Data Mining

Obviously huge data arrays accumulated in organizations contain much useful information which can and should be used for optimizing the workflow. To do so, it is necessary to summarize past experience, find regularities, extract rules and use this knowledge in the management process. Simple analytical methods such as visualization can fail to help in this case. They provide the answer to the question "what has happened", but we need to answer a different question: "what should be done". To do so, we need modeling mechanisms that are capable of finding non-trivial regularities in huge amounts of data — data mining systems.

Data Mining is the process used for searching raw data for previously unknown, non-trivial, practically useful and interpretable knowledge that is required for making decisions in various spheres of activity. Loginom has all the necessary tools for data mining. It contains the most advanced modeling algorithms as well as mechanisms covering the whole analytical treatment cycle (data pre-processing, modeling, appraisal of results, and implementation of results in the business process):

• Self-training algorithms and machine learning. Loginom has the most advanced adaptive modeling algorithms: decision trees, neural networks, self-organizing maps, association rules, etc. They are very easy to use. The analyst only has to formulate conjectures about a possible dependence, and the system automatically creates models using the available data.
• Time series analysis. Loginom contains algorithms used for finding seasonality, trends and autocorrelation dependences. Such models are most commonly used for resolving time series forecasting problems.
• Deviation analysis. After modeling, the user can find deviations or maverick cases. Such mechanisms allow the user to automatically detect events which need additional attention and to find patterns which do not fall under common regularities.
• Model quality assessment. The system contains integrated mechanisms for assessing model quality. They are used for comparing the results of modeling based on formal quality criteria, as well as on expert knowledge.
• Visualization. Loginom has numerous built-in specialized visualization methods designed for different Data Mining algorithms. The easy-to-use representation mechanisms considerably simplify the process of results interpretation and make the results of analysis more trustworthy for experts.

The use of data mining methods is in fact the only practical way to make use of the accumulated information; otherwise, the collected data will simply be a dead weight. Data mining allows knowledge to be extracted from data and converted into competitive advantages through performing high-quality analysis, more accurately identifying the target markets, predicting developments, managing risks and so on.