Applications of machine learning: matching problems to tasks and methods

The terminology of Machine Learning and Data Mining methods does not always allow a simple match between practical problems and methods. While some problems look similar from the user's point of view, but require different methods to be solved, some others look very different, yet they can be solved by applying the same methods and tools. Choosing appropriate Machine Learning methods for problem solving in practice is therefore largely a matter of experience and it is not realistic to expect a simple look-up table with matches between problems and methods. However, some guidelines can be given and a collection that summarizes other people's experience can also be helpful. A small number of definitions characterize the tasks that are performed by a large proportion of methods. Most of the variation in methods is concerned with differences in data types and algorithmic aspects of methods. In this paper, we summarize the main task types and illustrate how a wide variety of practical problems are formulated in terms of these tasks. The match between problems and tasks is illustrated with a collection of example applications with the aim of helping to express new practical problems as Machine Learning tasks. Some tasks can be decomposed into subtasks, allowing a wider variety of matches between practical problems and (combinations of) methods. We review the main principles for choosing between alternatives and illustrate this with a large collection of applications. We believe that this provides some guidelines.

Estimating the Costs of VA Ambulatory Care

This article reports how we matched Common Procedure Terminology (CPT)codes with Medicare payment rates and aggregate Veterans Affairs (VA)budget data to estimate the costs of every VA ambulatory encounter. Converting CPT codes to encounter-level costs was more complex than a simple match of Medicare reimbursements to CPT codes. About 40 percent of the CPT codes used in VA, representing about 20 percent of procedures, did not have a Medicare payment rate and required other cost estimates. Reconciling aggregated estimated costs to the VA budget allocations for outpatient care produced final VA cost estimates that were lower than projected Medicare reimbursements. The methods used to estimate costs for encounters could be replicated for other settings. They are potentially useful for any system that does not generate billing data, when CPT codes are simpler to collect than billing data, or when there is a need to standardize cost estimates across data sources.

Genetic distances in soybean based on RAPD markers

Four methods were applied to determine pairwise genetic distances among five soybean genotypes which are potential genitors for a mapping population. Additionally, individual plants from the most divergent pair of genotypes were evaluated by the RAPD technique to determine their degree of homozygosity. Genetic distances based on RAPD data were calculated by the modified Rogers' distance, and also by the following arithmetical complements of similarity: simple match, Nei and Li, and Gower. These genetic distances were similar, presenting a correlation coefficient ranging from 0.99 to 1.00. In all four methods lines UFV 91-717 and Ichigowase were the most divergent ones (4.53 to 21.43%). DNA samples from five plants from each of the two most divergent genotypes were amplified with 28 different primers. Among the amplified products, only five were polymorphic in each group (2.10%), demonstrating their high intragroup degree of homozygosity. These homozygosity were maintained when DNA samples from 12 plants from each of the two most divergent genotypes were amplified. These parameters were extremely useful for the confirmation of the chosen pair of genitors to generate a mapping population.