1. Warm-up: Cars Dataset Median

The built-in cars dataset contains speed and stopping distances of cars.
I’ll compute the median of the first column (speed).

data(cars)
median_speed <- median(cars[,1])
median_speed

## [1] 15

✅ Answer: The median of the first column is 15.

2. Extracting Bitcoin Data via API

I’ll use the CryptoCompare API to retrieve daily BTC-USD prices for the past 100 days.
This demonstrates working with JSON data.

# Construct the URL
url <- "https://min-api.cryptocompare.com/data/v2/histoday?fsym=BTC&tsym=USD&limit=100"

# Read JSON
btc_data <- fromJSON(url)

# Inspect structure
str(btc_data, max.level = 2)

## List of 6
##  $ Response  : chr "Success"
##  $ Message   : chr ""
##  $ HasWarning: logi FALSE
##  $ Type      : int 100
##  $ RateLimit : Named list()
##  $ Data      :List of 4
##   ..$ Aggregated: logi FALSE
##   ..$ TimeFrom  : int 1751155200
##   ..$ TimeTo    : int 1759795200
##   ..$ Data      :'data.frame':   101 obs. of  9 variables:

# Extract DataFrame
df <- btc_data$Data$Data

# Convert time to date
df$time <- as.POSIXct(df$time, origin="1970-01-01")

# Preview data
head(df)

##                  time     high      low     open volumefrom   volumeto    close
## 1 2025-06-28 20:00:00 108538.4 107233.6 107346.4    6094.35  657403121 108391.9
## 2 2025-06-29 20:00:00 108815.7 106756.4 108391.9   14220.51 1530756559 107169.8
## 3 2025-06-30 20:00:00 107574.1 105295.0 107169.8   15730.48 1673173145 105724.2
## 4 2025-07-01 20:00:00 109818.5 105143.1 105724.2   19611.45 2124516059 108886.6
## 5 2025-07-02 20:00:00 110584.4 108579.6 108886.6   16647.63 1825182142 109639.0
## 6 2025-07-03 20:00:00 109810.0 107283.9 109639.0   10440.09 1130647601 108027.7
##   conversionType conversionSymbol
## 1         direct                 
## 2         direct                 
## 3         direct                 
## 4         direct                 
## 5         direct                 
## 6         direct

Now I’ll find the maximum daily close price:

max_close <- max(df$close, na.rm = TRUE)
max_close

## [1] 124723

✅ Answer: This returns the max daily close over the 100 days (value depends on current data).

3. Mini Project: Bitcoin Price Trends (Data ETL + Cleaning)

3.1 Project Title and Research Questions

Title: Bitcoin Price Trends Over the Last 100 Days

Research Questions:
1. What is the overall trend in BTC closing prices?
2. Are there missing or special values in the dataset?
3. How does mean and kNN imputation affect missing values?
4. Can we visualize the price changes clearly over time?

3.2 Data Extraction

I’ll reuse the df dataset extracted above. This dataset includes OHLCV data (open, high, low, close, volume) for BTC.

summary(df)

##       time                          high             low        
##  Min.   :2025-06-28 20:00:00   Min.   :107574   Min.   :105143  
##  1st Qu.:2025-07-23 20:00:00   1st Qu.:112938   1st Qu.:110217  
##  Median :2025-08-17 20:00:00   Median :116361   Median :114140  
##  Mean   :2025-08-17 20:00:00   Mean   :116154   Mean   :113405  
##  3rd Qu.:2025-09-11 20:00:00   3rd Qu.:118927   3rd Qu.:116866  
##  Max.   :2025-10-06 20:00:00   Max.   :126287   Max.   :123143  
##       open          volumefrom       volumeto             close       
##  Min.   :105724   Min.   : 3465   Min.   :3.747e+08   Min.   :105724  
##  1st Qu.:111292   1st Qu.:10960   1st Qu.:1.291e+09   1st Qu.:111549  
##  Median :115392   Median :18335   Median :2.125e+09   Median :115401  
##  Mean   :114732   Mean   :18821   Mean   :2.168e+09   Mean   :114878  
##  3rd Qu.:117778   3rd Qu.:25660   3rd Qu.:2.999e+09   3rd Qu.:117853  
##  Max.   :124723   Max.   :50123   Max.   :5.882e+09   Max.   :124723  
##  conversionType     conversionSymbol  
##  Length:101         Length:101        
##  Class :character   Class :character  
##  Mode  :character   Mode  :character  
##                                       
##                                       
##

3.3 Detecting and Introducing Missing Values

To demonstrate imputation, I’ll introduce some artificial missing values into the close column.

set.seed(4310)
missing_idx <- sample(1:nrow(df), 5)
df$close[missing_idx] <- NA
table(is.na(df$close))

## 
## FALSE  TRUE 
##    96     5

3.4 Detecting Special Values

any(is.nan(df$close))

## [1] FALSE

any(is.infinite(df$close))

## [1] FALSE

No NaN or Inf values were detected.

3.5 Imputation: Mean and kNN

Mean Imputation

df_mean_imp <- df
mean_val <- mean(df_mean_imp$close, na.rm = TRUE)
df_mean_imp$close[is.na(df_mean_imp$close)] <- mean_val
table(is.na(df_mean_imp$close))

## 
## FALSE 
##   101

kNN Imputation

I’ll use the VIM package for kNN imputation (k = 5 by default).

df_knn_imp <- kNN(df, variable = "close")

##         high          low         open   volumefrom     volumeto         high 
## 1.075741e+05 1.051431e+05 1.057242e+05 3.465160e+03 3.746681e+08 1.262873e+05 
##          low         open   volumefrom     volumeto 
## 1.231435e+05 1.247230e+05 5.012319e+04 5.881848e+09

table(is.na(df_knn_imp$close))

## 
## FALSE 
##   101

3.6 Visualization

I’ll plot the closing price trends before and after imputation.

ggplot(df, aes(x = time, y = close)) +
  geom_line(color = "blue") +
  labs(title = "Bitcoin Daily Closing Prices (With Missing Values)", x = "Date", y = "Closing Price (USD)")

ggplot(df_mean_imp, aes(x = time, y = close)) +
  geom_line(color = "green") +
  labs(title = "Bitcoin Daily Closing Prices (Mean Imputation)", x = "Date", y = "Closing Price (USD)")

ggplot(df_knn_imp, aes(x = time, y = close)) +
  geom_line(color = "red") +
  labs(title = "Bitcoin Daily Closing Prices (kNN Imputation)", x = "Date", y = "Closing Price (USD)")

4. Quiz Questions & Answers

Below are the answers to the quiz associated with this assignment:

Q1: Median of first column in cars dataset?
✅ C. 15

Q2: Extract BTC data and find max close.
✅ Code shown in Section 2, answer depends on latest data.

Q3: Mini project steps (Title, Questions, Data, Cleaning, Imputation).
✅ Completed in Section 3.

Q4: Presentation Bonus.
✅ If presented, click True.

DATA 4310 Mini Project - Bitcoin Price Trends

Elias Bekdas