consistency check and add new features

add_feature_consistency.r

# let's add information to our dataset

# function to transform angles deg into radians
deg2rad <- function(deg) {(deg * pi) / (180)}

if(!require(geosphere)){
  install.packages("geosphere")
  library("geosphere")
}

# path to the file that generated the parsing_geolife_dataset.r script
perc_csv <- "dataset_raw.csv"

dati <- read.csv(perc_csv, header = TRUE, sep =",", quote = "\"", dec = ".")

# we check if there are null values
print(colSums(is.na(dati)))

# we remove the file that generates this script
file.remove("dataset_with_add_features.csv")

# number of tuples in the dataset
print(nrow(dati))
# we delete lines which have latitude > 90 0 <-90
# as they make no sense being out of range
dati <- dati[dati$Latitude < 90,]
dati <- dati[dati$Latitude > -90,]

# we delete lines which have latitude > 180 0 <-180
# as they make no sense being out of range
dati <- dati[dati$Longitude < 180,]
dati <- dati[dati$Longitude > -180,]

# number of tuples in the dataset after removing tuples with no consistency
print(nrow(dati))
# we create an array of true or false because R is faster than evaluating conditions
# we compare id_user and id_perc and label if all are equal to the previous line 
# is true and false when they are different it indicates the change of path
cond <- c(FALSE, (dati$Id_user[-nrow(dati)] == dati$Id_user[-1]) & (dati$Id_perc[-nrow(dati)] == dati$Id_perc[-1]) & (dati$Label[-nrow(dati)] == dati$Label[-1]))

# I copy the columns in support array to optimize the execution time
longitudine <- dati$Longitude
latitudine <- dati$Latitude
date_time <- dati$Date_Time

# new feature
dati$distance <- 0
dati$vel <- 0
dati$delta_time <- 0
dati$angle <- 0

distance <- dati$distance
vel <- dati$vel
delta_time <- dati$delta_time
angle <- dati$angle

for(i_row in 1:nrow(dati))
{
  if(i_row %% 10000 == 0)
  {
    # print for debug
    print(i_row)
  }
  if(cond[i_row])
  {
    # if the previous line and the current line have id_user and id_perc and label the same
    
    # we calculate the distance in meters between the previous point and the current point
    distance[i_row] <- distGeo(c(longitudine[i_row-1], latitudine[i_row-1]), c(longitudine[i_row], latitudine[i_row]))
    # we calculate in seconds the time between the previous point and the next
    delta_time[i_row] <- as.numeric(difftime(date_time[i_row], date_time[i_row-1], units = "secs"))
    # if the delta time or delta distance are at 0 I set the speed to 0
    if(distance[i_row] == 0 | delta_time[i_row] == 0)
    {
      vel[i_row] = 0
    }
    else
    {
      # calculate speed
      vel[i_row] <- distance[i_row]/delta_time[i_row]
    }
    # I calculate the angle between north and two coordinates
    bearing <- atan2(sin(deg2rad(longitudine[i_row]) - deg2rad(longitudine[i_row-1])) * cos(deg2rad(latitudine[i_row])),
                     cos(deg2rad(latitudine[i_row-1])) * sin(deg2rad(latitudine[i_row])) - sin(deg2rad(latitudine[i_row-1]))
                     * cos(deg2rad(latitudine[i_row]))
                     * cos(deg2rad(longitudine[i_row]) - deg2rad(longitudine[i_row-1])))
    bearing = bearing + 2.0 * pi
    while(bearing > 2.0 * pi)
    {
      bearing = bearing - 2.0 * pi
    }
    angle[i_row] <- bearing
  }
}

# I insert the new calculated values in the dataframe
dati$distance <- distance
dati$vel <- vel
dati$delta_time <- delta_time
dati$angle <- angle

# save the new CSV
write.csv(dati,file="dataset_with_add_features.csv" ,row.names=FALSE)