Douglas A. Talbert

Oral quinolones in hospitalized patients: an evaluation of a computerized decision support intervention*

Objective.  To determine whether a computerized decision support system could increase the proportion of oral quinolone antibiotic orders placed for hospitalized patients.

An empirical analysis of techniques for constructing and searching k-dimensional trees

fast computers with large memories have lessened the demand for program efficiency, but applications such as browsing and searching very large databases often have rate- limiting constraints and therefore benefit greatly from improvements in efficiency. This paper empirically evaluates several variants of a common k-dimensional tree technique to demonstrate how different algorithm options influence search cost for nearest neighbors.

A machine learning and data mining framework to enable evolutionary improvement in trauma triage

Trauma triage seeks to match injured patients with appropriate healthcare resources. Mistriage can be costly both in terms of money and lives. This paper proposes and evaluates a comprehensive model that uses both machine learning and data mining to support the process of trauma triage. The proposed model is more dynamic and adaptive than the typical guideline-based approach, and it incorporates a computer-assisted feedback loop to support clinician efforts to improve triage accuracy. This paper uses three years of retrospective data to compare multiple machine learning algorithms to the current standard triage decision guidelines. Then, the triage classifications from one of those experiments are used as input to demonstrate the potential of our data mining algorithm to provide a mapping between patient type and classifier performance.

Prefix and Suffix Sequential Pattern Mining

Sequential pattern mining is a challenging problem that has received much attention in the past few decades. The mining of large sequential databases can be very time consuming and produces a large number of unrelated patterns that must be evaluated. In this paper, we explore the problems of frequent prefix, prefix-closed, and prefix-maximal pattern mining along with their suffix variants. By constraining the pattern mining task, we are able to reduce the mining time required while obtaining patterns of interest. We introduce notations related to prefix/suffix sequential pattern mining while providing theorems and proofs that are key to our proposed algorithms. We show that the use of projected databases can greatly reduce the time required to mine the complete set of frequent prefix/suffix patterns, prefix/suffix-closed patterns, and prefix/suffix-maximal patterns. Theoretical analysis shows that our approach is better than the current existing approach, and empirical analysis on various datasets is used to support these conclusions.

Complex Patterns in Dynamic Attributed Graphs

In recent years, there has been huge growth in the amount of graph data generated from various sources. These types of data are often represented by vertices and edges in a graph with real-valued attributes, topological properties, and temporal information associated with the vertices. Until recently, most pattern mining techniques focus solely on vertex attributes, topological properties, or a combination of these in a static sense; mining attribute and topological changes simultaneously over time has largely been overlooked. In this work-in-progress paper, we propose to extend an existing state-of-the-art technique to mine for patterns in dynamic attributed graphs which appear to trigger changes in attribute values.

A Review of the Fourteenth International Conference on Machine Learning

We briefly review each paper of the Fourteenth International Conference on Machine Learning, along with some general observations on the conference as a whole. The major topics of papers include data reduction, feature selection, ensembles of classifiers, natural language learning, text categorization, inductive logic programming, stochastic models, and reinforcement learning.