Flickr Data Analysis

Analyzing Flickr data can provide insights into image content, tags, user behaviors, trends, and geographic information. Flickr's API makes it easy to access such data for analysis in R. Here's how to perform Flickr data analysis step by step:

1. Setting Up

Register for Flickr API Access

1. Go to Flickr API.

2. Create an API key for your project.

2. Installing Required R Packages

You’ll need packages to interact with the API and perform analysis:

install.packages(c("httr", "jsonlite", "tidyverse", "lubridate"))

3. Connecting to Flickr API

Use your API key to fetch data.

library(httr)

library(jsonlite)

# Set up API key

api_key <- "your_api_key"

# Example: Search for public photos with a specific tag

query <- "sunset"

url <- paste0("https://api.flickr.com/services/rest/",

              "?method=flickr.photos.search",

              "&api_key=", api_key,

              "&tags=", query,

              "&format=json&nojsoncallback=1")

# Fetch data

response <- GET(url)

data <- content(response, as = "text", encoding = "UTF-8")

parsed_data <- fromJSON(data)

# View structure of data

str(parsed_data)

4. Extracting Useful Information

Flickr API returns metadata such as photo IDs, titles, tags, and locations. Extract these fields for analysis.

Example:

# Extract photo metadata

photos <- parsed_data$photos$photo

photo_metadata <- data.frame(

  id = photos$id,

  title = photos$title,

  owner = photos$owner,

  tags = photos$tags

)

head(photo_metadata)

5. Downloading Images

To analyze the images, construct URLs using the metadata.

# Construct image URLs

photo_metadata$image_url <- paste0("https://live.staticflickr.com/",

                                   photos$server, "/", photos$id, "_", photos$secret, ".jpg")

# Download example image

download.file(photo_metadata$image_url[1], destfile = "example.jpg", mode = "wb")

6. Geotagged Data Analysis

If geotagged photos are available, analyze spatial distributions.

# Example: Get geolocation data for a photo

photo_id <- photo_metadata$id[1]

geo_url <- paste0("https://api.flickr.com/services/rest/",

                  "?method=flickr.photos.geo.getLocation",

                  "&api_key=", api_key,

                  "&photo_id=", photo_id,

                  "&format=json&nojsoncallback=1")

geo_response <- GET(geo_url)

geo_data <- fromJSON(content(geo_response, as = "text"))

coordinates <- geo_data$photo$location

coordinates

Plot Geotagged Data

Use the ggmap package to visualize photo locations on a map:

install.packages("ggmap")

library(ggmap)

# Example data frame for geotagged photos

geo_photos <- data.frame(

  lat = c(37.7749, 34.0522, 40.7128),  # Example latitudes

  lon = c(-122.4194, -118.2437, -74.0060)  # Example longitudes

)

# Plot locations on a map

map <- get_map(location = "USA", zoom = 4)

ggmap(map) +

  geom_point(data = geo_photos, aes(x = lon, y = lat), color = "red", size = 3)

7. Tag Analysis

Flickr tags are a great way to analyze trends and categorize photos.

Count and Visualize Tags:

# Extract tags

all_tags <- unlist(strsplit(paste(photo_metadata$tags, collapse = " "), " "))

# Count tag frequency

tag_count <- as.data.frame(table(all_tags)) %>%

  arrange(desc(Freq))

# Plot top tags

top_tags <- tag_count[1:10, ]

ggplot(top_tags, aes(x = reorder(all_tags, -Freq), y = Freq)) +

  geom_bar(stat = "identity", fill = "skyblue") +

  labs(title = "Top 10 Tags", x = "Tags", y = "Frequency") +

  theme_minimal()

Word Cloud for Tags:

library(wordcloud)

wordcloud(words = tag_count$all_tags, 

          freq = tag_count$Freq, 

          max.words = 100, 

          colors = brewer.pal(8, "Dark2"))

8. Time-Based Analysis

Analyze trends in photo uploads over time.

# Convert date fields (if available)

photo_metadata$upload_date <- as.Date(sample(seq(as.Date('2020-01-01'), as.Date('2023-01-01'), by="days"), nrow(photo_metadata), replace = TRUE)) # Example dates

# Group by year/month

time_trend <- photo_metadata %>%

  mutate(year_month = floor_date(upload_date, "month")) %>%

  group_by(year_month) %>%

  summarise(photo_count = n())

# Plot time trends

ggplot(time_trend, aes(x = year_month, y = photo_count)) +

  geom_line(color = "blue") +

  labs(title = "Photo Upload Trends", x = "Time", y = "Number of Photos") +

  theme_minimal()

9. Sentiment Analysis of Titles and Descriptions

Perform sentiment analysis on photo titles or descriptions.

library(tidytext)

library(textdata)

# Tokenize titles

tokens <- photo_metadata %>%

  unnest_tokens(word, title)

# Sentiment analysis

bing <- get_sentiments("bing")

sentiment_scores <- tokens %>%

  inner_join(bing, by = "word") %>%

  count(word, sentiment, sort = TRUE)

# Plot sentiment

ggplot(sentiment_scores, aes(x = reorder(word, -n), y = n, fill = sentiment)) +

  geom_bar(stat = "identity") +

  labs(title = "Sentiment in Photo Titles", x = "Words", y = "Count") +

  coord_flip()

10. Advanced Analysis

• Image Recognition: Use tools like Google Vision API or Python-based libraries (e.g., TensorFlow) to analyze image content.

• Clustering: Group similar photos based on tags or metadata.

• Trend Prediction: Apply machine learning techniques to predict popular tags or upload patterns.