Starbucks_Final.Rmd

---
title: "Data Brew: Brewing Success with Starbucks Customer Insights"
subtitle: "TEAM 6 - Chekitha Swayampu, Hrushikesh Sai Seshagiri Chowdary Uppalapati, Swathi Murali Srinivasan, Vaishnavi Tamilvanan"
date: "2023-10-20"
output:
  prettydoc::html_pretty:
    theme: hpstr
    toc: true
  pdf_document:
    toc: yes
    toc_depth: 3
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r}
# Load required libraries
library(jsonlite)
library(dplyr)

# Load data
portfolio <- read.csv("data/portfolio.csv", row.names = 1)
profile <- read.csv("data/profile.csv", row.names = 1)
transcript <- read.csv("data/transcript.csv", row.names =1)
```

## DATA CLEANING AND PREPROCESSING
```{r}
# Check the structure of each dataset
str(portfolio)
str(profile)
str(transcript)

# Check summary statistics
summary(portfolio)
summary(profile)
summary(transcript)

# Check for missing values
sum(is.na(portfolio))
sum(is.na(profile))
sum(is.na(transcript))
```


```{r}
# PORTFOLIO DATA
# Clean up 'channels' column
portfolio$channels <- gsub("\\[|\\]|'", "", portfolio$channels)

# Convert 'channels' to a list
portfolio$channels <- lapply(strsplit(portfolio$channels, ","), trimws)

# Extract unique channels
unique_channels <- unique(unlist(portfolio$channels))

# Create binary columns for each channel
for (channel in unique_channels) {
  # Remove leading and trailing spaces
  cleaned_channel <- trimws(channel)
  
  # Create binary columns for each cleaned channel
  portfolio[paste0("channel_", cleaned_channel)] <- as.integer(sapply(portfolio$channels, function(x) cleaned_channel %in% x))
}

portfolio
```


```{r}
# PROFILE 
# Convert 'gender' to factor
profile$gender <- as.factor(profile$gender)

# Impute missing values in 'income' with mean
profile$income[is.na(profile$income)] <- mean(profile$income, na.rm = TRUE)
# Check unique values in 'became_member_on' to identify any issues
unique(profile$became_member_on)

# Remove any non-numeric characters
profile$became_member_on <- gsub("[^0-9]", "", profile$became_member_on)

# Convert 'became_member_on' to Date
profile$became_member_on <- as.Date(profile$became_member_on, format = "%Y%m%d")

head(profile)
```
```{r}
head(transcript)

```

```{r}
# TRANSCRIPT 
# Load the jsonlite library
library(jsonlite)

# Extract offer id from 'value' column
transcript$value <- gsub("'", "\"", transcript$value)  # Replace single quotes with double quotes
transcript$offer_id <- sapply(transcript$value, function(x) {
  parsed_value <- fromJSON(x, simplifyVector = TRUE)
  if (!is.null(parsed_value) && 'offer id' %in% names(parsed_value)) {
    return(parsed_value[['offer id']])
  } else {
    return(NA)
  }
})

# Drop the 'value' column as we've extracted the offer id
transcript <- subset(transcript, select = -c(value))



```


```{r}
head(transcript)
```
## FEATURE ENTINEERING

```{r}
# FE PROFILE 

# Calculate membership duration in days
profile$membership_duration <- as.numeric(difftime(Sys.Date(), profile$became_member_on, units = "days"))
head(profile)
```


```{r}
# FE TRANSCRIPT
head(transcript)

# Create binary columns for different events
transcript$offer_received <- as.integer(transcript$event == "offer received")
transcript$offer_viewed <- as.integer(transcript$event == "offer viewed")
transcript$offer_completed <- as.integer(transcript$event == "offer completed")
transcript$transaction <- as.integer(transcript$event == "transaction")

```


```{r}
head(transcript)
```

## MERGING DATA
```{r}
# Merge profile and transcript datasets based on 'id' (customer ID)
merged_data <- merge(profile, transcript, by.x = "id", by.y = "person", all.x = TRUE)

# Merge the resulting dataset with the portfolio dataset based on 'offer_id'
merged_data <- merge(merged_data, portfolio, by.x = "offer_id", by.y = "id", all.x = TRUE)

# Check the structure and summary of the merged dataset
str(merged_data)
summary(merged_data)
```


```{r}
merged_data
```

## K-MEANS CLUSTERING

```{r}
# Select relevant features for clustering
features_for_clustering <- merged_data[, c("age", "income", "membership_duration", "reward", "difficulty", "duration")]

# Check for missing values
if (any(sapply(features_for_clustering, function(x) any(is.na(x))))) {
  # Impute missing values with mean
  features_for_clustering <- lapply(features_for_clustering, function(x) {
    x[is.na(x)] <- mean(x, na.rm = TRUE)
    x
  })
}

# Convert the list to a data frame
features_for_clustering <- as.data.frame(features_for_clustering)

# Check for infinite values
if (any(sapply(features_for_clustering, function(x) any(!is.finite(x))))) {
  # Replace infinite values with appropriate values or investigate the cause
  features_for_clustering <- lapply(features_for_clustering, function(x) {
    x[!is.finite(x)] <- 0
    x
  })
}

# Standardize the features after handling missing and infinite values
scaled_features <- scale(features_for_clustering)

# Check for missing or infinite values after scaling
any(is.na(scaled_features))
any(!is.finite(scaled_features))
```

```{r}
# Elbow Method to find the optimal number of clusters
wss <- numeric(4)  # Set the maximum number of clusters you want to try (e.g., 10)

for (i in 1:4) {
  kmeans_result <- kmeans(scaled_features, centers = i)
  wss[i] <- kmeans_result$tot.withinss
}

# Plot the elbow curve
plot(1:4, wss, type = "b", pch = 19, frame = FALSE, 
     xlab = "Number of Clusters", ylab = "Total Within-Cluster Sum of Squares",
     main = "Elbow Method")

# Add a line at the "elbow" or optimal number of clusters
abline(v = which.min(wss), col = "red", lty = 2)

# Print the optimal number of clusters
cat("Optimal Number of Clusters:", which.min(wss), "\n")


```

```{r}
# Example: K-means clustering with k clusters
k <- 4  # Replace with your desired number of clusters
kmeans_result <- kmeans(scaled_features, centers = k)

# View the clustering results
str(kmeans_result)

# Assign cluster labels to each observation in your data
merged_data$cluster <- kmeans_result$cluster

# Check the distribution of clusters
table(merged_data$cluster)

# Explore the characteristics of each cluster
cluster_summary <- aggregate(. ~ cluster, data = merged_data[, c("cluster", "age", "income", "membership_duration", "reward", "difficulty", "duration")], mean)
print(cluster_summary)

```


```{r}
# Bar plot of cluster distribution
barplot(table(merged_data$cluster), main = "Cluster Distribution", xlab = "Cluster", ylab = "Count", col = rainbow(k))
```



```{r}

```



```{r}


```



```{r}


```


```{r}


```



```{r}


```


```{r}


```


```{r}


```


```{r}


```