Data Mining

Deep Data Diver

DEEP DATA DIVERT - NOVEL DM TECHNOLOGY

The data analysis systems, referring to the area of Data Mining and KDD (Knowledge Discovery in Databases) include if-then rules search software as an important component. Systems like that help to solve problems of forecasting, classification, pattern recognition, database segmentation, extraction of the "hidden" knowledge out of data, data interpretation, finding associations in database, etc. If-then rules search methods have minimum requirements to a data type and can be applied to processing heterogeneous information. Their results are transparent for perception. But there are any unsolved problems of if-then rules search in data: the problem of attribute segmentation, the combinatorial problem, the lack of criteria for estimating certain rules, the problem of false laws in multidimensional data, etc.

DEEP DATA DIVERT SYSTEM

We have developed the Deep Data DiverT system which claims to solve some of the existing problems. The system uses a new principle and the technology of logical laws search in data. The principle is based upon the concepts of special local geometry. In this geometry each multidimensional object exists in its own local space of events with its individual metrics. The characteristics of the local spaces give geometrical interpretation to the combinatory problem of the logical laws search. The technology of this search is based upon the modified tools of linear algebra. Also, the procedure of data self-organization and the informational structural resonance effect are used.

The Deep Data DiverT has the following major characteristics:

1. Finding "the best" if-then rules (the most complete ones with the accuracy provided) for each database record
2. Constructing and testing data classifiers on the if-then rules basis
3. Constructing "fuzzy" if then rules
4. Developing dendrograms and exploring the metastructure of the ruleset
5. Linear dependency of the search algorithm working time on the data volume
6. No limits for a data type
7. Working with any amount of blanks in data
8. Working with "polluted" data
9. Using the "data + noise" method which helps to find out the stability laws in data
10. Finding non-periodical complex patterns in numerical and symbol sequences
11. Possibility of parallel processing of the if-then rules search process

Deep Data Diver v. 1.01 runs on a true 32-bit operating system - Windows 95/98/2000/XP/NT. A minimum of 128 MB of RAM is required to run Deep Data Diver with "average" performance.

Created by MaxMaster, 2003-2004