R Statistics Cookbook

There's more...

The number of anomalies can be adjusted using two parameters: the alpha that controls the amplitude of what is considered normal. If a value is outside those normal values (for the remainder, this is after the trend and seasonality have been removed), it is then considered to be an anomaly.

If the alpha is reduced to 0.1, we will be lowering the threshold, and more anomalies will appear. On the other hand, we can set max_anoms=0.02 meaning that, at most, 2% of the data could be considered to be anomalies:

results_anomalies = tibbletime::as_tbl_time(currency_sales,indice_tiempo) %>%time_decompose(promedio_diario, merge = TRUE) %>% anomalize(remainder,alpha=0.10,max_anoms=0.02) %>% time_recompose() results_anomalies %>% plot_anomaly_decomposition(ncol=3,alpha_dots ...