Weekly - Gasoline
Tobi Joshua Samuel
2023-03-07
OVERVIEW OF THE DATASET
The dataset contains information about weekly oil production for two consecutive years. It has 173 observations and 7 variables. Here is a brief description of each variable:
Fiscal.Year: The fiscal year of the observation.
Fiscal.Week: The fiscal week of the observation.
Current.Year.Production: The current year’s production in barrels of oil.
Previous.Year.Production: The previous year’s production in barrels of oil.
Difference.From.Same.Week.Last.Year: The difference in production between the current and previous year’s production for the same fiscal week.
Current.Year.Cumulative.Production: The cumulative production for the current year up to the current week.
Cumulative.Difference: The cumulative difference in production between the current and previous year’s production up to the current week.
(1).LOAD THE REQUIRED LIBRARIES
library(ggplot2) # A popular package for creating graphics in R, with a syntax based on the grammar of graphics
library(forecast) # Provides methods and tools for forecasting time series data## Registered S3 method overwritten by 'quantmod':
## method from
## as.zoo.data.frame zoolibrary(xts) # Provides an extensible time series class for use with R's time-series functions## Loading required package: zoo##
## Attaching package: 'zoo'## The following objects are masked from 'package:base':
##
## as.Date, as.Date.numeric##
## ################################### WARNING ###################################
## # We noticed you have dplyr installed. The dplyr lag() function breaks how #
## # base R's lag() function is supposed to work, which breaks lag(my_xts). #
## # #
## # If you call library(dplyr) later in this session, then calls to lag(my_xts) #
## # that you enter or source() into this session won't work correctly. #
## # #
## # All package code is unaffected because it is protected by the R namespace #
## # mechanism. #
## # #
## # Set `options(xts.warn_dplyr_breaks_lag = FALSE)` to suppress this warning. #
## # #
## # You can use stats::lag() to make sure you're not using dplyr::lag(), or you #
## # can add conflictRules('dplyr', exclude = 'lag') to your .Rprofile to stop #
## # dplyr from breaking base R's lag() function. #
## ################################### WARNING ###################################(2). LOADING AND CLEANING THE DAT
data <- read.csv("weekly-gasoline.csv", header = TRUE, sep = ",")
df <- data.frame(data)2.1 HEAD and STR
head(df,n=6)## Fiscal.Year Fiscal.Week Current.Year.Production Previous.Year.Production
