desc_analysis.Rmd

---
title: "Air Traffic Challenge"
output: 
  prettydoc::html_pretty:
    theme: cayman
    highlight: github
    toc: true 
    number_sections: true 
    df_print: paged
---

# Data manipulation (daily stats)
```{r,warning=FALSE, message=FALSE}
library(dplyr)
library(ggplot2)
library(sf)
library(varhandle)
library(tidyr)
library(ggplot2)
library(hrbrthemes)
library(lubridate)

twist_zrh_cleaned <-readRDS("twist_zrh_cleaned.RDS")

# flights per day, median / mean delays per day, precipitation etc.
twist_daily <-twist_zrh_cleaned %>%  
  group_by(date,start_landing) %>% 
  summarize(flights_n=n(),
            mean_delay=as.numeric(mean(diff_in_secs)), 
            median_delay=as.numeric(median(diff_in_secs)),
            precip=sum(precip),
            temp_avg=mean(temp_avg),
            precip=sum(precip),
            lightnings_hour_f=sum(lightnings_hour_f))
```

# Descriptive Analysis
```{r,warning=FALSE, message=FALSE}
# delay (mean)
ggplot(twist_daily, aes(as.Date(date), mean_delay))+
  geom_line()+
  facet_wrap(~start_landing)+
  theme_ipsum()

# delay  (median)
ggplot(twist_daily, aes(as.Date(date), median_delay))+
  geom_line()+
  geom_smooth(se=F)+
  facet_wrap(~start_landing)+
  theme_ipsum()

# flights per day
ggplot(twist_daily, aes(as.Date(date),flights_n))+
  geom_line()+
  geom_smooth(se=F)+
  facet_wrap(~start_landing)+
  theme_ipsum()+
  labs(title="Number of Flights")

# Temperature
ggplot(twist_daily, aes(as.Date(date),temp_avg))+
  geom_line()+
  geom_smooth(se=F)+
  theme_ipsum()+
  labs(title="Temperature (avg)")

```

# Visualization of Mean Delay per Flight Routes
```{r}

if (!require("pacman")) install.packages("pacman")

pacman::p_load(dplyr,sf,lwgeom,ggplot2,maps,maptools,rgeos)

#read in flight data
flight_sf <- readRDS("flight_sf.RDS")


#read in flight data
#filter flights for a single date
vizdata_del <- flight_sf %>% 
               group_by(origin_destination,start_landing) %>% 
               summarize(mean_delay=mean(diff/60),n=n())


#create a new geometry : lines connecting Zurich Airport and the destination/origin of the flight
vizdata_del$flightline <- sf::st_union(vizdata_del$geometry, st_as_sfc("POINT(8 47)",crs=st_crs("+proj=longlat +datum=WGS84 +no_defs"))) %>% 
  st_cast("LINESTRING")

#get worldmap
world1 <- sf::st_as_sf(maps::map('world', plot = FALSE, fill = TRUE))

#plot on top (new ggplot2 2.3.0 version needed!)
ggplot(vizdata_del)+
  geom_sf(aes(geometry=flightline))+
  geom_sf(data=world1)

#we can now set the "flightline"-geometry as the main geometry in our dataset
st_geometry(vizdata_del) = "flightline"

#transform worldmap and flights into spheric projection
world2 <- sf::st_transform(
  world1,
  "+proj=laea +y_0=0 +lon_0=8 +lat_0=47 +ellps=WGS84 +no_defs"
)

flights2 <- sf::st_transform(
  vizdata_del,
  "+proj=laea +y_0=0 +lon_0=8 +lat_0=47 +ellps=WGS84 +no_defs"
)  

#Plot!
ggplot() + 
  geom_sf(data=world2, color="white",size=0.2)+
  geom_sf(data=flights2 %>% filter(n>=100), aes(color=as.numeric(mean_delay), alpha=0.8),size=0.2,show.legend = "line")+
  theme_void()+
  scale_color_viridis_c(name="")+
  guides(alpha=F)+
  coord_sf(ndiscr=1000)+
  theme(plot.background = element_rect(fill="#f5f5f2"))+
  facet_wrap(~start_landing)+
  labs(title="Mean Delay per Flight Route",subtitle="Routes with >=100 Flights, 2017\n\n")


```

# Flights by Night (23:30-6:00)
```{r}

#dummy for nightflights
twist_zrh_night <- readRDS("twist_zrh_cleaned.RDS") %>% 
                        mutate(hour=hour(effective_time)+minute(effective_time)/60) %>% 
                        mutate(night=ifelse(hour<6 |hour>=23.5,1,0)) %>% 
  group_by(date,start_landing, night) %>% 
  summarize(flights_n=n())%>% 
   mutate(freq =flights_n / sum(flights_n)*100)

twist_zrh_night %>% 
  group_by(start_landing, night) %>% 
  summarize(n=sum(flights_n))

# ----

devtools::install_github("jayjacobs/ggcal")

library(ggcal)

cal <- twist_zrh_night %>%  filter(night==1)
```

# Flights between 23:30-6:00

## Number per day
```{r}
ggcal(cal$date,cal$flights_n)
```

## In proportion of daily flights
```{r}
ggcal(cal$date,cal$freq)

```

# 
```{r}

```




```{r}

```


<!-- # Modelling with xgboost --- -->
<!-- ```{r} -->

<!-- data <- twist_zrh_cleaned %>% mutate(delay=ifelse(diff_in_secs>1800,1,0)) %>%  -->
<!--   select(-geometry) -->



<!-- install.packages("xgboost", repos="http://dmlc.ml/drat/", type = "source") -->

<!-- library(xgboost) -->
<!-- library(magrittr) -->
<!-- library(dplyr) -->
<!-- library(Matrix) -->


<!-- # Partition data -->
<!-- set.seed(1234) -->
<!-- ind <- sample(2, nrow(data), replace = T, prob = c(0.8, 0.2)) -->
<!-- train <- data[ind==1,] -->
<!-- test <- data[ind==2,] -->

<!-- # Create matrix - One-Hot Encoding for Factor variables -->
<!-- trainm <- sparse.model.matrix(delay ~ .-1, data = train) -->
<!-- head(trainm) -->
<!-- train_label <- train[,"delay"] -->
<!-- train_matrix <- xgb.DMatrix(data = as.matrix(trainm),  -->
<!--                             label = train_label) -->

<!-- #  -->
<!-- testm <- sparse.model.matrix(admit~.-1, data = test) -->
<!-- test_label <- test[,"delay"] -->
<!-- test_matrix <- xgb.DMatrix(data = as.matrix(testm), label = test_label) -->

<!-- #### WORK IN PROGRESS  -->



<!-- ``` -->