We propose a method for generating rule sets as global and local explanations for tree-ensemble learning methods using answer set programming (ASP). To this end, we adopt a decompositional approach where the split structures of the base decision trees are exploited in the construction of rules, which in turn are assessed using pattern mining methods encoded in ASP to extract explanatory rules. For global explanations, candidate rules are chosen from the entire trained tree-ensemble models, whereas for local explanations, candidate rules are selected by only considering rules that are relevant to the particular predicted instance. We show how user-defined constraints and preferences can be represented declaratively in ASP to allow for transparent and flexible rule set generation, and how rules can be used as explanations to help the user better understand the models. Experimental evaluation with real-world datasets and popular tree-ensemble algorithms demonstrates that our approach is applicable to a wide range of classification tasks.

Fast-and-frugal trees (FFTs) are simple algorithms that facilitate efficient and accurate decisions based on limited information. But despite their successful use in many applied domains, there is no widely available toolbox that allows anyone to easily create, visualize, and evaluate FFTs. We fill this gap by introducing the R package FFTrees. In this paper, we explain how FFTs work, introduce a new class of algorithms called fan for constructing FFTs, and provide a tutorial for using the FFTrees package. We then conduct a simulation across ten real-world datasets to test how well FFTs created by FFTrees can predict data. Simulation results show that FFTs created by FFTrees can predict data as well as popular classification algorithms such as regression and random forests, while remaining simple enough for anyone to understand and use.

Glyphosate-resistant (GR) crops have been rapidly adopted in the United States and the evolution of GR weeds throughout the world has also been on the rise. With experience, weed scientists and crop advisers develop “intuition” on the basis of field history and current in-field conditions for predicting whether escaped weed biotypes may be herbicide resistant. However, there are no previous reports on the association of in-field crop management factors with the prediction of herbicide resistance. By using in-field survey data, we tested the accuracy of predicting glyphosate resistance in late-season horseweed escapes. We hypothesized that glyphosate resistance in late-season horseweed populations found in soybean fields could be predicted using in-field knowledge of crop residues and the appearance and distribution of weeds in the field. Field survey data were collected to determine the distribution and frequency of GR horseweed populations in Indiana soybean fields during September and October of 2003, 2004, and 2005. After the in-field survey, soil properties for sampled field locations were also collected from the U.S. Department of Agriculture Natural Resources Conservation Service Web Soil Survey. GR horseweed predictions used in-field presence of crop residues and the appearance, abundance, and distribution of weeds in the field. The significance of independent data factors were determined by chi-square statistics. The interactions and relative significance of multiple factors were modeled using classification and regression tree analysis. Our results indicated that the most important factor for predicting GR populations was the identification of an altered plant phenotype after injury from POST glyphosate. This was followed by crop rotation, field distribution, and the presence of other escaped weed species in the field in a model with a classification rate of 0.68.

Objectives: To compare the cost-effectiveness of alternative methods of screening for retinopathy of prematurity (ROP) in the United Kingdom, including the existing method of indirect ophthalmoscopy by ophthalmologists and digital photographic screening by nurses.

Methods: A decision tree model was used to compare five screening modalities for the UK population of preterm babies, using a health service perspective. Data were taken from published sources, observation at a neonatal intensive care unit (NICU), and expert judgment.

Results: We estimated that use of standard digital cameras by nurses in NICUs would cost more than current methods (£371 compared with £321 per baby screened). However, a specialist nurse visiting units with a portable camera would be cheaper (£172 per baby). These estimates rely on nurses capturing and interpreting the images, with suitable training and supervision. Alternatively, nurses could capture the images then transmit them to a central unit for interpretation by ophthalmologists, although we estimate that this would be rather more expensive (£390 and £201, respectively, for NICU and visiting nurses). Sensitivity analysis was used to examine the robustness of estimates.

Conclusions: It is likely that there is an opportunity to improve the efficiency of the ROP screening program. We estimate that screening by specialist nurses trained in image capture and interpretation using portable digital cameras is a cost-effective alternative to the current program of direct visualization by ophthalmologists. This option would require the development of a suitable portable machine. Direct comparative research is strongly needed to establish the accuracy of the various screening options.

Four bar mechanisms are basic components of many important mechanical devices. The kinematic synthesis of four bar mechanisms is a difficult design problem. A novel method that combines the genetic programming and decision tree learning methods is presented. We give a structural description for the class of mechanisms that produce desired coupler curves. Constructive induction is used to find and characterize feasible regions of the design space. Decision trees constitute the learning engine, and the new features are created by genetic programming.

A software tool for data mining (DMiner-I) is introduced, which integrates pattern recognition (PCA, Fisher, clustering, HyperEnvelop, regression), artificial intelligence (knowledge representation, decision trees), statistical learning (rough set, support vector machine), and computational intelligence (neural network, genetic algorithm, fuzzy systems). It consists of nine function models: pattern recognition, decision trees, association rule, fuzzy rule, neural network, genetic algorithm, HyperEnvelop, support vector machine and visualization. The principle, algorithms and knowledge representation of some function models of data mining are described. Nonmonotony in data mining is dealt with by concept hierarchy and layered mining. The software tool of data mining is realized by Visual C++ under Windows 2000. The software tool of data mining has been satisfactorily applied in the prediction of regularities of the formation of ternary intermetallic compounds in alloy systems, and diagnosis of brain glioma.

This paper proposes a novel hierarchical multi-layer decision tree for representing reactive robot navigation knowledge. In this representation, the perception space is decomposed into a hierarchical set of worlds reflecting environments which are homogeneous in nature and which vary in complexity in an ordered manner. Each world is used to produce a corresponding decision tree which is trained incrementally. The instantaneous perception of the robot is used to select an appropriate rule from the decision tree and a sequence of rule activations form the complete trajectory. The ability to keep the knowledge complexity manageable and under control is an important aspect of the technique.