Complete Decisions

for comprehensive data analysis and decision making IN THE MODERN WORLD


Complete Decisions, LLC © ALL RIGHTS RESERVED.


We are investigating some of the most challenging open problems in the areas of decision making and data mining.

The decision making problems we are analyzing are related to choice problems in the presence of a finite set of alternatives when each alternative can be described in terms of a finite set of evaluative criteria. This type of problem is formally known as multi-criteria decision-making (MCDM) or multi-criteria decision analysis (MCDA)

A special area of interest is that of shared decision making (SDM).  This area of decision making is in the interface of multi-criteria decision making / analysis and the management of medical treatments / procedures.  Advances in medicine and health care have created multiple treatments for many diseases and conditions.  However, choosing the best treatment for a given case can be a complex problem if there are multiple choices.  The application of one treatment may influence the way other treatments are applied in the future or even it may exclude the use of other treatments in the future.  The connection of treatments to different adverse effects, the way individual patients perceive such adverse effects and the stochastic nature of the way diseases respond to treatments make this problem to be a unique one in the decision making filed.  We are creating decision aids for deciding which treatment is the best and how patient preferences and values may be incorporated in the decision making process.  This is also done for deciding the best screening procedure for identifying if certain diseases and conditions are present. 
Another area of research interest is that of computer-aided diagnosis (CAD) in medicine.  We are developing methods that capitalize on the availability of large amounts of data to design cost-efficient diagnostic systems.  Such systems consider data in a step-wise manner that aims at reaching a diagnostic decision by first using easily obtainable data and if a decision is not made with high confidence, then to gradually increase the inclusion of data that are more expensive to obtain.  A related problem is how to optimally balance false-positive, false-negative, and undecided cases when using data for diagnostic purposes.  

Regarding data mining we formulate the problem of maximum accuracy as an optimization problem which aims at minimizing the weighted sum of all the types of errors involved when making classification decisions.  We approach solutions to this problem by employing advanced search strategies for finding optimal or semi-optimal results.  We also exploit properties in the data that have a very high potential to lead to excellent classification results when such properties are identified and utilized properly.