Icon

01_​Semi_​Automated_​ML

Read and visualize data tables Select threshold andcalculate similarityscore Manual matching: one per loop iteration Clean up, show results,and write tables Semi-Automated MLThis workflow is for semi-automated data blending from two different datasets. Each dataset consists of unique columns and overlapping columns that appear in both datasets. The first part of the workflow represents themachine learning algorithm, which matches corresponding rows (using numeric and string distance metrics to calculate the distance between the rows from tables 1 and 2, based on the selected columns. Based on thisdistance, the domain expert can decide whether they want to trust the prediction or inspect the results and correct them if needed. This is handled in the second part of the workflow.This workflow generalizes well, as the algorithmic part serves simply as an example and can be exchanged by any ML algorithm; the interactive views can be adapted easily to different use cases. For each data pointFull matched tableFlag for manualmapping Save & View Results Group Loop Start Loop End Concatenate InteractiveManual Inspection Empty Table Switch End IF ConstantValue Column View Tables &Choose Columns Select Threshold forManual Matching Read Data Read and visualize data tables Select threshold andcalculate similarityscore Manual matching: one per loop iteration Clean up, show results,and write tables Semi-Automated MLThis workflow is for semi-automated data blending from two different datasets. Each dataset consists of unique columns and overlapping columns that appear in both datasets. The first part of the workflow represents themachine learning algorithm, which matches corresponding rows (using numeric and string distance metrics to calculate the distance between the rows from tables 1 and 2, based on the selected columns. Based on thisdistance, the domain expert can decide whether they want to trust the prediction or inspect the results and correct them if needed. This is handled in the second part of the workflow.This workflow generalizes well, as the algorithmic part serves simply as an example and can be exchanged by any ML algorithm; the interactive views can be adapted easily to different use cases. For each data pointFull matched tableFlag for manualmapping Save & View Results Group Loop Start Loop End Concatenate InteractiveManual Inspection Empty Table Switch End IF ConstantValue Column View Tables &Choose Columns Select Threshold forManual Matching Read Data

Nodes

Extensions

Links