# library(dplyr) untuk melakukan agregasi lebih lanjut
# library(ggplot2) untuk keperluan visualisasi dan plotting
# library(ppclust) salah satu package yang digunakan untuk membuat model fuzzy clustering
# library(factoextra) digunakan untuk membuat Principal Component Analysis
# library(fclust) package yang membentuk model fuzzy clustering yang akhirnya digunakan
# library(readr) digunakan untuk membaca data csv
# library(lubridate) digunakan untuk mengambil informasi kuartal
# library(tidyverse) untuk melakukan agregasi lebih lanjut
# library(FactoMineR) digunakan untuk melakukan Principal Component Analysis
# library(ggpubr) tambahan fitur untuk package ggplot2
# library(plotly) untuk membuat plot menjadi interaktif
# library(glue) untuk membuat "tag" pada visualisasi

ratio_grouped <- read_csv("grouped.csv") # dataframe utama yang digunakan untuk visualisasi dan pembentukan model machine learning, dan

unique_stocks <- read_csv("unique_stocks.csv") # daftar perusahaan yang ada pada ratio_grouped, digunakan untuk menunjukkan jumlah unik yang ada

ratio_grouped<- ratio_grouped %>% 
  column_to_rownames("...1") %>% # mengubah kolom tersebut menjadi nama baris
  mutate_if(is.numeric, round, digits = 2) # melakukan pembulatan terhadap angka agar lebih mudah dilihat

unique_stocks <- unique_stocks %>% 
  select(-"...1") #menghilangkan kolom "...1" karena tidak digunakan

head(ratio_grouped)

rata2_data<- ratio_grouped %>% 
  summarise_all(.funs = "mean") %>% 
  mutate(across(where(is.numeric), round, 2)) %>% 
  pivot_longer(cols = c("GPM":"OPM"),names_to = "Variable", values_to = "Rata-Rata")

median_data<- ratio_grouped %>% 
  summarise_all(.funs = "median") %>% 
  mutate(across(where(is.numeric), round, 2)) %>% 
  pivot_longer(cols = c("GPM":"OPM"),names_to = "Variable", values_to = "Rata-Rata")

min_data <- ratio_grouped %>% 
  summarise_all(.funs = "min") %>% 
  mutate(across(where(is.numeric), round, 2)) %>% 
  pivot_longer(cols = c("GPM":"OPM"),names_to = "Variable", values_to = "Rata-Rata")

max_data <- ratio_grouped %>% 
  summarise_all(.funs = "max") %>% 
  mutate(across(where(is.numeric), round, 2)) %>% 
  pivot_longer(cols = c("GPM":"OPM"),names_to = "Variable", values_to = "Rata-Rata")

sd_data <- ratio_grouped %>% 
  summarise_all(.funs = "sd") %>% 
  mutate(across(where(is.numeric), round, 2)) %>% 
  pivot_longer(cols = c("GPM":"OPM"),names_to = "Variable", values_to = "Rata-Rata")

df_list <- list(rata2_data, median_data, min_data, max_data, sd_data)

ringkasan_data <- df_list %>% 
  reduce(full_join, by = "Variable") %>% 
  rename("Mean" = "Rata-Rata.x",
         "Median" = "Rata-Rata.y",
         "Min" = "Rata-Rata.x.x",
         "Max" =  "Rata-Rata.y.y",
         "Standard Deviation" = "Rata-Rata")

ringkasan_data

set.seed(149)
ratio_fclust <- Fclust(ratio_grouped, k = 2, ent = T, stand = 1, type = "Standard")

## Scatter Plot, untuk menunjukkan titik-titk data
for_plotly<- data.frame(pca_object3$x)
for_plotly$clus <- ratio_fclust$clus[,1]
percentage <- round(pca_object3$sdev^2 / sum(pca_object3$sdev^2)*100, 2)
percentage <- paste(colnames(for_plotly),"(",paste(as.character(percentage),"%",")", sep=""))
sample_name <- rownames(for_plotly)

## add frame
hull_group <- for_plotly %>% 
  mutate(cluster = ratio_fclust$clus[,1]) %>% 
  group_by(cluster) %>% 
  slice(chull(PC1, PC2))

## Variables
pca_loadings <- data.frame(Variables = rownames(pca_object3$rotation), pca_object3$rotation)

## Merging
for_plotly <- merge(ratio_grouped, for_plotly, by = 'row.names') %>% 
  column_to_rownames("Row.names")

## ggplotly object
final_viz <- ggplotly(ggplot()+
                        geom_polygon(hull_group,mapping = aes(x = PC1,
                                                              y = PC2,
                                                              fill = as.factor(cluster),
                                                              group = as.factor(cluster)),
                                     alpha = 0.2)+
                        scale_fill_discrete(labels = c("1", "2"))+
                        geom_point(for_plotly,mapping = aes(x = PC1,
                                                            y = PC2,
                                                            text = glue("Stock & Quarter : {rownames(for_plotly)}
                                                  Gross Profit Margin : {round(for_plotly$GPM,2)}
                                                  Working Capital : {round(for_plotly$WC,2)}
                                                  Retained Earnings : {round(for_plotly$`RE/TA`,2)}
                                                  Return on Assets : {round(for_plotly$ROA,2)}
                                                  Return on Equity : {round(for_plotly$ROE,2)}
                                                  Debt Ratio : {round(for_plotly$`TD/TA`,2)}
                                                  Total Asset Turnover : {round(for_plotly$ATR,2)}
                                                  Debt to Equity : {round(for_plotly$`TD/TE`,2)}
                                                  Net Profit Margin : {round(for_plotly$PM,2)}
                                                  Market Value of Equity/Book Value of Liability : {round(for_plotly$`Marcap/TL`,2)}
                                                  Current Ratio : {round(for_plotly$CR,2)}
                                                  Cash Ratio : {round(for_plotly$Csh_Rat,2)}
                                                  Operating Cash Flow : {round(for_plotly$OCF,2)}
                                                  Interest Coverage Ratio : {round(for_plotly$ICR,2)}
                                                  Operating Profit Margin : {round(for_plotly$OPM,2)}
                                                  "),
                                                  fill = as.factor(clus),
                                                  stroke = 0),
                                   size = 1.5,
                                   show.legend = F)+
                        geom_segment(data = pca_loadings, aes(
                          x = 0,
                          y = 0,
                          xend = (PC1*6),
                          yend = (PC2*6)
                        ),
                        arrow = arrow(length = unit(1/2, "picas")),
                        color = "black")+
                        annotate("text", x = (pca_loadings$PC1*8),
                                 y = (pca_loadings$PC2*7),
                                 label = pca_loadings$Variables)+
                        guides(fill = guide_legend(title = "Cluster"))+
                        labs(title = "Hasil Clustering")+
                        theme(plot.title = element_text(hjust = 0.5)), tooltip = "text")

SqEucDs <- function(x,y) sum( (x-y)^2) # Squared Euclidean Distance
unseen_data <- function(newData, centroids = ratio_fclust$H) { 
  dists = apply(centroids, 1, function(y) SqEucDs(y,newData))
  order(dists)[1]
}

ratio_fclust$H