Random Forest Voting Calculator

1. Prediction and Forest Settings

Prediction mode

Classification type

Voting method

Class labels

Separate labels with commas.

Random seed

Tie-breaking method

Class priority

2. Thresholds and Advanced Classification Options

Majority threshold

Minimum confidence

Minimum winning margin

Positive-class threshold

Probability temperature

Per-class thresholds JSON

Class cost multipliers JSON

Rejected prediction label

Missing prediction handling

Normalise tree weights

Apply reliability weighting

Use only OOB-marked trees

Reject uncertain predictions

Compare voting methods

Run weight sensitivity analysis

3. Tree Predictions

Tree	Predicted class	Probabilities	Confidence	Weight	Reliability	Actual	OOB	Abstain

Tree	Prediction	Weight	Variance	Actual target	OOB	Abstain

4. Import, Batch, and Utility Options

Import CSV

Classification columns: tree,class,probabilities,confidence,weight,reliability,actual,oob,abstain.

Import JSON

Use an array of tree objects matching the table fields.

Batch prediction JSON

Batch mode uses the selected classification method and current class labels.

Saved Calculation History

Time	Mode	Method	Prediction	Confidence
No saved calculations yet.

Formula Used

Hard voting selects the class receiving the most tree predictions. Weighted voting multiplies each vote by its tree weight. Soft voting averages class probabilities before selecting the largest value.

Hard vote: ŷ = mode(h₁(x), h₂(x), ..., hₜ(x))
Weighted score: Score(c) = Σ wᵢ I(hᵢ(x) = c)
Soft probability: P(c) = (1/T) Σ Pᵢ(c)
Regression mean: ŷ = (1/T) Σ hᵢ(x)
Weighted regression: ŷ = Σ(wᵢhᵢ(x)) / Σwᵢ

How to Use

Select classification or regression mode.
Choose the ensemble voting method.
Enter each tree prediction and optional evidence.
Adjust thresholds, weights, reliability, and tie rules.
Submit the form and inspect every diagnostic.

Example Data

Tree	Class	Probabilities	Weight
Tree 1	Positive	{"Negative":0.20,"Positive":0.80}	1.00
Tree 2	Negative	{"Negative":0.60,"Positive":0.40}	0.80
Tree 3	Positive	{"Negative":0.30,"Positive":0.70}	1.20

Frequently Asked Questions

What is hard voting?

Hard voting counts predicted labels. The most frequent label becomes the forest prediction.

What is soft voting?

Soft voting averages class probabilities. It uses more information than labels alone.

When should weights be used?

Weights help when some trees are more reliable. Larger weights increase their influence.

How are ties resolved?

Ties use your selected rule. Probability, weight, order, priority, or randomness can decide.

What does vote entropy show?

Entropy measures uncertainty across class votes. Higher entropy suggests stronger disagreement.

What is consensus ratio?

Consensus is the winning vote share. Higher values indicate broader tree agreement.

Can trees abstain?

Yes. Abstaining trees are excluded from the active ensemble calculation.

What is OOB voting?

OOB voting uses trees where a record was out-of-bag. It estimates generalisation behaviour.

How does regression voting work?

Regression forests average numeric tree predictions. Weighted averaging favours selected trees.