WALMART PROJECT.Rmd

---
title: "R Notebook"
output:
  html_notebook: default
  html_document:
    df_print: paged
  pdf_document: default
  word_document: default
---

This is an [R Markdown](http://rmarkdown.rstudio.com) Notebook. When you execute code within the notebook, the results appear beneath the code. 

Try executing this chunk by clicking the *Run* button within the chunk or by placing your cursor inside it and pressing *Cmd+Shift+Enter*. 

```{r}
#Loading the dataset in R
library(readxl)
library(readxl)
getwd()
Walmart_Retail_store <- read_excel("Walmart Retail store.xlsx")
walmart <- Walmart_Retail_store
View(walmart)
summary(walmart)
str(walmart)
library(dplyr)
```
```{r}
# Which Store has max sales
# Aggregating Weekly Sales for the Stores
max_sales <- summarize(group_by(walmart, Store), Weekly_Sales = sum(Weekly_Sales))
max_sales
```
```{r}
# Which Store has max sales
# Max Weekly Sales
max(max_sales$Weekly_Sales)
```
```{r}
# Which Store has max sales
# OR
View(arrange(max_sales, desc(Weekly_Sales)))
# Store 20 has the max sales of $301397792
```
```{r}
# Which Store has max standard deviation? What is the coefficient of mean to standard deviation?
# Std. Deviation of Store's Weekly Sales
ssd <- summarize(group_by(walmart, Store), sd_weeklysales = sd(Weekly_Sales))
ssd
```
```{r}
# Which Store has max standard deviation? What is the coefficient of mean to standard deviation?
# Max SD
max(ssd)
# Store 14 has the max Standard Deviation
```
```{r}
# Which Store has max standard deviation? What is the coefficient of mean to standard deviation?
# Mean of Weekly Sales of each store
smean <- summarize(group_by(walmart, Store), m_weeklysales = mean(Weekly_Sales))
smean
```
```{r}
# Which Store has max standard deviation? What is the coefficient of mean to standard deviation?
# Max Mean
max(smean)
# Store 20 has the max mean
```
```{r}
# Which Store has max standard deviation? What is the coefficient of mean to standard deviation?
# Calculating Coefficient of Mean to Standard Deviation
# St. Deviation / Mean
str(walmart)
cmsd <- summarize(group_by(walmart, Store), coeff_weeklysales = mean(Weekly_Sales)/sd(Weekly_Sales))
cmsd
```
```{r}
# Max Coefficient of Mean to Standard Deviation
max(cmsd$coeff_weeklysales)
# Store 37 has the max coefficient of mean to standard deviation
```
```{r}
# Adding a new column for Quarters in the Walmart Data
install.packages(lubridate)
library(lubridate)
date.new <- as.Date(walmart$Date, format='%m-%d-%Y')
Quarter <- quarter(date.new)
walmart$Quarter <- Quarter
str(walmart)
```
```{r}
# Calculating Quarterly Sales for each store
q.sales <- summarize(group_by(walmart, Store, Quarter), w.sales = sum(Weekly_Sales))
q.sales
```
```{r}
#Creating a new DATE variable with Date formate
walmart.date <- as.Date(walmart$Date)
walmart$DATE <- walmart.date
str(walmart)
```
```{r}
# Which holidays have higher mean sales than the mean sales in non-holiday season for all the stores together?
# Some holidays have a negative impact on sales
# Part 1 - Summarized the average of non-holiday sales of all the stores 
sales <- summarize(group_by(walmart, DATE), avg.sales = mean(Weekly_Sales))
sales
```
```{r}
# Which holidays have higher mean sales than the mean sales in non-holiday season for all the stores together?
# Average of all the non-holiday sales
mean.sales <- mean(sales$avg.sales)
mean.sales
```
```{r}
# Which holidays have higher mean sales than the mean sales in non-holiday season for all the stores together?
# Part - Summarized sales with Holiday = 1
holiday.sales <- filter(walmart, Holiday_Flag == 1)
holiday.sales1 <- summarize(group_by(holiday.sales, DATE), avg.sales = mean(Weekly_Sales))
holiday.sales1
```
```{r}
