Binary Log Display Using Prediction Bands for Anomaly Detection

The patent introduces a new dimension (vertical axis representing prediction accuracy or confidence level) to the traditional time-series graph. This allows binary log data to be visualized with an additional dimension of information, where the vertical position indicates the reliability of the prediction rather than just the binary value. This dimensional transformation enables simultaneous display of both the binary state and the deviation degree without losing information.

The patent introduces prediction data as an intermediary element between the actual binary log data and the display representation. By showing both the actual binary values and the predicted values (with associated accuracy indicators), the system creates an intermediate layer that reveals deviation degrees. The prediction accuracy metric serves as a mediator that quantifies how much the actual data deviates from expected normal behavior.

The patent segments the display information into distinct visual elements: actual binary values, predicted values, and accuracy indicators. This segmentation allows each component to be represented separately, with the accuracy component providing additional information without complicating the core binary data representation. Users can easily distinguish between normal and anomalous states by looking at the accuracy segment.

The patent applies local quality by adding detailed information (prediction accuracy) only where needed - specifically at points where deviation from normal state occurs. The display maintains simple binary representation for normal data points while enriching the information locally at anomalous points with accuracy indicators, thus improving anomaly detection ease without uniformly increasing overall display complexity.