The next R script shown four methodologies to classificate time series from smartwatch data. The objective is to distinguish between two states: Sleep/Awake. 1. First approximation by conditional: Steps < 4 and hour in [10,20] hours. 2. Run lenght encoding in both directions over the first approximation applying a condition: Ignore changes which last less than 1 hour. 3. Support Vector Machine (R package e1071). 4. Recursive partitioning (R package rpart).