oil_price$value)

oil_price$date <- as.Date(oil_price$date)

oil_price <- cbind(oil_price, as.data.frame(rep("OilPrice", 84)))

colnames(oil_price) <- c("Date", "Value", "Name")

corona_bydate$confirmed <- scale(corona_bydate$confirmed)

corona_bydate <- corona_bydate %>%

select("date","confirmed")

corona_bydate$date <- as.Date(corona_bydate$date, format = "%m-%d")

corona_bydate <- cbind(corona_bydate, as.data.frame(rep("ConfirmedCovid", 104)))

colnames(corona_bydate) <- c("Date", "Value", "Name")

Alldata <- rbind(number_of_flights, oil_price)

Alldata <- rbind(Alldata, corona_bydate)

ggplot(Alldata, aes(x=Date, y=Value, color=Name)) +

geom_point(size=3) +

theme_classic()

```

Percent change in oil price, # of flights and # of confirmed COVID patients by month

```{r}

number_of_flights <-read.csv("C:/Users/codud/Desktop/CL/DATAFEST/total-number-of-flights.csv",

fileEncoding="UTF-8-BOM",

check.names=FALSE)

oil_price <- read.csv("C:/Users/codud/Desktop/CL/DATAFEST/wti-crude-oil-prices-10-year-daily-chart.csv")

oil_price <- oil_price[2435:2518,]

corona_bydate <- read.csv("C:/Users/codud/Desktop/CL/DATAFEST/covid_US.txt")

corona_bydate$date <- as.Date(corona_bydate$date, format = "%m-%d")

corona_bydate <- corona_bydate %>%

group_by(date=floor_date(date, "month")) %>%

summarise(confirmed = mean(confirmed), deaths = mean(deaths)) %>%

mutate(newconfirm = (confirmed-lag(confirmed)))

corona_bydate

number_of_flights$Date <- as.Date(number_of_flights$Date)

number_of_flights <- number_of_flights %>%

group_by(Date=floor_date(Date, "month")) %>%

summarise(Number = mean(`Number of flights`)) %>%

mutate(pct_change = (Number/lag(Number) - 1) * 100)

number_of_flights

oil_price$date <- as.Date(oil_price$date)

oil_price <- oil_price %>%

group_by(date=floor_date(date, "month")) %>%

summarise(Number = mean(value)) %>%

mutate(pct_change = (Number/lag(Number) - 1) * 100)

oil_price

```

bottom of page