## 1 2023 16 2393748000 2417856000
## 2 2023 15 2367876000 2324364000
## 3 2023 14 2222052000 2402568000
## 4 2023 13 2209116000 2858856000
## 5 2023 12 2742138000 2641884000
## 6 2023 11 2561916000 2784768000
## Difference.From.Same.Week.Last.Year Current.Year.Cumulative.Production
## 1 -24108000 39649722000
## 2 43512000 37255974000
## 3 -180516000 34888098000
## 4 -649740000 32666046000
## 5 100254000 30456930000
## 6 -222852000 27714792000
## Cumulative.Difference
## 1 -3024672000
## 2 -3000564000
## 3 -3044076000
## 4 -2863560000
## 5 -2213820000
## 6 -2314074000str(df) # Examine the structure of the data## 'data.frame': 173 obs. of 7 variables:
## $ Fiscal.Year : int 2023 2023 2023 2023 2023 2023 2023 2023 2023 2023 ...
## $ Fiscal.Week : int 16 15 14 13 12 11 10 9 8 7 ...
## $ Current.Year.Production : num 2.39e+09 2.37e+09 2.22e+09 2.21e+09 2.74e+09 ...
## $ Previous.Year.Production : num 2.42e+09 2.32e+09 2.40e+09 2.86e+09 2.64e+09 ...
## $ Difference.From.Same.Week.Last.Year: chr "-24108000" "43512000" "-180516000" "-649740000" ...
## $ Current.Year.Cumulative.Production : num 3.96e+10 3.73e+10 3.49e+10 3.27e+10 3.05e+10 ...
## $ Cumulative.Difference : chr "-3024672000" "-3000564000" "-3044076000" "-2863560000" ...Data has 173 observations and 7 variables of interest
2.2 Convert All Character To Numeric For Easier Analysis
# Clean the data by converting "Difference.From.Same.Week.Last.Year" and "Cumulative.Difference" columns from character to numeric
df$Difference.From.Same.Week.Last.Year <- as.numeric(gsub(",", "", df$Difference.From.Same.Week.Last.Year))## Warning: NAs introduced by coerciondf$Cumulative.Difference <- as.numeric(gsub(",", "", df$Cumulative.Difference))## Warning: NAs introduced by coerciondf<-na.omit(df) # remove all NAs2.3 Check For Missing Values
sum(is.na(df)) # check if any value is missing## [1] 0No missing value
2.4 Quick Summary Statistics Of Each Variables
summary(df) # Summary statistics for all variables## Fiscal.Year Fiscal.Week Current.Year.Production
## Min. :2020 Min. : 1.00 Min. :2.209e+09
## 1st Qu.:2021 1st Qu.:10.25 1st Qu.:2.541e+09
## Median :2022 Median :24.00 Median :2.632e+09
## Mean :2022 Mean :24.68 Mean :2.628e+09
## 3rd Qu.:2022 3rd Qu.:38.00 3rd Qu.:2.741e+09
## Max. :2023 Max. :52.00 Max. :2.953e+09
## Previous.Year.Production Difference.From.Same.Week.Last.Year
## Min. :1.489e+09 Min. :-649740000
## 1st Qu.:2.420e+09 1st Qu.: -77322000
## Median :2.587e+09 Median : 59976000
## Mean :2.527e+09 Mean : 100947737
## 3rd Qu.:2.710e+09 3rd Qu.: 226306500
## Max. :2.953e+09 Max. :1135722000
## Current.Year.Cumulative.Production Cumulative.Difference
## Min. :2.433e+09 Min. :-7.050e+09
## 1st Qu.:2.736e+10 1st Qu.: 1.416e+08
## Median :6.196e+10 Median : 2.118e+09
## Mean :6.354e+10 Mean : 1.717e+09
## 3rd Qu.:9.702e+10 3rd Qu.: 4.250e+09
## Max. :1.350e+11 Max. : 5.863e+09(3). EXPLORATORY DATA ANALYSIS
3.1 Calculation of mean
Fiscal Year
mean(df$Fiscal.Year)## [1] 2021.588Fiscal Week
mean(df$Fiscal.Week)## [1] 24.68421Current.Year.Production
mean(df$Current.Year.Production)## [1] 2628006684Previous.Year.Production
mean(df$Previous.Year.Production)## [1] 2527058947Difference.From.Same.Week.Last.Year
mean(df$Difference.From.Same.Week.Last.Year)## [1] 100947737Current.Year.Cumulative.Production
mean(df$Current.Year.Cumulative.Production)## [1] 63544045404Cumulative.Difference
mean(df$Cumulative.Difference)## [1] 17167072633.2 Calculation of median
Fiscal Year
median(df$Fiscal.Year)## [1] 2022Fiscal Week
median(df$Fiscal.Week)## [1] 24Current.Year.Production
median(df$Current.Year.Production)## [1] 2632035000Previous.Year.Production
median(df$Previous.Year.Production)## [1] 2587200000Difference.From.Same.Week.Last.Year
median(df$Difference.From.Same.Week.Last.Year)## [1] 59976000Current.Year.Cumulative.Production
