Poulin-Hugin Patterns & Predictions 1.6

Category Intelligent Software>Bayesian Network Systems/Tools and Intelligent Software>Data Mining Systems/Tools

Abstract Poulin-Hugin (P-H) Patterns & Predictions is an integrated command line/GUI based automation tool for Bayesian pattern discovery and prediction modeling based on the Hugin Decision Engine (see G6G Abstract Number 20163). It implements data-driven modeling through the use of Naive Bayes models.

Products key features/capabilities include:

1) Construction of Naive Bayes, Tree-augmented Naive Bayes (TAN) and Hierarchy Models.

2) Model Update.

3) Inference (Prediction).

4) Value Information Analysis of Cases (Data files) and Variables (Data objects).

5) Sensitivity Analysis (What-If scenarios).

6) Data Parser [Flat File Databases, Text, Uniform Resource Locators (URLs)].

P-H Patterns and Predictions suite of command-line based tools allow the user or automation environment to build and use sophisticated statistical models efficiently.

The underlying technology as stated above is based on Bayesian analysis and 'Bayesian networks' in particular. Bayesian networks are derived from Bayes’ Theorem. Bayes’ Theorem allows the inferring of a future event based on the prior evidence. The theorem was discovered by Rev. Thomas Bayes (1702-1761).

The P-H Patterns and Predictions software package automates the Bayesian knowledge discovery and prediction process. This advanced tool allows the user to quickly perform extensive analysis of both structured and unstructured data. In addition, the analysis capabilities of the tool will produce rapid conclusions.

The Bayesian knowledge discovery and prediction process can be divided into three (3) subsequent steps:

(1) Data preparation and manipulation.

(2) Model construction.

(3) Inference and analysis.

The command-line driven tools included in the software package may be grouped according to the above three (3) steps:

1) Data preparation and manipulation --

The first step in any analysis is data preparation and manipulation. The tools for constructing the statistical models assume a certain data format. The goal of the data preparation and manipulation step is to organize the data to be analyzed into a data format suitable for pattern discovery and prediction.

A number of data parsing tools for formatting both structured and unstructured data are included in the package.

The tools for data preparation and manipulation can be organized into four groups: parsing data, feature extraction, case creation and generation, and pulling web pages.

The model construction tools assume data to adhere to the format of Hugin data files. If the data does Not adhere to the required format, then data may be prepared and manipulated to adhere to the required format using the tools provided with this package.

2) Model construction --

The second step is model construction. The goal of the model construction is to construct statistical models based on the data prepared in the first step.

Tools for building Naive Bayes Models, Tree-Augmented Naive Bayes Models, and Hierarchical Naive Bayes Models are also included in the package.

3) Inference and Analysis --

The third and last step is inference and analysis. The goal of inference and analysis is to perform inference in and analysis on the statistical models constructed in the second step. The inference and analysis is performed using data prepared in the first step.

Tools for computing the probability of events given observations, performing value of information analysis on both scenarios and variables, performing scenario-based sensitivity analysis, and for determining classification accuracy are included in the package.

Documentation provided with P-H Patterns and Predictions include:

1) The PHJ UI/Wizard Walkthrough Document.

2) A Patterns & Predictions User Manual 1.6.

3) A Bayesian Technology Primer.

4) Finance: Credit Risk Case Study.

5) Finance: Futures Case Study.

6) Full Technical Specification.

7) Case Study in Epidemiology -- Influenza Prediction - The manufacturers’ latest example is in epidemiological prediction.

Specifically, in this initial case study the manufacturer demonstrates the building of a predictive model for Influenza Mortality using the Patterns and Predictions™ Bayesian algorithms based tool.

The scope of this predictor currently deals with influenza as defined by the International Classification of Diseases 9th revision (ICD-9) diagnostic codes specific to the various common strains of influenza.

System Requirements

Contact manufacturer.

Manufacturer

Patterns and Predictions is a joint technology partnership. The software is a data mining/automation tool utilizing both Hugin technology, adopted by R&D departments in 25 countries, and from Poulin Holdings, LLC.


Manufacturer Web Site Poulin-Hugin Patterns & Predictions 1.6

Price Price guide

G6G Abstract Number 20206

G6G Manufacturer Number 102163