Fig. 1

Effect of the percent of mixed segments on classifier accuracy. Bars indicate 95% confidence intervals and solid lines the average. Simulation conditions—pure-pure (blue): train and test with pure, single-behaviour segments only; pure-mixed (orange): train with pure only and test with mixed and pure; mixed-pure (green): train with mixed and pure and test with pure only; mixed-mixed (red): train and test with mixed and pure segments. The relevant comparisons are between conditions (lines) of same test type: mixed-mixed (red) versus pure-mixed (orange), and mixed-pure (green) versus pure-pure (blue)