median(df$Current.Year.Cumulative.Production)## [1] 61960353000Cumulative.Difference
median(df$Cumulative.Difference)## [1] 21178290003.3 Calculation of minimum,maximum
Fiscal Year
min(df$Fiscal.Year)## [1] 2020max(df$Fiscal.Year)## [1] 2023Fiscal Week
min(df$Fiscal.Week)## [1] 1max(df$Fiscal.Week)## [1] 52Current.Year.Production
min(df$Current.Year.Production)## [1] 2209116000max(df$Current.Year.Production)## [1] 2952642000Previous.Year.Production
min(df$Previous.Year.Production)## [1] 1489110000max(df$Previous.Year.Production)## [1] 2952642000Difference.From.Same.Week.Last.Year
min(df$Difference.From.Same.Week.Last.Year)## [1] -649740000max(df$Difference.From.Same.Week.Last.Year)## [1] 1135722000Current.Year.Cumulative.Production
min(df$Current.Year.Cumulative.Production)## [1] 2433144000max(df$Current.Year.Cumulative.Production)## [1] 1.35e+11Cumulative.Difference
min(df$Cumulative.Difference)## [1] -7050414000max(df$Cumulative.Difference)## [1] 58626540003.4 Calculation of standard deviation
Fiscal Year
sd(df$Fiscal.Year)## [1] 0.910231sd(df$Fiscal.Year)## [1] 0.910231Fiscal Week
sd(df$Fiscal.Week)## [1] 15.61048sd(df$Fiscal.Week)## [1] 15.61048Current.Year.Production
min(df$Current.Year.Production)## [1] 2209116000max(df$Current.Year.Production)## [1] 2952642000Previous.Year.Production
sd(df$Previous.Year.Production)## [1] 272720976sd(df$Previous.Year.Production)## [1] 272720976Difference.From.Same.Week.Last.Year
sd(df$Difference.From.Same.Week.Last.Year)## [1] 304899984sd(df$Difference.From.Same.Week.Last.Year)## [1] 304899984Current.Year.Cumulative.Production
sd(df$Current.Year.Cumulative.Production)## [1] 39799555792sd(df$Current.Year.Cumulative.Production)## [1] 39799555792Cumulative.Difference
sd(df$Cumulative.Difference)## [1] 3041539561sd(df$Cumulative.Difference)## [1] 30415395613.5 Graphs
Visualize the trend of current oil production over time
ggplot(df, aes(x = Fiscal.Week, y = Current.Year.Production, color = factor(Fiscal.Year))) +
geom_line() +
labs(title = "Weekly Oil Production by Fiscal Year", x = "Fiscal Week", y = "Current Year Production")
Visualize the trend of previous oil production over time
ggplot(df, aes(x = Fiscal.Week, y = Previous.Year.Production, color = factor(Fiscal.Year))) +
geom_line() +
labs(title = "Weekly Oil Production by Fiscal Year", x = "Fiscal Week", y = "Previous Year Production")
Visualize the trend in Current Year Cumulative Production of oil production over time
ggplot(df, aes(x = Fiscal.Week, y = Current.Year.Cumulative.Production, color = factor(Fiscal.Year))) +
geom_line() +
labs(title = "Current Year Cumulative Production by Fiscal Year", x = "Fiscal Week", y = "Current.Year.Cumulative.Production")
Compare the difference between Difference From Same Week Last Year and Cumulative Difference over time
ggplot(df, aes(x = Difference.From.Same.Week.Last.Year, y =Cumulative.Difference, color = factor(Fiscal.Year))) +
geom_line() +
labs(title = "Difference From Same Week Last Year and Cumulative.Difference", x = "Fiscal Week", y = "Difference From Same Week Last Year Cumulative.Production")
Compare the current year’s production to the previous year’s production
# Compare the current year's production to the previous year's production
ggplot(df, aes(x = Previous.Year.Production, y = Current.Year.Production,color=factor(Fiscal.Year))) +
geom_point() +
geom_smooth(method = "lm", se = FALSE) +
labs(title = "Comparison of Current Year's Production to Previous Year's Production", x = "Previous Year Production", y = "Current Year Production")## `geom_smooth()` using formula = 'y ~ x'
3.6 Creating Histogram For Current Year Of Production
qplot(data =df, x=Current.Year.Production , geom = "histogram", bins = 15, color =I("yellow"))## Warning: `qplot()` was deprecated in ggplot2 3.4.0.
