Project2 on Eagle Data

#this loads the data
options(width=10000)
load('eagle_data.Rdata')
(eagle_data
  %>% as_tibble() 
  ) %>% head

movement <- eagle_data %>% select(KPH:absVR) #select movement variables
movement_scaled <- scale(movement) # scale variables

movement_pca <- prcomp(movement, center = TRUE, scale. = TRUE) #pca of movement varaiables and scales and centers
movement_pca_scores <- as_tibble(movement_pca$x) #creates scores of the movement variables

movement_vars <- c("KPH", "Sn", "AGL", "VerticalRate", "abs_angle", "absVR") #this is to select the movement variables if need
sqrt_vars <- c("KPH", "Sn", "AGL", "abs_angle", "absVR") #selects variables where it is safe to use a sqrt transformation

kmeans_movement <- kmeans(movement_scaled, centers = 2) # use k means to make 2 clusters for first pca

fviz_pca_biplot( #create pca plot
  movement_pca, #data
  geom.ind = "point", #point plot
  col.ind = factor(kmeans_movement$cluster), #adds clusters to data
  palette = c("blue", "orange"), #color code clusters
  label = "var", #create labels for arrows
  col.var = "black", #black color arrows
  repel = TRUE #repel the text
) +
  labs(title = "PCA Biplot with K-means Clusters") #title name

set.seed(123) #set the seed
# this function makes 6 sets of data that randomly gets 1000 rows of data
samples <- lapply(1:6, function(i) {
  eagle_data %>% sample_n(1000)
})
# combined all 6 data sets
combined_samples <- bind_rows(
  samples,
  .id = "sample_id"   # optional: marks which sample each row came from
)
#select all the movement variables
combined_samples_movement <- combined_samples %>% select(all_of(movement_vars))
combined_transformed <- combined_samples_movement %>% # sqrt transformation for skewness
  mutate(across(all_of(sqrt_vars), sqrt)) %>%
  drop_na() # drop null

combined_scaled <- scale(combined_transformed) #scale
combined_pca <- prcomp(combined_scaled, center = TRUE, scale. = FALSE) #pca
kmeans_combined4 <- kmeans(combined_scaled, centers = 4) # 4 clusters using k means
pca_scores_combined <- as.data.frame(combined_pca$x[, 1:2]) %>%
  mutate(cluster = factor(kmeans_combined4$cluster)) #gets pc1 and pc2

#create sil plot to find optimal clusters
fviz_nbclust(combined_scaled, 
             FUNcluster = kmeans,
             method='silhouette',
             )

# creates a data frame of the scores
pca_eagle_clustered <- movement_pca_scores %>%
  as.data.frame() %>%
  mutate(cluster = factor(kmeans_movement$cluster))

# takes pc1 and gets the variable contributions in a %
contribution_pc1 <- fviz_contrib(combined_pca, choice = 'var', axes = 1) + 
  theme_classic(base_size = 16) +
  labs(x = 'Variable', 
       title = 'Contribution to first dimension') +
  # add this line to make the x-axis label text smaller
  theme(axis.text.x = element_text(size = 10))

# takes pc2 and gets the variable contributions in a %
contribution_pc2 <- fviz_contrib(combined_pca, choice = 'var', axes = 2) + 
  theme_classic(base_size = 16) +
  labs(x = 'Variable', 
       title = 'Contribution to second dimension') +
  # Add this line to make the x-axis label text smaller
  theme(axis.text.x = element_text(size = 10))

contribution_pc1 / contribution_pc2 #patches the 2 plots together

kmeans_movement4 <- kmeans(movement_scaled, centers = 4) # get 4 clusters using k means of the movement variables

#create data frame of the clusters form the kmeans
eagle_clustered <- eagle_data %>%
  as.data.frame() %>%
  mutate(cluster = factor(kmeans_movement4$cluster))

#similar as above code
eagle_clustered_combined <- combined_samples %>%
  as.data.frame() %>%
  mutate(cluster = factor(kmeans_combined4$cluster))

#this code is used to group the clusters in the data frame
cluster_means_combined <- eagle_clustered_combined %>%
  group_by(cluster) %>%
  summarize( #summarize the data to find the means of the variables
    across(all_of(movement_vars), mean, na.rm = TRUE)
  )
cluster_means_combined

#this package makes a table of the above code ouput for better visual
knitr::kable(cluster_means_combined, caption = "Table of cluster means") %>%
  kable_styling(
    font_size = 18, # Adjust this number (e.g., 8, 10, or 'small', 'x-small')
    full_width = FALSE # Keeps the table from stretching to 100% width
  )

