- TargetColumn
- Specify the name of the input table column that contains the time series data.
- Accumulate
- Specify the names of the input table columns to copy to the output table.Tip: To identify change points in the output table, specify the columns that appear in partition_exp and order_by_exp.
- SegmentationMethod
- [Optional] Specify the segmentation method:
'normal_distribution' (Default) In each segment, data is in normal distribution. 'linear_regression' In each segment, data is in linear regression. - SearchMethod
- [Optional] Specify the search method, binary search.
- MaxChangeNum
- [Optional] Specify the maximum number of change points to detect.
- Cost
- [Optional] Specify the penalty function, which is used to avoid over-fitting:
Option Condition for Change Point Existence Condition for Normal Distribution and Linear Regression 'BIC' (Default) ln(L1)−ln(L0) > (p1-p0)*ln(n)/2 (p1-p0)*ln(n)/2 = ln(n) 'AIC' ln(L1)−ln(L0) > p1-p0 p1-p0 = 2
threshold, a DOUBLE PRECISION value Function compares specified value to ln(L1)−ln(L0).
L1 and L0 are the maximum likelihood estimation of hypotheses H1 and H0.
For normal distribution, the definition of Log(L1 ) and Log(L0) are in ChangePointDetection (ML Engine). p is the number of additional parameters introduced by adding a change point. p1 and p0 represent this parameter in hypotheses H1 and H0, respectively.
- OutputType
- [Optional] Specify the output table columns. See ChangePointDetection Output.
- Granularity
- Specify the difference between the indexes of consecutive candidate change points, an INTEGER value greater than or equal to 1.