3.7 Creating Histogram For Previous Year Of Production
qplot(data =df, x=Previous.Year.Production , geom = "histogram", bins = 15, color =I("pink"))
(4) TIME SERIES ANALYSIS
4.1 Create a time series object using the Current.Year.Production variable
oil_ts <- ts(na.omit(df$Current.Year.Production), start = c(2020, 1), end = c(2023, 25), frequency = 52)4.2 Plot the time series object
plot(oil_ts, main = "Weekly Oil Production Time Series", xlab = "Fiscal Week", ylab = "Production (barrels)")
4.3 Decompose the time series object into its components
oil_decomp <- decompose(oil_ts)4.4 Visualize the trend, seasonal, and random components of the time series
autoplot(oil_decomp) +
labs(title = "Decomposition of Current Weekly Oil Production Time Series")
From the exploratory data analysis and time series analysis, we can see that there is a clear upward trend in oil production over the observed period. We can also see that there is a seasonal component to the data, with production peaking in the summer months and decreasing in the winter month.
(5) LINEAR REGRESSION MODEL
The provided dataset does not have any categorical variable, we cannot perform a classification analysis. Therefore, we will perform linear regression analysis. First, we need to split our dataset into training and testing sets. We will use 80% of the data for training and 20% for testing.
5.1 Split the dataset
set.seed(42)
train_index <- sample(1:nrow(df), 0.8*nrow(df))Train Data
train_data <- df[train_index,]
head(train_data)## Fiscal.Year Fiscal.Week Current.Year.Production Previous.Year.Production
## 50 2022 20 2519580000 2471658000
## 153 2020 21 2656290000 2640414000
## 66 2022 4 2740962000 2512230000
## 26 2022 44 2511054000 2873850000
## 75 2021 47 2814168000 2693334000
## 152 2020 22 2700684000 2664228000
## Difference.From.Same.Week.Last.Year Current.Year.Cumulative.Production
## 50 47922000 52795932000
## 153 15876000 55975542000
## 66 228732000 11045580000
## 26 -362796000 115000000000
## 75 120834000 119000000000
## 152 36456000 58676226000
## Cumulative.Difference
## 50 5038278000
## 153 153468000
## 66 959616000
## 26 3799068000
## 75 3529470000
## 152 189924000Test Data
test_data <- df[-train_index,]
head(test_data)## Fiscal.Year Fiscal.Week Current.Year.Production Previous.Year.Production
## 1 2023 16 2393748000 2417856000
## 7 2023 10 2426970000 2635122000
## 11 2023 6 2570148000 2722146000
## 14 2023 3 2625420000 2832396000
## 24 2022 46 2748312000 2743902000
## 45 2022 25 2539278000 2533104000
## Difference.From.Same.Week.Last.Year Current.Year.Cumulative.Production
## 1 -24108000 39649722000
## 7 -208152000 25152876000
## 11 -151998000 15375318000
## 14 -206976000 7609896000
## 24 4410000 120000000000
## 45 6174000 65715174000
## Cumulative.Difference
## 1 -3024672000
## 7 -2091222000
## 11 -1186584000
## 14 -694722000
## 24 3713220000
## 45 58626540005.2 Create a linear regression model using the lm() function in R. We will use the Current.Year.Production variable as our response variable and the Previous.Year.Production, Fiscal.Week, and Cumulative.Difference variables as our predictor variables.
lm_model <- lm(Current.Year.Production ~ Previous.Year.Production + Fiscal.Year + Difference.From.Same.Week.Last.Year, data = train_data)5.3 We can now use our trained model to make predictions on our testing dataset and calculate the accuracy of our predictions using the mean squared error.
predictions <- predict(lm_model, newdata = test_data)
mse <- mean((test_data$Current.Year.Production - predictions)^2)
mse## [1] 5.338338e-13The MSE value is very small, this shows that our model is a good fit.
5.4 Visualize the relationship between the two variables using a scatter plot
ggplot(df, aes(x = Previous.Year.Production, y = Current.Year.Production)) +
geom_point() +
geom_smooth(method = "lm") +
labs(title = "Scatter Plot of Current vs. Previous Year's Gasoline Production",
x = "Previous Year's Production (barrels)", y = "Current Year's Production (barrels)")## `geom_smooth()` using formula = 'y ~ x'
(6) CLUSTERING ANALYSIS USING K-MEANS ALGORITHM.
