Miner Feature List

What's New in Miner

Association Rules

Association Rules help uncover relationships between variables in large data sets, most commonly to analyze customer behavior such as purchasing patterns (Market Basket Analysis), but also in many other areas, such as web site usage.

Spotfire Integration

In addition to new nodes to read and write Spotfire Text Data files, Miner now provides examples of using Spotfire Professional to visualize, explore & share model results. The Spotfire platform makes it easier to communicate results with business professionals in your organization and to provide them the tools they need to make better decisions based on the insights from your models.

New Deployment and Integration Options

Custom Java & C++ nodes: Spotfire Miner now supports custom nodes written using Java and C++.

Remote Script Execution: Leveraging the new functionality in Spotfire S+, Miner workflows can execute S+ scripts remotely on Statistics Services to offload and distribute intensive jobs. The Spotfire Miner workflow interface provides a convenient way to organize and track the progress of these jobs.

Global Worksheet Parameters: Spotfire Miner users can now set global worksheet parameters as a property of a workflow. These parameters can be accessed by interactive and batch applications, and open up new flexibility and reusability for workflows.

Data Import/Export and Data Preparation Enhancements

New Data File Types: Spotfire Miner provides nodes to access new data formats, including Spotfire Text Data, Microsoft Excel 2007, Microsoft Access 2007, and Matlab 7 data files.

JDBC: Spotfire Miner an also import and export to JDBC using the sjdbc library, opening up many new data sources for analysis.

Recode Values Node: Handling and preparing data from multiple sources is now easier, allowing you to change the values in a column to a new value, including renaming the levels of a categorical variable.

Improved graphics tools such as a new trellis hexbin plot and hexbin matrix, and the ability to create charts without the need for sampling.

Extended file format support including support for 64-bit SAS® and compressed SAS and new report and graphics output formats

The S+ Script Node and over 20 charting nodes are now included, no separate license of S+ required

Miner Feature List

Visual Workflow Environment

• Create self-documenting visual programs
• Intuitive drag-and-drop interface
• Link nodes together to describe analytic process
• On-screen annotations
• Node-level change-tracking for multi-user collaboration
• Visual confirmation of validity and caching
• Save and share worksheets as templates for best practices
• Export worksheet image to a file

• Delimited ASCII files
• Fixed format ASCII
• Data dictionary support
• SAS®, SPSS®, Excel® & many other flat file formats
• ODBC access to compliant databases (Windows®)
• Native access Oracle®, DB2, Microsoft® SQL Server, Sybase

Data Manipulation

• Powerful sampling, including stratified methods
• Row: Aggregate, Append, Filter, Partition, Sample, Shuffle, Sort, Stack, and Unstack
• Column: Bin, Create, Filter, Join, Modify, Reorder, Transpose and Normalize
• Automatically bin continuous variables
• Continuous, date, categorical and string data types
• Create or modify columns and filter rows using powerful expression language

Data Cleaning

• Detect and repair missing values with variance-preserving methods
  
• Detect duplicates
  
• Missing value handling: drop, replace, impute and last observation carried forward
  
• Detect multi-dimensional outliers with leading-edge robust methods
  

Exploratory Data Analysis and Visualization

• Trellis graphics quickly show structure of high-dimension data
  
• Univariate descriptive statistics, plus Correlation and Covariance calculations
  
• Table views and Visual Crosstabs rapidly slice and dice data
  
• Compare datasets for validation purposes
  
• 1-D Charts: Pie, Bar, Column, Dot, Histogram, Boxplot
  
• 2-D Charts: Scatterplot, Boxplot, Strip plot, Quantile-Quantile, Density
  
• Hexagonal Binning chart to view relationships between variables of very large data sets
  
• 3-D Charts: Contour, Level plot, Surface plot, Cloud plot
  
• Multivariate charts: Multiple 2-D plot, Scatterplot matrix, Hexbin Matrix, Parallel plot
  
• Time series charts: Line plot, High-Low plot, Stacked Bar plot
  

Model Types, Algorithms and Visualizers

• Prediction and classification outcome models with basic and advanced model options
  
• Highly scalable algorithms: train models on very large data sets without the need for sampling or aggregation
  
• Decision trees for classification and regression with single-tree or ensemble techniques using Block Model Averaging™; K-Fold cross-validation, plus Gini and Entropy splitting rules
  
• Linear and logistic regression implemented as QR decomposition with Householder transformations
  
• Neural Networks with Multi-layer perceptrons
  
• Neural Network training methods: Resilient Propagation, Quick Propagation, Delta-Bar-Delta, Conjugate Gradient, and Online methods.
  
• Neural Networks: up to three hidden layers with user-specified number of nodes per layer
  
• Interactive Neural Network visualizer allows real-time control over learning process
  
• Naïve Bayes Classifier
  
• Principal components analysis
  
• Cox Proportional Hazard models for censored data with time-varying covariates
  
• Customer segmentation models with K-Means Clustering
  
• Collapsible tree viewer with interactive dendrogram
  
• Assess models with gain charts, lift charts, ROC charts and agreement matrices
  
• Variable importance tool for selection of the most significant variables
  
• Automatic calculation of dummy variable and interaction columns
  

Scalability

• All components operate out-of memory and in-memory
  
• Unique "Pipeline Architecture" moves data in blocks through processing components
  
• Classical incremental techniques
  
• Block Model Averaging™ techniques
  
• Tailor size of blocks to optimize use of computing resources
  
• Automatic and manual control of caching to balance quick response with massive scalability
  

Extensibility

• Compound nodes: create an entire process within a single node
  
• Create new nodes using S programming language
  
• Complete access to all S-PLUS 8 Enterprise Developer functions and libraries through S programming language
  
• Create custom predictive models, charts and reports
  
• Create and share user libraries of custom nodes
  
• Manage multiple custom libraries
  

Deployment and Scoring

• Web-ready graphical reports
  
• HTML. PDF, PostScript and RTF model summary exports
  
• Non-interactive batch execution of all components*
  
• Model ports support automatically-updating scoring components
  
• Score custom predictive models created using S-PLUS on very large databases
  
• Predictive Model Markup Language (PMML) model import and export
  
• Generate C code for run-time model scoring*
  
  Note: * Requires Statistics Services