Fig. 1From: Multiclass anomaly detection in imbalanced structural health monitoring data using convolutional neural networkComparison between the number of data points needed between (a) non-overlapping hourly time shift and (b) shortened intervalBack to article page