Matej Sprogar

Evolution in Medical Decision Making

The classical approach to medical decision making can be limited by the underlying theories. The evolutionary computation is a different concept, which can find many different solutions of the problem. In medicine, this is useful because of different expectations the decision system must face. We implemented a tool for genetic induction of vector decision trees, which are a good choice for a medical decision model because of their simplicity and transparency. The vector decision tree gives multiple classifications in one single pass. Evolutionary development of such trees achieved good results when the results were statistically compared to those of other classical methods. For medical interpretation however a cooperation with doctors is needed to verify the model build.

Finding the right decision tree's induction strategy for a hard real world problem

Decision trees have been already successfully used in medicine, but as in traditional statistics, some hard real world problems can not be solved successfully using the traditional way of induction. In our experiments we tested various methods for building univariate decision trees in order to find the best induction strategy. On a hard real world problem of the Orthopaedic fracture data with 2637 cases, described by 23 attributes and a decision with three possible values, we built decision trees with four classical approaches, one hybrid approach where we combined neural networks and decision trees, and with an evolutionary approach. The results show that all approaches had problems with either accuracy, sensitivity, or decision tree size. The comparison shows that the best compromise in hard real world problem decision trees building is the evolutionary approach.

The Art of Building Decision Trees

Decision support systems that help physicians are becoming a very important part of medical decision making. They are based on different models and the best of them are providing an explanation together with an accurate, reliable, and quick response. One of the most viable among models are decision trees, already successfully used for many medical decision-making purposes. Although effective and reliable, the traditional decision tree construction approach still contains several deficiencies. Therefore we decided to develop and compare several decision support models using four different approaches. We took statistical analysis, a MtDeciT, in our laboratory developed tool for building decision trees with a classical method, the well-known C5.0 tool and a self-adapting evolutionary decision support model that uses evolutionary principles for the induction of decision trees. Several solutions were evolved for the classification of metabolic and respiratory acidosis (MRA). A comparison between developed models and obtained results has shown that our approach can be considered as a good choice for different kinds of real-world medical decision making.Art (from Latin ars meaning skill) is the skill in doing or performing that is attained by study, practice, or observationMicrosoft Bookshelf. 1999 Edition

Evolutionary design of decision trees

Decision tree (DT) is one of the most popular symbolic machine learning approaches to classification with a wide range of applications. Decision trees are especially attractive in data mining. It has an intuitive representation and is, therefore, easy to understand and interpret, also by nontechnical experts. The most important and critical aspect of DTs is the process of their construction. Several induction algorithms exist that use the recursive top‐down principle to divide training objects into subgroups based on different statistical measures in order to achieve homogeneous subgroups. Although being robust and fast, generally providing good results, their deterministic and heuristic nature can lead to suboptimal solutions. Therefore, alternative approaches have developed which try to overcome the drawbacks of classical induction. One of the most viable approaches seems to be the use of evolutionary algorithms, which can produce better DTs as they are searching for globally optimal solutions, evaluating potential solutions with regard to different criteria. We review the process of evolutionary design of DTs, providing the description of the most common approaches as well as referring to recognized specializations. The overall process is first explained and later demonstrated in a step‐by‐step case study using a dataset from the University of California, Irvine (UCI) machine learning repository.

Evaluating breastfeeding advantages using decision trees

Decision support systems that help physicians are becoming a very important part of medical decision making. They are based on different models and the best of them are providing an explanation together with an accurate, reliable and quick response. One of the most viable among these models are decision trees, already successfully used for many medical decision making purposes. Although effective and reliable, the traditional decision tree construction approach still contains several deficiencies. The real-world problem that was a background of our research has a special structure. According to the same attributes we have to determine five different decisions. We decided for two different approaches: the traditional decision trees and a completely new approach of multiple-outcome decision making where we get a decision vector as a result of the decision tree. For constructing such decision trees we used evolutionary programming and we have developed a tool named DecRain. A comparison between developed models and obtained results has shown that our approach can be considered as a good choice for different kinds of real-world medical decision making.

Evolving groups of basic decision trees

A decision tree is a good classifier with a transparent decision mechanism. Decision-tree building methods usually have problems in splitting the learning samples into more subsets, because of the nature of the tree. If the classification into such subsets is not possible, it is better to put the classification decision on to some other classifier. This leads to the introduction of a null classification, which simply means that no classification is possible in this step. This approach is sensible with evolutionary methods, as they can handle a number of trees simultaneously. In the process of construction, we have to address the problem of whether a classification is sensible. The performance of the proposed model has been tested on several data sets and the results presented on one such data set show its potential.

Prudent alignment and crossover of decision trees in genetic programming

Crossover is the central search operator responsible for navigating through unknown problem landscapes while at the same time the main conservation operator, which is supposed to preserve the already learned lessons. This paper is about a novel homologous decision tree crossover operator. Contrary to other tree crossover operators it defines the context for a decision tree node and elaborates a fast one-sample-based tree alignment procedure. The idea is to replace a sub-tree with a better one from the same context, as defined by the decision tree training process. This operator does not rely on the topological properties of the tree but rather on its behavioral properties. During empirical testing the new operator showed the best generalization capabilities.

A Study of GP's Division Operators for Symbolic Regression

Article explores three possible implementations of the division operator in genetic programming -- protected division, division throwing an exception and division returning an undefined result. The article proposes a simple classification scheme and discusses the subtle differences between operators. Results of a case study on the well known binomial-3 problem are presented. They suggest that the (most common) protected division returning 1 is not always the safest bet.

Agent-oriented framework for decision tree evolution

An autonomous evolutionary framework for construction of decision trees that requires no or minimal human interaction is presented. The framework evolves two types of agents which hold the discovered knowledge, and uses a non-standard implicit fitness evaluation in a co-evolving environment. Together with self-adaptation of evolutionary parameters and with some other improvements it can monitor and adjust its own behavior. This framework is a base for a specific implementation of a program for induction of decision trees. The programs capability to self-adapt to a given problem is used as a measure to predict if some dataset is difficult or even impossible to analyze. On average it produces very general solutions or gives no solution if the dataset is prone to the overfitting problem.

Notes on Medical Decision Model Creation

To evaluate some intelligent method for decision-making we need to compare the method against the competing methods to get some idea of its performance and capabilities. In everyday practice this is enough and the method, if proven good, can be used for different problems. In medicine however we have a demand for the best solution in all circumstances. It is therefore impossible to declare one method as acceptable for every type or even every single problem. In this article we would like to stress some important aspects of machine learning in medicine, especially the creation of specific decision models. We believe the evolutionary concept is a good approach to this as it creates many diverse solutions.