Icon

Interpretable ML - Binary Classifier Inspector Node

This is an overview of the Binary Classification Inspector node in Knime. This node let's you visually explore the performance metrics and also provides an easy way to change probability thresholds and see how the performance metrics change as you change the cutoffs.

Data Reading Data PartioningWe split the data intotraining and test sets usingStratified sampling on theChurn response variable TrainingTrain the models using the training data. Webuild models using:Gradien Boosted TreeDecision TreeRandom Forest PredictionUse the Models built in the previoussteps to make predictions on the testdata. Model EvaluationWe are using the Binary Classification Inspector node to visual theperformance of the predictors. This node takes the predicted probablities fromthe predictor ndoes as input. Looking at the metrics from the Inspector node,we find that all the models perform equally well. We will go with Decision TreePredictor as it is easier to interpret. Final PredictionsHere we get the final predictionsfrom the Decision Tree modeland plot the tree. OverviewThis workflow uses the customer data to predict the churn. We use tree basedpredictors, and then evaluate the performance using the Binary ClassifierInspector node.A video explanation of this available at:https://youtu.be/mahvrUUGs20 City Latitude and LongitudeChurn DataJoin on CityConvert Churnto String FactorNode 10Node 1180% for training set20% for testing set.Stratified on theChurn response varNode 21Node 27Node 28Node 30Node 31Node 34Write Predicted values to a file Table Reader CSV Reader Joiner Rule Engine DecisionTree Learner Decision TreePredictor Partitioning Binary ClassificationInspector Joiner Column Rename Column Rename Gradient BoostedTrees Learner Gradient BoostedTrees Predictor Random ForestLearner Random ForestPredictor Column Rename Joiner Decision TreeTo Image CSV Writer Data Reading Data PartioningWe split the data intotraining and test sets usingStratified sampling on theChurn response variable TrainingTrain the models using the training data. Webuild models using:Gradien Boosted TreeDecision TreeRandom Forest PredictionUse the Models built in the previoussteps to make predictions on the testdata. Model EvaluationWe are using the Binary Classification Inspector node to visual theperformance of the predictors. This node takes the predicted probablities fromthe predictor ndoes as input. Looking at the metrics from the Inspector node,we find that all the models perform equally well. We will go with Decision TreePredictor as it is easier to interpret. Final PredictionsHere we get the final predictionsfrom the Decision Tree modeland plot the tree. OverviewThis workflow uses the customer data to predict the churn. We use tree basedpredictors, and then evaluate the performance using the Binary ClassifierInspector node.A video explanation of this available at:https://youtu.be/mahvrUUGs20 City Latitude and LongitudeChurn DataJoin on CityConvert Churnto String FactorNode 10Node 1180% for training set20% for testing set.Stratified on theChurn response varNode 21Node 27Node 28Node 30Node 31Node 34Write Predicted values to a file Table Reader CSV Reader Joiner Rule Engine DecisionTree Learner Decision TreePredictor Partitioning Binary ClassificationInspector Joiner Column Rename Column Rename Gradient BoostedTrees Learner Gradient BoostedTrees Predictor Random ForestLearner Random ForestPredictor Column Rename Joiner Decision TreeTo Image CSV Writer

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: saqibali
Created at: 2019-12-13
On NodePit since: 2019-12-23
Last update: 2024-12-21
Created with KNIME version: v4.1.0
Tags: Binary Classifier Inspector NodeBinary Classifier InspectorInterpretable ML

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!