[Ежемесячное количество браков] #R #code #metrics #time_series

marriage_analysis.R

library(rio) # загрузка данных
library(tidyverse) # манипуляции
library(forecast) # прогнозировани
library(tsibble) # ряды богатой структуры
library(stringr) # строками
library(skimr) # описательные статистики
library(lubridate) # работа с датами

# данные скачиваются с fedstat.ru
# tools - install packages: rio, tidyverse, forecast

d = import("c:/Users/teacher.MAIN.000/Downloads/data.xls")

2006:2018
paste0("y", 2006:2018)

colnames(d) = c("region", "unit", "period", paste0("y", 2006:2018))
glimpse(d)

d2 = tail(d, -2)
glimpse(d2)

d3 = mutate_at(d2, vars(y2006:y2018), as.numeric)
glimpse(d3)
skim(d3)

d4 = select(d3, -unit) %>% filter(nchar(period) < 14)
# d4 = select(d3, -unit) %>% filter(!str_detect(period, "-"))
glimpse(d4)
skim(d4)


report = group_by(d4, region) %>% summarise(n_obs = n())
table(report$n_obs)
qplot(data = report, x = n_obs)

d5 = group_by(d4, region) %>% filter(n() == 12) %>% ungroup()
glimpse(d5)

d6 = mutate(d5, month = rep(1:12, nrow(d5)/12))
glimpse(d6)

table(d6$period, d6$month)
d7 = select(d6, -period)
glimpse(d7)

d8 = gather(d7, key = "year", value = "marriage", -month, -region)
glimpse(d8)

d9 = mutate(d8, date = ymd(
  paste0(str_remove(year, "y"), "-", month, "-01")))
glimpse(d9)

d10 = select(d9, -year, -month)
glimpse(d10)

export(d10, "c:/Users/teacher.MAIN.000/Downloads/marriage.csv")

d11 = mutate(d10, date = yearmonth(date))
glimpse(d11)

mar_tsibble = as_tsibble(d11, index = date, key = id(region))
mar_russia = filter(mar_tsibble, region == "643 Россия (код по ОКСМ)")
mar_ts = as.ts(mar_russia)
mar_ts


ggtsdisplay(mar_ts)
ggseasonplot(mar_ts)

model = auto.arima(mar_ts)
model

fcst = forecast(model, h = 12)
fcst

autoplot(fcst)