rapids/src/features/bluetooth/bluetooth_base.R

48 lines
1.9 KiB
R
Raw Normal View History

2020-06-24 21:10:42 +02:00
library(dplyr)
library(tidyr)
filter_by_day_segment <- function(data, day_segment) {
if(day_segment %in% c("morning", "afternoon", "evening", "night"))
data <- data %>% filter(local_day_segment == day_segment)
return(data %>% group_by(local_date))
}
compute_bluetooth_feature <- function(data, feature, day_segment){
if(feature %in% c("countscans", "uniquedevices")){
data <- data %>% filter_by_day_segment(day_segment)
data <- switch(feature,
"countscans" = data %>% summarise(!!paste("bluetooth", day_segment, feature, sep = "_") := n()),
"uniquedevices" = data %>% summarise(!!paste("bluetooth", day_segment, feature, sep = "_") := n_distinct(bt_address)))
return(data)
} else if(feature == "countscansmostuniquedevice"){
# Get the most scanned device
data <- data %>% group_by(bt_address) %>%
mutate(N=n()) %>%
ungroup() %>%
filter(N == max(N))
return(data %>%
filter_by_day_segment(day_segment) %>%
summarise(!!paste("bluetooth", day_segment, feature, sep = "_") := n()))
}
}
base_bluetooth_features <- function(bluetooth_data, day_segment, requested_features){
# Output dataframe
features = data.frame(local_date = character(), stringsAsFactors = FALSE)
# The name of the features this function can compute
base_features_names <- c("countscans", "uniquedevices", "countscansmostuniquedevice")
# The subset of requested features this function can compute
features_to_compute <- intersect(base_features_names, requested_features)
for(feature_name in features_to_compute){
feature <- compute_bluetooth_feature(bluetooth_data, feature_name, day_segment)
features <- merge(features, feature, by="local_date", all = TRUE)
}
features <- features %>% mutate_at(vars(contains("countscansmostuniquedevice")), list( ~ replace_na(., 0)))
return(features)
}