Milan Zorman

The Limitations of Decision Trees and Automatic Learning in Real World Medical Decision Making

The decision tree approach is one of the most common approaches in automatic learning and decision making. The automatic learning of decision trees and their use usually show very good results in various “ theoretical” environments. But in real life it is often impossible to find the desired number of representative training objects for various reasons. The lack of possibilities to measure attribute values, high cost and complexity of such measurements, and unavailability of all attributes at the same time are the typical representatives. For this reason we decided to use the decision trees not for their primary task—the decision making—but for outlining the most important attributes. This was possible by using a well-known property of the decision trees—their knowledge representation, which can be easily understood by humans. In a delicate field of medical decision making, we cannot allow ourselves to make any inaccurate decisions and the “tips,” provided by the decision trees, can be of a great assistance. Our main interest was to discover a predisposition to two forms of acidosis: themetabolic acidosis and respiratory acidosis, which can both have serious effects on childs health. We decided to construct different decision trees from a set of training objects. Instead of using a test set for evaluation of a decision tree, we asked medical experts to take a closer look at the generated trees. They examined and evaluated the decision trees branch by branch. Their comments show that trees generated from the available training set mainly have surprisingly good branches, but on the other hand, for some, no medical explanation could be found.

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.

Mining diabetes database with decision trees and association rules

Searching for new rules and new knowledge in problem areas, where very little or almost none previous knowledge is present, can be a very long and demanding process. In our research we addressed the problem of finding new knowledge in the form of rules in the diabetes database using a combination of decision trees and association rules. The first question we wanted to answer was, if there are significant differences in sets of rules both approaches produce, and how rules, produced by decision trees behave, after being a subject of filtering and reduction, normally used in association rule approaches. In order to accomplish that, we had to make some modifications to both the decision tree approach and association rule approach. From the first results we can conclude, that the sets of rules, built by decision trees are much smaller than the sets created by association rules. We could also establish, that filtering and reduction did not effect the rules derived from decision trees in the same scale as association rules.

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

Decision tree's induction strategies evaluated on a hard real world problem

Decision trees have been already been successfully used in medicine, but as in traditional statistics, some hard real-world problems cannot be solved successfully using the traditional method 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 concerning 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, with a hybrid approach (where we combined neural networks and decision trees) and with an evolutionary approach. The results show that all the approaches had problems with either accuracy or decision tree size. The comparison shows that the best compromise in hard real-world decision-tree building is the evolutionary approach.

Computer and natural language texts—a comparison based on long-range correlations

“Long‐range power low correlation” (LRC) is defined as a maximal propagation distance of the effect of some disturbance within a system found in many systems that can be represented as strings of symbols. LRC between characters has also been identified in natural language texts. The aim of this article is to show that long‐range power law correlation can also be found in computer programs, meaning that some common laws hold for both natural language texts and computer programs. This fact enables one to draw parallels between these two different types of human writings, and also enables one to measure the differences between them.

Analysis of approaches to structured data on the web

The early concept of the World Wide Web was the network of related (linked) documents represented in human readable form. The ongoing development leads to another aspect of the web, the web of data. The goal being that the network will provide first-class, machine readable data. Therefore the current network will be transformed to a network where the machines will not only serve as the platform that hosts human readable data but as a true machine-machine network. In this paper, we review and compare the formats, technologies and approaches that are used today for publishing semantic, machine readable data, on the web.

Symbol-Based Machine Learning Approach for Supervised Segmentation of Follicular Lymphoma Images

Lymphoma is a broad term encompassing a variety of cancers of the lymphatic system. Lymphoma is differentiated by the type of cell that multiplies and how the cancer presents itself. It is very important to get an exact diagnosis regarding lymphoma and to determine the treatments that will be most effective for the patients condition. Our work was focused on the identification of lymphomas by finding follicles in microscopy images provided by the Laboratory of Pathology in the University Hospital of Tenerife, Spain. Roughly we can divide our work in two stages: in the first stage we did image pre-processing and feature extraction, and in the second stage we used different rough set approaches for pixel classification. These results were compared to decision tree results we obtained earlier. Symbolic machine learning approaches are often neglected when looking for image analysis tools. The results we got are very promising and show that symbolic approaches can be successful in image analysis applications.

Improved medical decision support with multimethod approach

An enormous proliferation of databases in almost all areas of human life has created great need to develop tools for automatic knowledge extraction. Extracted knowledge can be used for categorizing, organizing or predictive purposes. One of the problems encountered is how to make a good induction with good generalization and knowledge representation what is especial important in medical domain. Although the research filed is very active, it is mainly focused on a specific method or on a specific combination of those methods. In this paper a multimethod approach is presented. This approach unlike other conventional hybrid approaches applies different methods on the same knowledge base where each method may contain inherent limitations with the expectation that the combined multiple methods may produce better results. It also addresses unbalanced nature of medical data.

Does Size Really Matter—Using a Decision Tree Approach for Comparison of Three Different Databases from the Medical Field of Acute Appendicitis

Decision trees have been successfully used for years in many medical decision making applications. Transparent representation of acquired knowledge and fast algorithms made decision trees one of the most often used symbolic machine learning approaches. This paper concentrates on the problem of separating acute appendicitis, which is a special problem of acute abdominal pain, from other diseases that cause acute abdominal pain by use of an decision tree approach. Early and accurate diagnosing of acute appendicitis is still a difficult and challenging problem in everyday clinical routine. An important factor in the error rate is poor discrimination between acute appendicitis and other diseases that cause acute abdominal pain. This error rate is still high, despite considerable improvements in history-taking and clinical examination, computer-aided decision-support, and special investigation such as ultrasound. We investigated three databases of different size with ca ses of acute abdominal pain to complete this task as successful as possible. The results show that the size of the database does not necessary directly influence the success of the decision tree built on it. Surprisingly we got the best results from the decision trees built on the smallest and the biggest database, where the database with medium size (relative to the other two) was not so successful. Despite this we were able to produce decision tree classifiers that were capable of producing correct decisions on test data sets with accuracy up to 84%, sensitivity to acute appendicitis up to 90%, and specificity up to 80% on the same test set.