Machine learning simulation lab

Tune models, thresholds, and business decisions.

Retention · Random Forest · F1 67%
Decision recommendation
Lower the threshold if missing positive cases is more costly.

Scenarios

4 cases
Saved runs
No saved runs yet.

Model Console

Churn risk
Scenario objective
Prioritize accounts that are likely to churn in the next 30 days.
Cost tradeoff
False positives waste success time. False negatives lose revenue.
Model type
Classification threshold
25%
precision100%
recall50%
accuracy70%
flagged3

Prediction Results

10 records
RecordProbabilityPredictionActualOutcome
Acme
43%
Retention playWill churnTP
Northstar
22%
No actionWill churnFN
Zenith
6%
No actionNegativeTN
BrightCo
19%
No actionWill churnFN
Nova
6%
No actionNegativeTN
Atlas
34%
Retention playWill churnTP
Kinetic
10%
No actionNegativeTN
Omni
25%
Retention playWill churnTP
Pulse
9%
No actionNegativeTN
Vertex
21%
No actionWill churnFN

Confusion Matrix

evaluation
True Positive
3
False Positive
0
False Negative
3
True Negative
4

Feature Importance

Random Forest
Low usage122
Support tickets98
Late payment84
Product adoption80

Action Queue

top risk
Acme43%
Retention play
Atlas34%
Retention play
Omni25%
Retention play
Northstar22%
Monitor only
Vertex21%
Monitor only

Experiment Notes

local