#plot of the pca of the combined samples
fviz_pca_biplot(
    combined_pca,
    geom.ind = "point",
    habillage = kmeans_combined4$cluster, # Ccolor by cluster
    pointsize = 1, #point size is 1
    pointshape = 19, # points are shape 19
    label = "var", # show variable arrows
    col.var = "black", # color of the variable arrows
    repel = TRUE # avoid overlapping labels
) +
  #color codes the clusters on plot
scale_color_manual(
    values = c(
        "1" = "blue", # blue
        "2" = "darkorange", # orange
        "3" = "darkgreen",  # green
        "4" = "red"  # red
    )
) +
labs(
    title = "PCA Biplot (Combined 6-Sample Dataset)", #title 
    color = "Cluster"
) +
theme_minimal(base_size = 14) # minimal theme

boxplot_variables <- c('KPH', 'AGL', 'VerticalRate', 'abs_angle')#variables that were used for the boxplot

# define the colors once to reuse them easily and avoid typos
cluster_colors <- c
  "blue" = "blue" # map color names to color values if needed or just keep as is
  "darkorange" = "darkorange"
  "darkgreen" = "darkgreen"
  "red" = "red"


eagle_clustered <- combined_samples %>%
  as.data.frame() %>%
  mutate(
    cluster = factor(
      kmeans_combined4$cluster,
      levels = c("1", "2", "3", "4"), # original levels
      labels = c("Ascending", "Flapping", "Gliding", "Perching") # ew labels
    )
  )

# define the variable labels for facets
variable_labels <- c(
  'abs_angle' = '|Angle|',
  'AGL' = 'Above ground level',
  'VerticalRate' = 'Vertical rate',
  'KPH' = 'KPH'
)

eagle_clustered %>%
  pivot_longer( #pivoting the data
    cols = all_of(boxplot_variables),
    names_to = "variable",
    values_to = "value"
  ) %>%
ggplot(aes(x = cluster, y = value, fill = cluster)) +
  geom_boxplot(outlier.alpha = 0.2) + #create boxplot
  
  # use the labeller argument to apply custom labels to the facets
  facet_wrap(~ variable, scales = "free_y", labeller = as_labeller(variable_labels)) +
  
  # update scale fill manual to map the new labels to the desired colors
  scale_fill_manual(
    values = c(
      "Ascending" = "blue",
      "Flapping" = "darkorange",
      "Gliding" = "darkgreen",
      "Perching" = "red"
    )
  ) + 
  
  theme_minimal() +
  labs(
    title = "Inter-Cluster Differences Across Movement Variables",
    x = "Behavioral Cluster", #changed x axis label to be more descriptive
    y = "Value"
  ) +
  theme(# this is the theme
    strip.text = element_text(size = 12, face = "bold"),
    legend.position = "none"
  )

#this is to create a new data frame
eagle112 <- eagle_data %>% 
  filter(Animal_ID == '112') %>% #filter by animal id 112
  filter(segment_length == '608') #filter by segment length 608

eagle112_numeric <- eagle112 %>% select(where(is.numeric)) #select numeric values
eagle112_numeric_clean <- eagle112_numeric %>% select(-Animal_ID, -segment_id, -segment_length) # this variables turned null so they needed to be removed
eagle112_scaled <- scale(eagle112_numeric_clean)#scale data
kmeans_eaggle112 <- kmeans(eagle112_scaled, centers = 4)# 4 clusters using k means
eagle112$cluster_k4 <- factor(kmeans_eaggle112$cluster) #get the clusters

# this is a plot that will plot the eagle 112 flight path for the flight segment
ggplot(eagle112, aes(x = X, y = Y, color = cluster_k4)) + #color are the different clusters
  geom_point(size = 1) + # dots for each GPS fix
  coord_equal() +
  theme_minimal() +
  scale_color_manual( #color code clusters
    values = c(
      "1" = "blue",      # Map color to original cluster '1'
      "2" = "darkorange", # Map color to original cluster '2'
      "3" = "darkgreen",  # Map color to original cluster '3'
      "4" = "red"       # Map color to original cluster '4'
    ),
    labels = c(# label the clusters
      "1" = "perching", # Map original cluster '1' to new label 'perching'
      "2" = "ascending",  # Map original cluster '2' to new label 'ascending'
      "3" = "flapping",  # Map original cluster '3' to new label 'flapping'
      "4" = "gliding"   # Map original cluster '4' to new label 'gliding'
    )
  ) +
  labs(
    title = "Flight Path of Eagle 112 Colored by Movement Cluster", #title
    x = "X Coordinate", #x axis title
    y = "Y Coordinate",# y axis title
    color = "Movement Behavior"
  )