Bayesian Network Tool

Model uncertain events, conditional links, and evidence-driven outcomes confidently. Test prior and posterior changes instantly. Review clear metrics, tables, and exports for smarter decisions.

Enter Network Inputs

This tool models a three-node binary chain: A → B → C. Enter percentages from 0 to 100.

Root node label.
Middle node label.
Outcome node label.
Prior probability for the root node.
Conditional probability when A is true.
Conditional probability when A is false.
Conditional probability when B is true.
Conditional probability when B is false.
Set evidence or leave unobserved.
Additional evidence updates all marginals.
Reset

Example Data Table

Scenario P(A) P(B|A) P(B|¬A) P(C|B) P(C|¬B) Evidence Posterior P(A) Posterior P(C)
Strong positive chain 40% 75% 20% 85% 15% B=True, C=True 71.43% 100.00%
Observed final outcome only 55% 80% 35% 70% 25% C=True 64.66% 100.00%
Conflicting evidence 30% 90% 10% 60% 20% B=False, C=True 4.55% 100.00%

Formula Used

1) Joint probability for the network

For any complete state, P(A, B, C) = P(A) × P(B | A) × P(C | B). This factorization follows the directed chain structure.

2) Evidence probability

P(E) = Σ P(A, B, C) across every state matching the observed evidence. The tool sums valid states automatically.

3) Posterior update

P(X | E) = P(X, E) / P(E). The calculator normalizes matching states to produce posterior probabilities for each node.

4) Entropy change

H(p) = -p log₂(p) - (1-p) log₂(1-p). Lower posterior entropy means evidence reduced uncertainty.

How to Use This Calculator

  1. Enter custom labels for the three binary nodes.
  2. Provide the prior probability for the first node.
  3. Enter conditional percentages for the middle and final nodes.
  4. Choose observed evidence for node B, node C, or both.
  5. Click Calculate Network to update all probabilities.
  6. Review the prior versus posterior summary table.
  7. Inspect the full joint state table for exact enumeration.
  8. Download the results as CSV or PDF when needed.

8 FAQs

1) What does this tool calculate?

It computes prior, joint, and posterior probabilities for a three-node binary Bayesian network. It also shows evidence probability, entropy change, lift, and the most likely posterior state.

2) Why is the network arranged as A → B → C?

This structure keeps the model understandable while still showing dependency propagation. Evidence on later nodes can still update earlier beliefs through exact Bayesian normalization.

3) Can I leave evidence fields unknown?

Yes. Unknown evidence means the tool uses only the prior and conditional tables. Posterior results then match the original marginal probabilities.

4) What is evidence probability?

Evidence probability is the total probability of all states consistent with your observations. It tells you how likely the observed combination is under the current network.

5) What does lift mean in the summary table?

Lift compares posterior probability to prior probability. A value above 1 means evidence increased belief. A value below 1 means evidence reduced belief.

6) Why can evidence become impossible?

If your probabilities assign zero chance to every state matching the chosen evidence, normalization cannot occur. The tool warns you when that configuration is impossible.

7) Should I enter percentages or decimals?

Enter percentages from 0 to 100. The calculator converts them internally into probability values between 0 and 1 before performing Bayesian calculations.

8) What do the CSV and PDF exports include?

The exports include evidence details, the prior versus posterior summary, and the complete joint state table. This makes review, reporting, and sharing much easier.

Related Calculators

linear model calculatorpolynomial regression modelnash equilibrium calculatorhazard function calculatormodel selection aicpetri net simulatorgravity model calculatornonlinear regression tooldiscrete event simulator

Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.