|
BayesiaLab
 |
BayesiaLab
Bayesian Network technology at your service
|
With BayesiaLab, Bayesia gives you a complete laboratory for manipulating
Bayesian networks:
- Develop your decision models through expertise
- Ergonomic node edition panel providing:
- wizards for modality generation and naming of both Label
and Interval nodes
- various entry modes for conditional probability distribution: probabilistic,
deterministic and equation
- powerful formula editor with a complete function and operator library
(discrete & continuous probability distributions, arithmetic and
trigonometric functions, ...)
- table completion and normalization tools, cut & paste between
tables and external applications
- Constraint nodes to express constraint that hold between nodes
- Enhanced traceability and documentation thanks to hypertext comments
associated to the graph, the nodes and the arcs
- Color node and arc tagging to semantically group your variables and
probabilistic relations
- Automatic learning or updating of
your models from your data (text files and databases)
- Learning conditional probabilities for a given network
- Discovering of all the probabilistic relations that hold in your data
base (Association discovery)
- Supervised learning entirely devoted to characterizing a target variable
- Selection of the minimal subset of variables correlated to the target
variable
- Bayesian Clustering to invent new concepts
- Robust Missing value processing
- Validation tools qualifying the obtained models (confusion matrix, lift
and Roc curves)
- Quickly assimilate the represented knowledge using a set of original
analytical tools
- Strength of the probabilistic relations (arc's thickness and HTML report)
- Amount of information brought to the target node/modality
- Type of probabilistic relations
- Complete HTML analysis report of the target variable
- HTML report of the evidence set analysis
- Causal analysis (essential graphs)
- Automatic network layouting algorithms
- Use the models in interactive or batch mode
- Positive, negative and soft evidences on the variable states
- Simulation of "What-if" scenarios with probability variation
highlighting
- Adaptive questionnaires taking into account Costs and Information Gains
- Off-line Tagging of new cases contained in a file
- Robust imputation algorithm to complete data with missing values
- Introduce the temporal dimension into your models
- Compact representation of Dynamic Bayesian networks
- Time node for an explicit use of time in the equations
- Temporal simulation step by step or by period with a graphical view
of the probability evolution
- Observations file to specify the context of the scenarios
- Representation, evaluation and learning of your action policies
- Decision nodes for modeling your actions
- Quality tables associated to Decision nodes for a direct representation
of action policies
- Utility nodes to valuate the states and to associate cost/gains
to node modalities
- Reinforcement learning algorithms to automatically discover policies
that optimize the expected sum of utilities
- Complete interoperability
- Connection with your databases by using JDBC/ODBC
- SQL Interface
- Exportation of Bayesian networks, tables, equations, graphs, matrices
and reports by using simple copy&paste, as images, numerical data
and HTML texts
|
|
Presentation
Trial Version