6.1 Visualize the three variables of interest in 3D
library("plot3D")
scatter3D(df$Fiscal.Year ,df$Current.Year.Production ,df$Previous.Year.Production)
dt <- cbind(df$Fiscal.Year ,df$Current.Year.Production ,df$Previous.Year.Production, col = NULL)
head(dt)## [,1] [,2] [,3]
## [1,] 2023 2393748000 2417856000
## [2,] 2023 2367876000 2324364000
## [3,] 2023 2222052000 2402568000
## [4,] 2023 2209116000 2858856000
## [5,] 2023 2742138000 2641884000
## [6,] 2023 2561916000 2784768000kmeans(dt,centers=3)## K-means clustering with 3 clusters of sizes 44, 18, 52
##
## Cluster means:
## [,1] [,2] [,3]
## 1 2022.227 2525927727 2688957409
## 2 2021.444 2565525667 2038710333
## 3 2021.096 2736009231 2559111692
##
## Clustering vector:
## [1] 1 2 1 1 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 1 3 1 1 1 1 1 1 3 1 1 1 1
## [38] 1 1 1 1 1 1 1 3 3 3 2 1 3 2 2 1 2 2 3 3 3 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
## [75] 3 3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 1 1 3 3 3 3 3 1 3 3 3 3 3 3 3 3
## [112] 3 3 3
##
## Within cluster sum of squares by cluster:
## [1] 1.216676e+18 1.660199e+18 1.119528e+18
## (between_SS / total_SS = 62.4 %)
##
## Available components:
##
## [1] "cluster" "centers" "totss" "withinss" "tot.withinss"
## [6] "betweenss" "size" "iter" "ifault"kmeans = kmeans(dt,centers=3)$cluster # vector of cluster belonging
dt$cm <- kmeans ## Warning in dt$cm <- kmeans: Coercing LHS to a listscatter3D(df$Fiscal.Year ,df$Current.Year.Production ,df$Previous.Year.Production, colvar=df$cm)
We will use the kmeans() function in R to cluster the data into 4 clusters based on the Current.Year.Production and Previous.Year.Production variables.
set.seed(42)
oil_cluster <- df[,c("Fiscal.Year","Current.Year.Production", "Previous.Year.Production")]
kmeans_model <- kmeans(oil_cluster, centers = 3)6.2 Visualize the clusters using a scatterplot and color the points according to their assigned cluster.
ggplot(oil_cluster, aes(x = Current.Year.Production, y = Previous.Year.Production, color = factor(kmeans_model$cluster))) +
geom_point() +
labs(title = "K-Means Clustering of Oil Production Data", x = "Current Year Production", y = "Previous Year Production")
(7) DISCUSSION
Based on the time series analysis of the
Current.Year.Production variable, we can see an upward trend in oil
production over the observed period. Additionally, there is a clear
seasonal component to the data, with production peaking in the summer
months and decreasing in the winter months. This information can be
useful for companies involved in the oil industry, as they can adjust
their production schedules to align with the seasonal patterns.
The linear regression model, which uses
Previous.Year.Production, Fiscal.Year, and
Difference.From.Same.Week.Last.Year as predictor variables, shows that
the model has a low mean squared error (MSE) of 5.338338e-13. The
predicted scatterplot shows that most points are close to the line,
indicating that the model provides a good fit for the data. This model
can be useful for predicting future oil production levels based on the
values of the predictor variables.
The clustering analysis using the kmeans
algorithm shows that most of the data points are close to each other.
The cluster means for Fiscal.Year, Current.Year.Production, and
Previous.Year.Production suggest that there are three distinct groups of
data points. This information can be useful for identifying patterns and
trends in the data, and for making decisions about resource allocation
and investment.
(8) CONCLUSION
The data analysis reveals useful information about the weekly oil production data. The time series analysis highlights the seasonal and upward trend of the production, which can help inform production scheduling. The linear regression model provides a useful tool for predicting future production levels based on predictor variables. The clustering analysis can help identify patterns and trends in the data, which can inform decisions about resource allocation and investment. Overall, the results of the data analysis can be used to inform strategic decision-making in the oil industry.