Train data ratio: 80%
Anomaly Detection finds contextual anomaly and displays it on the chart. Computes the score for anomalies in the context of that point. Anomaly Score is defined by $$ \mbox{Score}(X) = \begin{cases} \frac{ 100 }{ n } \sum _{ i=1 }^{ n }{ { |{ x }_{ i } - \hat { { x } } _{ i }| } } , & { if x_{ max } < 1 } \\ \ \frac { 100 }{ n } \sum _{ i=1 }^{ n }{ \frac { |{ x }_{ i } - \hat { { x } } _{ i }| }{ { x }_{ max }} } , & { otherwise } \end{cases} $$
RUL computes the time remaining to reach the threshold.
Accuracy measures the difference between actuals and predicted. Error is measured using mMAPE(maximum Mean Absolute Percentage Error, $0 < err. < 100$) $$ \begin{align} \mbox{Accuracy} &= 100 - Error(X) \\ \mbox{Error}(X) &= \begin{cases} \frac{ 100 }{ n } \sum _{ i=1 }^{ n }{ { |{ x }_{ i } - \hat { { x } } _{ i }| } } , & { if x_{ max } < 1 } \\ \ \frac { 100 }{ n } \sum _{ i=1 }^{ n }{ \frac { |{ x }_{ i } - \hat { { x } } _{ i }| }{ { x }_{ max }} } , & { otherwise } \end{cases} \end{align} $$
Model time is the time generating the time series forecast model.
Predicted time means the amount of time spent generating predictions with a time series prediction model that has already been generated.