| Efficiency |
- fast algorithms
- tailored to big data
|
| Portability |
|
| Consistent data format |
- standard spreadsheet form
- same data format for all methods
|
| Comprehensiveness |
- data preparation
- data reduction
- random sampling
- prediction methods: math, logic, and distance
- methods for classification, regression and clustering
- averaging and voting multiple solutions
|
| Scientific principles |
- see Predictive Data Mining: A Practical Guide
|
| State-of-the-art methods |
- decision rules and trees
- neural nets
- voted and averaged random samples for near-optimal results
|
| High-quality solutions |
- algorithms with successful track record for predictive performance
- prediction methods that test on independent data
|
| Extensibility
|
- programs: basic building blocks
- scripts: user can compose new methods with building blocks
- scripts: parallel execution where supported
|
| Network Computing Option
|
- Certified 100% Pure Java version of EDM
- GUI and command line
|
| Value
|
- single-user license: $25 US dollars
|