# Which holidays have higher mean sales than the mean sales in non-holiday season for all the stores together?
mean.holidaysales <- mean(holiday.sales1$avg.sales)
mean.holidaysales
# Mean of holiday sales is higher than the mean of non-holiday sales
```
```{r}
# Which holidays have higher mean sales than the mean sales in non-holiday season for all the stores together?
# Which mean is higher
mutate(holiday.sales1, Which.Mean.Is.Higher = ifelse(avg.sales > 1046965, "Holiday Sales", "Non-Holiday Sales"))
```

```{r}
#  Provide a monthly and a semester view of sales in units and give insights
w.month <- month(date.new)
walmart$Month <- w.month
w.semester <- semester(date.new)
walmart$Semester <- w.semester
walmart
```
```{r}
# Provide a monthly and a semester view of sales in units and give insights
# grouping the data by month over the years
group.month <- summarize(group_by(walmart, Month), tot.sales = sum(Weekly_Sales))
group.month
```
```{r}
#  Provide a monthly and a semester view of sales in units and give insights
plot(group.month, type="l")
# The graph shows a rapid increase in the sales in month of February and a decline in sales in the month of November and an excellerated increase in sales in December
```
```{r}
# Provide a monthly and a semester view of sales in units and give insights
# Grouping the data by semesters over the years
group.sem <- summarize(group_by(walmart, Semester), tot.sales = sum(Weekly_Sales))
group.sem
```
```{r}
# Provide a monthly and a semester view of sales in units and give insights
plot(group.sem, type="l")
```
```{r}
# Provide a monthly and a semester view of sales in units and give insights
# The graph shows an increasing trend from semester 1 to semester 2
```

```{r}
####### PART 2 - Model Development ########
```
```{r}
# Step 1 - Filter store 1
w.model <- filter(walmart, Store == 1)
w.model
```
```{r}
# Changing Dates in days by creating a new variable
date.num = c(1:143)
cbind(w.model, date.num)
```
```{r}
# Plotting the Weekly Sale of store 1 over the years for store 1 to see the trend
library(ggplot2)
with(w.model, plot(date.num, Weekly_Sales, type = "l"))
```
```{r}
# Equation
# predicted y hat(Weekly_Sales) = Intercept + B1(CPI) + B2(Unemployment) + B3(Fuel Price) + B4(date.num) + E

# Hypothesis
# H0: B1 = B2 = B3 = 0    <- This means that the 3 variables have no impact on y
# H1: B1 not equal to B2 not equal to B3 not equal to 0  <- meaning that they have an impact on y
```
```{r}
# Linear Regression Model
lm.walmart <- lm(formula = Weekly_Sales ~ CPI + Unemployment + Fuel_Price + date.num, data = w.model )
lm.walmart
```
```{r}
summary(lm.walmart)
# Variable unemployment is significant at 5% level
# Variables CPI and Fuel_Price are not significant at 5% level
# R Squared = 0.08517
# Adjusted R Squared = 0.05865
```
```{r}
# Let's see if we get better coefficients omitting CPI and Fuel_Price
# Equation
# Predicted y hat (Weekly Sales) = Intercept + B1(Unemployment) + B2(date.num) + E
lm2.walmart <- lm(formula = Weekly_Sales ~ Unemployment + date.num, data = w.model)
lm2.walmart
```
```{r}
summary(lm2.walmart)
```
```{r}
# The first model has a higher R squared as compared to the second model 
# The second model has better coefficients than the first model based on looking at the t value and p value. Model 2 is better as all the coefficients have a higher significance

```
```{r}
# Checking how accurately the second model predicts as compared to the first model
predicted_sales1 <- predict(lm.walmart)
View(predicted_sales1)
plot(predicted_sales1, type ="l")
```
```{r}
# Predicting using 2 model
predicted_sales2 <- predict(lm2.walmart)
View(predicted_sales2)
plot(predicted_sales2, type = "l")
```