Build Bayesian networks from data and/or expert opinion. Perform advanced queries and analytics.
Fast & efficient software for Bayesian networks
Automate the decision making process. Also known as influence diagrams.
Model multi-variate time series and sequences
Model hidden patterns / automate feature extraction
Guide diagnostic / troubleshooting applications.
Automatically extract important information from data
How unusual/anomalous is your current data? Can your model be trusted?
100x faster than a naive implementation. Over a decade of research.
Connect to databases, spreadsheets, No-Sql stores and even custom data sources
Missing data natively supported during both learning and prediction, and for both discrete and continuous variables
Support for both discrete & continuous variable, as well as latent variables for expressing complex distributions
Perform queries on entire data sources
How strong are the links/relationships in your model?
How susceptible are your predictions/queries to changes in evidence?
Connect to many different databases, spreadsheets and No-Sql stores
A robust and efficient pure .NET API for building intelligent applications and services
A robust and efficient pure Java API for building intelligent applications and services
Automatically evaluate more expressive models and hidden patterns using cluster analysis for latent variables
Learn the parameters of your model from data. Includes support for missing data, discrete, continuous & hybrid networks
Learn the structure (links) in a network. 5 different algorithms
Assign a weight to each row (or case). Great for historic data weighting, and learning from data with probabilities attached
How sensitive are you queries/predictions to the model parameters?
Analyze the log-likelihood with different evidence subsets
Identify anomalous data within a training data set.
Perform advanced queries/prediction such as P(X,Y[t+3],Y[t+4] | Z)
Model nodes with large numbers of parents
Nodes can contain multiple variables, for powerful modeling
Easily explore scenarios in the User Interface.
Bayes Server supports Big Data platforms & technologies such as Apache Spark & Hadoop
The Bayes Server API can be called from R. Our code center contains examples.
The Bayes Server API can be called from Python. Our code center contains examples.
The Bayes Server API can be called from Matlab. Our code center contains examples.
The Bayes Server API can be called from within Excel functions. Our code center contains examples.
Generate data from a model. Help understand/visualize your model, or generate test data.
Incrementally update (adapt) your model as new data arives.
Although continuous variables are fully supported, sometimes it can be useful to discretize. 3 discretization algorithms included.
Advanced support for creating variables from a wide variety of data sources.
Predict a variable as if it did not have evidence set on it. Great for prediction and anomaly detection.
Determine the most likely scenario (or sequence) given that you have an incomplete picture
We have built networks with over 10000 variables
Automatically adjust your model to achieve a desired goal.
Evidence can be uncertain on discrete variables
The API is so flexible you can connect to custom/bespoke data sources
Bayes Server supports distributed processing on Apache Spark
Evaluate the performance of your model
Evaluate the performance of your model. Useful when the target is rare.
Does you current data have conflicting evidence?
Navigate large numbers of predictions/queries in the user interface
4 different advanced inference algorithms
Multi-variate and conditional mutual information calculator (Discrete/continuous/hybrid).
Determine the most influential features from data
Multi-variate and conditional entropy calculator (Discrete/continuous/hybrid).
Multi-variate Kullback-Leibler calculator. (Discrete or continuous)
Reverse the direction of links in your network, while maintaining the same overall network distribution
Visualize predictions across 2 variables to easily identify patterns
Our exact inference algorithms only touch what they need, increasing performance
Evaluate the complexity of your inference problem, given specific evidence and queries
Our inference algorithms can auto propagate evidence through deterministic distributions, improving performance
Networks can contain disconnected groups of nodes/variables