library(dplyr)
  library(ggplot2)
  library(car)
  library(xts)
  library(scales)
  library(gridExtra)

  data_h <- read.csv("C:\\Users\\migue\\Desktop\\THEFUTURE\\DM1\\Assignment\\hour.csv", header=T)
  data_h
  data_d <- read.csv("C:\\Users\\migue\\Desktop\\THEFUTURE\\DM1\\Assignment\\day.csv", header=T)


  ##Seasons were represented by numbers in the original dataset
  data_d$season <- factor(data_d$season, levels = c(1, 2, 3 ,4), labels = c("spring", "summer", "fall", "winter"))
  data_h$season <- factor(data_h$season, levels = c(1, 2, 3 ,4), labels = c("spring", "summer", "fall", "winter"))

  ##The weathersit variable describes the "quality" of the weather" with numbers:
  ##- 1: Clear, Few clouds, Partly cloudy, Partly cloudy
  ##- 2: Mist + Cloudy, Mist + Broken clouds, Mist + Few clouds, Mist
  ##- 3: Light Snow, Light Rain + Thunderstorm + Scattered clouds, Light Rain + Scattered clouds
  ##- 4: Heavy Rain + Ice Pallets + Thunderstorm + Mist, Snow + Fog
  ##- temp : Normalized temperature in Celsius. The values are derived
  ##TODO: should we change this to a nominal variable?

  ##Converting weathersit to a factor of four levels to make the plotting of the data easier.
  data_d$weathersit <- factor(data_d$weathersit, levels = c(1,2,3,4), labels = c("1", "2", "3", "4"))
  data_h$weathersit <- factor(data_h$weathersit, levels = c(1,2,3,4), labels = c("1", "2", "3", "4"))
  ##holiday was 0 and 1. Now is nominal for clarity.
  data_d$holiday <- factor(data_d$holiday, levels = c(0,1), labels = c("work-day", "holiday"))
  data_h$holiday <- factor(data_h$holiday, levels = c(0,1), labels = c("work-day", "holiday"))

  ##Converting the dates to a timeseries.
  ##data_d$dteday <- xts(data_d$dteday, order.by=as.POSIXct(data_d$dteday))
  ##data_h$dteday <- xts(data_h$dteday, order.by=as.POSIXct(data_h$dteday))
  ##TODO check incomplete cases
  ##TODO check outliers
  ##TODO check further variables that need some kind of preprocessing

  summary(data_h)
  summary(data_d)


  ##check if total number of users is a normal distribution
  ##It doesn't seem to follow a Normal distribution because of the extreme values.
  qqPlot(data_d$cnt, main = "QQ plot of total number of users per day.")

  ##Since the total number of users is not random, let's study possible predictors.

  ##Total users by season
  ggplot(data_d, aes(x = cnt)) + geom_histogram(binwidth = 500, col="black", fill = "blue") + facet_wrap(~ season)

  ggplot(data_d, aes(x = cnt)) + geom_histogram(binwidth = 500, col="black", fill = "blue") + facet_wrap(~ holiday) + ggtitle("Distribution of total user when holiday.")
  ggplot(data_d, aes(x = cnt)) + geom_histogram(binwidth = 500, col="black", fill = "blue") + facet_wrap(~ weathersit) + ggtitle("Distribution of total users by weather conditions.") + xlab("Nº users") + ylab("Nº occcurences")

  ##Temperature graph. Can we add the medians of each variable?
  ggplot(data_d, aes(windspeed,cnt))    + stat_density_2d(aes(fill = ..level..), geom="polygon")+
    scale_fill_gradient(low="blue", high="red") +  geom_density_2d(color = "black") + ggtitle("Temperature graph between windspeed and number of users.")


  ggplot(data_d, aes(x = hum, y = cnt, shape = weathersit, colour = weathersit)) + geom_point() + ggtitle("Scatter plot between humidity levels and number of users.") + xlab("Humidity level") + ylab("Nº users")
  ggplot(data_d, aes(x = temp, y = cnt, shape = weathersit, colour = weathersit)) + geom_point() + ggtitle("Scatter plot between temperature measures and number of users.") + xlab("Temperature (ºC)") + ylab("Nº users")

  ggplot(data_d, aes(x = as.Date(dteday) , y = cnt, shape = weathersit, colour = weathersit)) + geom_point() + ggtitle("Scatter plotd of the total number of users throughout the days.") + xlab("Date (dd-mm-yy)") + ylab("Nº users") + scale_x_date(labels = date_format("%d-%m-%y"))



  ##data_y

  ggplot(data_d , aes(x = as.Date(dteday), y = cnt, colour = format(as.Date(data_d$dteday), "%Y"))) + geom_point(alpha=0.5) + geom_smooth(span = 0.2) +
    scale_color_manual(values = c('#999999','#E69F00')) +
    theme(legend.position=c(0,1), legend.justification=c(0,1))

  ##TODO study the fact that the number of total users was overall higher in 2012. Temperature, weatherit, windspeed, humidty? Popularity?
  ##TODO at what hours there are more users? This can make the prediction of the reliability of the system better.
  ##TODO multiple boxplots of temp, weather, windsp, hum, etc...

  ##Can we split the bins in the histograms in two colors, one for the casual users, other for the registered?
  ##Check http://r-statistics.co/Top50-Ggplot2-Visualizations-MasterList-R-Code.html for insight

  ##IMPORTANT plot hours and dates https://stackoverflow.com/questions/7160565/how-to-create-a-time-scatterplot-with-r