---
title: "Covid/Hotel Data (HD)"
output: 
  flexdashboard::flex_dashboard:
    orientation: columns
    vertical_layout: fill
    #theme: sandstone
    css: fd.css
    storyboard: false
    social: menu
    source: embed
    keep_tex: yes
---

```{r setup, include=FALSE, eval=TRUE}
library(flexdashboard)
library(tidyverse)
library(plotly)
library(dplyr)
library(lubridate)
library(plotly)
library(ggplot2)
library(modelr)
library(corrplot)
library(corrgram)
library(janitor)
library(DT)
library(knitr)
library(formattable)
library(rmarkdown)
library(sqldf)
library(party)
library(partykit)
library(caret)
library(packrat)
library(rsconnect)
library(skimr)
library(sjmisc)
library(kableExtra)
library(highcharter)
library(psych)
library(ggcorrplot)


covid_ds1<- read_csv(file = "C:/Reference Material/Machine_Learning/covid-data.csv")
hotel_ds1 <-read_csv(file = "C:/Reference Material/Machine_Learning/Hotel_Reservations.csv")

#########################################################################################
nc <- ncol(covid_ds1) # number of columns in dataset

## add columns year_no,month_no,day_no
covid_ds1 <- covid_ds1 %>% 
  mutate(year_no  = factor(year(date))) %>% 
  mutate(month_no = factor(month(date))) %>% 
  mutate(day_no   = factor(day(date))) %>% 
  mutate(total_row_cnt = nrow(.)) %>% 
  mutate_if(is_character,factor) ##converts all character colmuns to facto

hotel_ds1 <- hotel_ds1 %>% 
   mutate(date = make_date(arrival_year,arrival_month,arrival_date)) %>% 
   mutate(wday = wday(date,label = TRUE)) %>% 
   mutate_if(is.character,factor) ##converts all character colmuns to factors
  

  
```

# **Covid -NA's analysis**


Column {data-width=550}
-----------------------------------------------------------------------

### **NA summary in data set**
###### We need to clarify if columns with too many NA's are of any use. So, the calculation is as follows .. any columns with more than 20% total NA's per total rows will be discarded for now - see code below for logic. For reference - the totalnumber of columns in the original data set is `r nc`. The case for bringing back columns is up for discussion and how these NAs should be replaced & with what methodology - "Imputation". One possibility is to take the mean for each column and replace NA with the mean - but will need careful consideration. If there is a need to use these columns that have high counts of NA then we can analyse each column on it's merit by only using the rows that have no NA.
**Early detection of too many NA's (missing data points) - therefore, can not produce a meaningful dashboard**

```{r, echo = TRUE,class.source = "small"}
## NA SUMMARY
## Can use .. filter(!complete.cases(.)) this filters all rows with an NA
column_na_cnt <- covid_ds1 %>% 
  summarise_all(~ sum(is.na(.))) ## counts all NAs in columns

t1 <- column_na_cnt %>% 
  rotate_df() %>%  ## needs library(sjmisc) .. rotates columns to rows or visa versa
  arrange((V1))

t1 <- t1 %>% 
  mutate(tot_row_count = nrow(covid_ds1)) %>% 
  mutate(perc_na = (V1/tot_row_count)*100) %>% 
  mutate(in_out_na = case_when(V1/tot_row_count > 0.2 ~ 'dismiss',
                                TRUE ~ 'keep' )) %>% 
  filter(in_out_na == 'keep') 

t1 <-plyr::rename(
  t1,
  replace      = c(V1="number_of_nas", tot_row_count="total_rows", perc_na = "NA%",
                   in_out_na = "keep_or_discard"),
  warn_missing = FALSE
)

t1 <- rownames_to_column(t1, var = "column_name") ## 1st column had no name

# **Below is the table now showing columns kept for further analysis**

t1 %>% 
  kable %>%
  kable_styling("striped", full_width = F, font_size = 10) %>% ## needs library(kableExtra)
 scroll_box(width = "800px", height = "400px") ## needs library(kableExtra)


```

Column {data-width=450}
-----------------------------------------------------------------------

### Histogram of NA's - kept columns

```{r}
## Histogram of nas
t1a <- t1 %>% 
  select(everything()) %>% 
  filter(t1$number_of_nas >=0)
  
  
hist1 <-ggplot(t1a, aes(x=t1a$number_of_nas)) + geom_histogram(binwidth=500,color="black", fill="red")+
  labs(x = "x-axis -> Bins = 500", y = "y-axis -> Count")

ggplotly(hist1)%>%
  config(displayModeBar = FALSE)
```

### Tree Map showing kept columns with NA totals descending order

```{r}
## Tree map

t1%>%
  arrange(number_of_nas)%>%
  hchart(type = "treemap", hcaes(x = column_name, value = number_of_nas, color = number_of_nas))%>%
  hc_colorAxis(stops = color_stops(colors = viridis::inferno(20)))
 
```

# **HD - Categorical Summary**

##Column {data-width=650}
-----------------------------------------------------------------------

