pre_aggregated: Aggregation functions

In ilangurudev/approxmapR: ApproxMAP (APPROXimate Multiple Alignment Pattern)

Aggregation functions

Description

Used whenever the df to be analyzed is preaggregated, i.e. the data has already by grouped into periods (corresponding to itemsets).

Usage

aggregate_sequences(df, include_date = FALSE, multiset = FALSE, summary_stats = TRUE, output_directory = "~")

Arguments

df	A dataframe that has either 3 or 4 columns; 3 columns in the order of id, date, event if the date is not desired to be included; or 4 columns in the order of id, date, period, event if the date is to be included.
include_date	Logical indicator which controls the inclusion of the date variable in the returning data. If creating reports using the -generate_reports- function of approxmapR, then the dates will be included in the alignment_with_date output file if this argument is equal to TRUE - default value is FALSE.
multiset	Beta; Logical indicator which controls the exclusion of multiple events within the same event set.
summary_stats	Logical controlling printing of summary statistics regarding aggregation. Defaults to TRUE
output_directory	The path to where the exports should be placed.

Value

Returns a dataframe that has the properly classes dataframe

Examples

library(approxmapR)
library(tidyverse)

data("demo1")
demo1 <- data.frame(do.call("rbind", strsplit(as.character(demo1$id.date.item), ",")))
names(demo1) <- c("id", "period", "event")

# Identifying the earliest date per -id- and setting it as the -index_dt-
demo1 <- demo1 %>% group_by(id) %>% mutate(index_dt = min(as.Date(period, "%m/%d/%Y"))) 

# Creating an Index from the earliest date
demo1 <- demo1 %>%
          mutate(date = as.Date(period, "%m/%d/%Y")) %>%
          mutate(period = as.numeric(difftime(date, index_dt, units = "days"))) %>%
          select(id, period, event) %>% arrange(id, period)


# Aggregating custom aggregation frames with the following groupings:
#    [] index date will be first period (1),
#    [] the first 28 days after the index date will be grouped into weekly periods (2 - 4), and then
#    [] events which occurred on the 29th day or more from the index day will be grouped in a monthly frame (5+)
demo1 <- demo1 %>% group_by(id) %>% mutate(date = period,
                                          n_ndays7 = period / 7,
                                          period = as.integer(case_when(period == 0 ~ 1,
                                                             ceiling(n_ndays7) < 5 ~ ceiling(n_ndays7) + 1,
                                                             TRUE ~ floor(n_ndays7) + 2))
                                          ) %>% select(id, date, period, event)

# Since -demo1- has the date column, need to select only the id, period, and event columns if the dates are not
#    to be included
agg <- demo1 %>% select(id, period, event) %>% pre_aggregated()

# No need to select specific columns if the dates are desired to be included
agg <- demo1 %>% pre_aggregated(include_date = TRUE)

ilangurudev/approxmapR documentation built on March 22, 2022, 1:15 p.m.