ROC-AUC & Log Loss Node

R

AI/ML/Metrics

Threshold-free evaluation of a binary classifier: area under the ROC curve, log loss and the curve points. This is the payoff for Logistic Regression producing calibrated probabilities instead of bare class labels.

ml_roc_aucml

Inputs5

Outputs4

Security exposure8/10

Packageml

Ratings

Scores range from 0 to 10. Higher values mean more impact, exposure, or operational weight.

SecurityAttack surface and exposure impact.

8/10High

PrivacyPotential sensitivity of processed data.

8/10High

PerformanceRuntime or resource pressure.

7/10High

GovernancePolicy, audit, or compliance impact.

8/10High

ReliabilityOperational stability considerations.

9/10High

CostExternal or compute cost impact.

9/10High

Input Pins

5

Input

Execution

exec_in

Execution trigger that begins the ROC evaluation

Database

Struct

database

Database connection containing the predicted probabilities and the true labels

SchemaNodeDBConnectionNodeDBConnectionNodeDBConnection1 fields

cache_keystringrequired

Schema enforced

Probabilities Column

String

probabilities_col

Column holding P(positive class) for each row, between 0 and 1 — the probability of the class named in Positive Label, NOT the probability of whichever class was predicted. No node writes this column for you: Predict in Database mode writes the predicted class only, and `confidence` is a field on the struct its Vector mode returns for one row, so build the column by looping rows through Vector mode. Convert as you go, because `confidence` is the winning class's probability: use it directly where the prediction is the positive class, and 1 - confidence elsewhere. A raw decision value or an uncalibrated score produces a meaningless curve.

Default probability

Actuals Column

String

actuals_col

Column holding the true binary label of each sample

Default target

Positive Label

String

positive_label

Value of the actuals column that counts as the positive class. Strings are compared literally, numbers numerically; booleans are always taken as-is.

Default 1

Output Pins

4

Done

Execution

exec_out

Activated once the ROC evaluation completes

AUC

Float

auc

Area under the ROC curve (0.5 = random, 1.0 = perfect)

Log Loss

Float

log_loss

Mean binary cross-entropy of the predicted probabilities (lower is better)

Result

Struct

result

AUC, log loss and the ROC curve points ordered by ascending false positive rate

SchemaRocAucResultRocAucResultRocAucResult6 fields

aucnumber:doublerequired

Area under the ROC curve. 0.5 is a coin flip, 1.0 is a perfect ranking.

format double

log_lossnumber:doublerequired

Mean binary cross-entropy of the predicted probabilities. Lower is better.

format double

curveArray<RocPoint>required

Curve points ordered by ascending false positive rate, ready to be charted

itemsRocPointarray item

A single point of the Receiver-Operating-Characteristic curve

thresholdnumber | null

Probability threshold this point was measured at. `null` for the closing (0, 0) endpoint, which corresponds to classifying everything as negative.

format double

false_positive_ratenumber:doublerequired

False positive rate at this threshold, the x axis of the curve

format double

true_positive_ratenumber:doublerequired

True positive rate (recall) at this threshold, the y axis of the curve

format double

n_samplesinteger:uintrequired

Number of samples evaluated

format uintmin 0

n_positiveinteger:uintrequired

Number of samples whose true label is positive

format uintmin 0

n_negativeinteger:uintrequired

Number of samples whose true label is negative

format uintmin 0

Node Info

Internal name

ml_roc_auc

Category

AI/ML/Metrics