### HD - Summary of  Category factors in dataset

```{r, echo = TRUE,class.source = "small"}

## type_of_meal_plan .. room_type_reserved .. market_segment_type .. booking_status

f1 <- hotel_ds1 %>% 
  count(type_of_meal_plan)
f2 <- hotel_ds1 %>% 
  count(room_type_reserved)
f3 <- hotel_ds1 %>% 
  count(market_segment_type)
f4 <- hotel_ds1 %>% 
  count(booking_status)

## count rows in dataset
nr <- nrow(hotel_ds1)
## now get percentage proportion for each factor
f1_p <- f1 %>% 
  mutate(t_rows = nr) %>% 
  mutate(perc = (n/t_rows)*100) %>% 
  arrange(perc)
##
f2_p <- f2 %>% 
  mutate(t_rows = nr) %>% 
  mutate(perc = (n/t_rows)*100) %>% 
  arrange(perc)
##
f3_p <- f3 %>% 
  mutate(t_rows = nr) %>% 
  mutate(perc = (n/t_rows)*100) %>% 
  arrange(perc)
##
f4_p <- f4 %>% 
  mutate(t_rows = nr) %>% 
  mutate(perc = (n/t_rows)*100) %>% 
  arrange(perc)


```

##Column {data-width=350}
-----------------------------------------------------------------------

### HD - Categorical proportion in percentages

``` {r, echo = TRUE,class.source = "small"}
## type_of_meal_plan 
f1_p

## room_type_reserved 
f2_p


```

##Column {data-width=350}
-----------------------------------------------------------------------

### HD - Categorical proportion in percentages

``` {r, echo = TRUE,class.source = "small"}

## market_segment_type 
f3_p

## booking_status
f4_p

```


# **HD - Corr and Mkt Seg Analysis**

Inputs {.sidebar}
---------------------
LEGEND:

a_p_r   -> Average Price per Room  
a_y     -> Arrival Year
a_m     -> Arrival Month
a_d     -> Arrival Date
l_time  -> Lead Time
n_ad    -> Number of Adults          
n_ch    -> Number of Children        
n_wen   -> Number of Weekend Nights          
n_wn    -> Number of Week Nights         
n_pc    -> Number of Previous Cancellations        
n_pbnc  -> Number of Previous Bookings Not Cancelled
n_sr    -> Number of Special Requests
r_cps   -> Required Car Parking Space
r_g     -> Repeated Guest


```{r, echo = TRUE}


```

## Row {.tabset .tabset-fade}
-----------------------------------------------------------------------

### HD - Correlation Chart

```{r}
cor_plot1 <- hotel_ds1 %>% 
  select(market_segment_type,lead_time,avg_price_per_room,starts_with(c("no","arr","re"))) %>% 
  group_by(market_segment_type) %>%
  summarise_all(~ mean(.,na.rm = TRUE))
  

cor_plot2 <- cor_plot1 %>% 
  select(everything(),-market_segment_type) 

colnames(cor_plot2)<- c("l_time","a_p_r","n_ad","n_ch","n_wen","n_wn","n_pc","n_pbnc","n_sr",
                        "a_y", "a_m","a_d","r_cps","r_g")

cor_plot3 <- cor(cor_plot2)
##corrplot(cor_plot3,method = "pie", type = "upper", tl.cex=1.2)

##preferred plot type below
```



```{r}

ggcorrplot(cor_plot3, 
           hc.order = TRUE, 
           type = "lower",
           lab = TRUE,
           lab_size = 2)

```

### HD - Summary of  factors in dataset

```{css}
.pre {
  max-height: 800px;
  float: left;
  width: 910px;
  overflow-y: auto;
  overflow-x: auto;
}
```
``` {r, class.output = "pre",echo = FALSE,class.source = "small"}
## select all columns with numeric data

nd <- hotel_ds1 %>% 
  select_if(is_double)


describe(nd)
             
```

## Column {data-width=400}
-----------------------------------------------------------------------

### HD - Market segment percentages by Meal Plan - all %

```{r}
mat1 <- hotel_ds1 %>% 
  select(market_segment_type,type_of_meal_plan,arrival_year)

##
## GENDER RACE %
mat1a <-mat1 %>%
  tabyl(market_segment_type,type_of_meal_plan) %>%
  adorn_percentages("all") %>% ## CAN BE col, row, all
  adorn_pct_formatting(digits = 2) %>%
  adorn_ns("front")
print(mat1a)
```

### HD - Market segment percentages by Meal Plan - column %

```{r}
mat1b <-mat1 %>%
  tabyl(market_segment_type,type_of_meal_plan) %>%
  adorn_percentages("col") %>% ## CAN BE col, row, all
  adorn_pct_formatting(digits = 2) %>%
  adorn_ns("front")
print(mat1b)
```

### HD - Market segment percentages by Meal Plan - row %

```{r, echo=FALSE}
mat1c <-mat1 %>%
  tabyl(market_segment_type,type_of_meal_plan) %>%
  adorn_percentages("row") %>% ## CAN BE col, row, all
  adorn_pct_formatting(digits = 2) %>%
  adorn_ns("front")
print(mat1c)
```

# **HD - Facet grid**

```{r}

facet1 <- hotel_ds1 %>% 
  filter(wday != 'NA')

sp <- ggplot(facet1, aes(x=avg_price_per_room, y=lead_time,colour = wday)) + geom_point(shape=1)+
facet_grid(room_type_reserved ~ type_of_meal_plan) +
  theme(strip.text.x = element_text(size=6, angle=0),
        strip.text.y = element_text(size=6, face="bold"),
        strip.background = element_rect(colour="red", fill="#CCCCFF"))
ggplotly(sp)

```