Add Rolling Window Statistics

Compute rolling window statistics (sum, standard deviation, min, max) and rolling trend slopes for the target variable. All statistics are computed within groups if specified.

Usage

feat_rolling_dt(
  df,
  date,
  target,
  groups = NULL,
  windows = c(7, 28),
  stats = c("sum", "sd", "min", "max"),
  trend_windows = NULL
)

Arguments

df

A data frame containing time series data

date

Symbol or character naming the date column

target

Symbol or character naming the target variable column

groups

Character vector naming grouping columns

windows

Integer vector of window sizes for rolling statistics (e.g., c(7, 28, 90))

stats

Character vector specifying which statistics to compute: "sum", "sd", "min", "max"

trend_windows

Integer vector of window sizes for rolling linear trend slopes

Value

Data frame with rolling statistic features added

Details

Rolling statistics use right-aligned windows (including current observation). Trend slopes are computed by fitting y ~ x within each window. Missing values are handled with na.rm = TRUE for statistics.

Examples

if (FALSE) { # \dontrun{
# Add 7-day and 28-day rolling statistics
df_roll <- feat_rolling_dt(df, date = date, target = sales,
                            groups = "store_id",
                            windows = c(7, 28),
                            stats = c("sum", "sd", "min", "max"))

# Add rolling trend slopes
df_roll <- feat_rolling_dt(df, date = date, target = sales,
                            groups = "store_id",
                            windows = c(7, 28),
                            stats = "sum",
                            trend_windows = c(7, 14))
} # }