Creates tidy, discrete data from the hessi solar flares data. Also creates five random samples (KEVIN)

solar_flares_tidy.R

install.packages('lubridate')
install.packages('RCT')
install.packages('RWeka')

solarFlares <- read.csv("hessi.solar.flare.up_to_2018.csv")

# install lubridate, which can be used to convert HH:MM:SS to seconds
library(lubridate)

##### Remove instances where not solar flare #####
# instances of NS (not solar event) or PS (potential solar event) are when
# position and radial data are not collected (which is bad for analysis)
solarFlares = subset(solarFlares, x.pos.asec != 0 & y.pos.asec != 0 & radial != 0)

##### Convert timestamps to seconds #####

#add a 'start_time_seconds' column by converting the timestamp
# in the 'start.time' and 'start.date' column to seconds
# (seconds since January 1, 1970)
solarFlares$start_seconds = with(solarFlares, (as.numeric(as.Date(start.date)) * 24 * 60 * 60) + period_to_seconds(hms(start.time)))

get_next_time_seconds = function(start_date, start_hms, next_hms) {
  table_start_date_seconds = (as.numeric(as.Date(start_date)) * 24 * 60 * 60);
  table_start_time_seconds = period_to_seconds(hms(start_hms));
  table_next_time_seconds = period_to_seconds(hms(next_hms));
  
  # if next_seconds is less than start_seconds, it is the next day
  # so add an entire day of seconds to next_seconds
  # (because next seconds is on the day after start_date)
  for(i in 1:length(table_next_time_seconds)) {
    next_time_seconds = table_next_time_seconds[[i]];
    start_time_second = table_start_time_seconds[[i]];
    if (next_time_seconds < start_time_second) {
      table_next_time_seconds[[i]] = next_time_seconds + (24 * 60 * 60);
    }
  }
  
  # return next_seconds + the start date in seconds
  return(table_next_time_seconds + table_start_date_seconds)
}

# convert peak time to seconds
solarFlares$peak_seconds =  get_next_time_seconds(
                              solarFlares$start.date,
                              solarFlares$start.time,
                              solarFlares$peak
                            )

# convert end time to seconds
solarFlares$end_seconds = get_next_time_seconds(
                            solarFlares$start.date,
                            solarFlares$start.time,
                            solarFlares$end
                          )

##### Remove undesired columns #####
library(dplyr)
solarFlares <- solarFlares %>% select(-one_of('start.date', 'start.time', 'peak', 'end')) 

##### Rename existing columns #####
library(dplyr)
solarFlares <- rename(solarFlares, duration_seconds = duration.s)
solarFlares <- rename(solarFlares, total_photons = total.counts)
solarFlares <- rename(solarFlares, position_arc_seconds.x = x.pos.asec)
solarFlares <- rename(solarFlares, position_arc_seconds.y = y.pos.asec)
solarFlares <- rename(solarFlares, sun_region = active.region.ar)
solarFlares <- rename(solarFlares, peak_max_photons = peak.c.s)
solarFlares <- rename(solarFlares, highest_energy_discrete = energy.kev)

##### Determine time to peak #####
solarFlares$seconds_to_peak <- with(solarFlares, peak_seconds - start_seconds)

##### Discretize seconds #####
library(RCT)
solarFlares$seconds_to_peak_discrete <- ntile_label(solarFlares$seconds_to_peak, 10)
solarFlares$duration_discrete <- ntile_label(solarFlares$duration_seconds, 10)
solarFlares$peak_max_photons_discrete <- ntile_label(solarFlares$peak_max_photons, 10)
solarFlares$total_photons_discrete <- ntile_label(solarFlares$total_photons, 10)

##### Export changes to csv and arff #####
write.csv(solarFlares, 'solar_flares_before_2018-03-03.csv')
library(RWeka)
write.arff(solarFlares, 'solar_flares_before_2018-03-03.arff')

##### Create five random samples #####
set.seed(11)
k <- sample_n(solarFlares, 25000)
write.arff(k, 'sample_k.arff')
set.seed(5)
e <- sample_n(solarFlares, 25000)
write.arff(e, 'sample_e.arff')
set.seed(22)
v <- sample_n(solarFlares, 25000)
write.arff(v, 'sample_v.arff')
set.seed(9)
i <- sample_n(solarFlares, 25000)
write.arff(i, 'sample_i.arff')
set.seed(14)
n <- sample_n(solarFlares, 25000)
write.arff(n, 'sample_n.arff')

@medenton