Plenary Lecture

Classification with Incomplete Information

Prof. Amaury A. Caballero
Florida International University
Miami, Florida, USA

Abstract: In many different fields like finance, business, pattern recognition, communication and many other applications, analysts are often faced with the task of classifying items based on historical or measured data. A major difficulty faced by such analysts is that the data to be classified can often be quite complex, with numerous interrelated variables, or incomplete. The time and effort required to develop a model to solve accurately such classification problems can be significant.
There could be three major categories that affect directly the performance of the classifier:
• Ambiguous class labels in the sample data set.
• Values corrupted by noise or not enough precise sensor measurements
• Missing values in the incoming information.
Many methods exist for solving the problem:
• Imputation techniques
• Factorial analysis.
• Decision tree methods
• Rule-based methods, fuzzy logic
• Neural networks
• Bayesian and Dependency Networks
The most important characteristic of a classifier is its generalization ability, permitting to produce decisions based on data not previously seen during the training process. The use of neural networks and fuzzy logic give the analyst a powerful tool for solving the proposed task. The work is focused on analyzing the advantages of these methods, from the point of view of their simplicity and time consuming.