Advances in Data Mining,
Reasoning and Problem Solving


Boris Kovalerchuk
Jim Schwing


Back Next

    A typical example of goal 1 is the animation of a known algorithm for a novice. Here the intention is to show visually the algorithmís steps. A different situation arises relative to goal 2 when we use animation to discover properties of an algorithm visually such as the number of loops and the amount of space required for the task. If the animation tool permits viewing loops and space used, then it can serve as a visual problem-solving tool. Thus, both tasks might use the same technique. This observation shows that the essence of the transition from goal 1 to goal 2 is not the visual technique itself (e.g. animation). The essence of achieving goal 2 is in matching a decision-making/problem-solving task with a visual technique. The same type of matching may be required to convince a decision maker. Note that it is likely that a simple animation technique appropriate for the novice would not be sufficient for an advanced visual decision-making on algorithm efficiency by an experienced analyst. On the other hand, there are also situations, which show that after a solution is discovered through the use of sophisticated analytical and visual means a very simple visualization can be sufficient and desirable for convincing a decision maker.

    We use the term brute force visual problem solving for the approach in which every available visual and/or analytical technique is tried for the task. A task-driven approach is a better alternative and is one theme that is discussed throughout the book. It can be implemented at a variety of levels from the global decision level to subpixel sensor level. Such an approach can involve the automatic generation/selection of a visual tool based on a userís query to the automatic generation of composite icons matched to userís queries. Task driven approaches can also involve of the use of hierarchical, visual, decision-making models and the recording the analystís visual decision-making procedures.

    The book emphasizes the difference between visual decision making and visual data mining. Visual data mining discovers useful regularities visually or visualizes patterns discovered by common data mining tools such as neural networks and decision trees. We will cover this subject along with a look at spatial data mining, which combines analytical and visual mining of spatial data. Visual decision making relies heavily on visual data mining, but useful regularities are only a part of the entire decision-making process. In finance, visual data mining helps to discover market trends. Visual investment decision-making is used to produce buy/sell signals and to help select a portfolio. In medicine, visual data mining helps with patient diagnosis. Decision on a course of treatment is also heavily based on the diagnosis, but it is only a part of a treatment decision. For instance, a treatment decision should avoid a patientís allergic reactions and other negative side effects. Diagnostic data mining does not cover this issue. In defense applications, visual data mining can provide valuable intelligence clues for planning an operation. A visual decision about a military strike requires that much more information be considered, including the availability of military forces and political realities. Visual data mining is also useful in breaking drug trafficking rings, but the actual decision on these operations needs to involve more aspects of the problem.

Back Next




Table of Contents


Authors List

Links Page

Publisher Flyer

Springer (Buying Info) (Buying Info)