Knowledge Discovery in Databases

When resolving the problems of modeling, forecasting and searching for patterns, users almost invariably have to combine different analytical treatment methods. The reason is that real-life raw data are often unsuitable for analysis. The information has to be collected from various sources, systematized, cleaned and transformed — only after that can Data Mining methods be applied.

Knowledge discovery in databases (KDD) is a process of searching for useful knowledge in raw data. KDD involves the following actions: data pre-processing, selection of informative indicators, data cleansing, use of Data Mining (DM) methods, data post-processing and interpretation of the results. Loginom contains all the necessary tools for implementing the KDD process:

• Scenario-based approach. The process of analysis involves experts developing a processing scenario. All the operations are performed using wizards. When developing a scenario the analyst can arbitrarily combine any processing mechanisms implemented in Loginom.
• Data cleansing and transformation. Loginom contains a complete set of data cleansing and pre-processing algorithms, including gap-filling, outlier editing, filtering, replacement, grouping, merging and many others.
• Data Mining. Loginom has the most powerful self-training, pattern-finding algorithms, including neural nets, decision trees, self-organizing maps, associatiщт rules, etc.
• Visualization. The program contains dozens of convenient visualizers: both general (tables, charts, OLAP etc.) and specialized ones, which take into account the specifics of the analytical treatment and facilitate the interpretation of analysis results (trees, maps, rules etc.)
• Integration. Loginom supports data exchange with various data sources and receivers, from files and office applications to commercial DBMS and OLTP systems.

Due to the implementation of the KDD approach, Loginom enables users to resolve the problem of expert knowledge formalization and replication. The end-users do not have to think about how the data are obtained; the system will automatically extract the data necessary for analysis, run the prepared scenario and display the results in the form most convenient for the end-user.