Benchmarking attribute selection techniques for discrete class data mining

Data engineering is generally considered to be a central issue in the development of data mining applications. The success of many learning schemes, in their attempts to construct models of data, hinges on the reliable identification of a small set of highly predictive attributes. The inclusion of irrelevant, redundant, and noisy attributes in the model building process phase can result in poor predictive performance and increased computation. Attribute selection generally involves a combination of search and attribute utility estimation plus evaluation with respect to specific learning schemes. This leads to a large number of possible permutations and has led to a situation where very few benchmark studies have been conducted. This paper presents a benchmark comparison of several attribute selection methods for supervised classification. All the methods produce an attribute ranking, a useful devise for isolating the individual merit of an attribute. Attribute selection is achieved by cross-validating the attribute rankings with respect to a classification learner to find the best attributes. Results are reported for a selection of standard data sets and two diverse learning schemes C4.5 and naive Bayes.     

Consideration of the errors inherent in mapping historical glacier positions in Austria from the ground and space (1893-2001)

The historical record of in situ measurements of the terminus positions of the Pasterze and Kleines Fleißkees glaciers in the eastern Alps of Austria is used to assess uncertainties in the measurement of decadal scale changes using satellite data. Topographic maps beginning in 1893, and satellite data from 1976 to 2001, were studied in concert with ground measurements to measure glacier changes. Ground measurements show that the tongue of the Pasterze Glacier receded ∼1150 m from 1893 to 2001, while satellite-derived measurements, using August 2001 Landsat Enhanced Thematic Mapper Plus (ETM+) data registered to an 1893 topographic map, show a recession of 1300-1800 m, with an unknown error. The measurement accuracy depends on the registration technique and the pixel resolution of the sensor when two satellite images are used. When using topographic maps, an additional source of error is the accuracy of the glacier position shown on the map. Between 1976 and 2001, Landsat-derived measurements show a recession of the terminus of the Pasterze Glacier of 479±136 m (at an average rate of 19.1 m a⁻¹) while measurements from the ground showed a recession of 428 m (at an average rate of 17.1 m a⁻¹). Four-meter resolution Ikonos satellite images from 2000 and 2001 reveal a shrinkage of 22,096±46 m² in the Pasterze tongue. The nearby Kleines Fleißkees glacier lost 30% of its area between 1984 and 2001, and the area of exposed ice increased by 0.44±0.0023 km², according to Landsat satellite measurements. As more recent satellite images are utilized, especially data that are geocoded, the uncertainty associated with measuring glacier changes has decreased. It is not possible to assess the uncertainty when an old topographic map and a satellite image are coregistered.     

Scheduling and Lot Streaming in Flowshops with No-Wait in Process

Lot streaming involves splitting a production lot into a number of sublots, in order to allow the overlapping of successive operations, in multi-machine manufacturing systems. In no-wait flowshop scheduling, sublots are necessarily consistent, that is, they remain the same over all machines. The benefits of lot streaming include reductions in lead times and work-in-process, and increases in machine utilization rates. We study the problem of minimizing the makespan in no-wait flowshops producing multiple products with attached setup times, using lot streaming. Our study of the single product problem resolves an open question from the lot streaming literature. The intractable multiple product problem requires finding the optimal number of sublots, sublot sizes, and a product sequence for each machine. We develop a dynamic programming algorithm to generate all the nondominated schedule profiles for each product that are required to formulate the flowshop problem as a generalized traveling salesman problem. This problem is equivalent to a classical traveling salesman problem with a pseudopolynomial number of cities. We develop and computationally test an efficient heuristic for this problem. Our results indicate that solutions can quickly be found for flowshops with up to 10 machines and 50 products. Moreover, the solutions found by our heuristic provide a substantial improvement over previously published results.     

RCC-Full: An Effective Network for Parallel Computations

A new interconnection network is proposed for the construction of a massively parallel computer system. The systematic construction of this interconnection network, denoted RCC-FULL, is performed by methodically connecting together a number of basic atoms where a basic atom is a set of fully interconnected nodes. Key communication characteristics are derived and evaluated for RCC-FULL and efficient routing algorithms, which need only local information to route messages between any two nodes, are also derived. AnO(log (N)) sorting algorithm is shown for RCC-FULL and RCC-FULL is shown to emulate deterministically the CRCW PRAM model, with onlyO(log (N)) degradation in time performance. Finally, the hardware cost for the RCC-FULL is estimated as a function of its pin requirements and compared to that of the binary hypercube and most instances of RCC-FULL have substantially lower cost. Hence, RCC-FULL appears to be a particularly effective network for PRAM emulation, and might be considered as a universal network for future supercomputing systems.     

Unstructured Grid-Based Discontinuous Galerkin Method for Broadband Electromagnetic Simulations

A parallel, unstructured, high-order discontinuous Galerkin method is developed for the time-dependent Maxwell's equations, using simple monomial polynomials for spatial discretization and a fourth-order Runge–Kutta scheme for time marching. Scattering results for a number of validation cases are computed employing polynomials of up to third order. Accurate solutions are obtained on coarse meshes and grid convergence is achieved, demonstrating the capabilities of the scheme for time-domain electromagnetic wave scattering simulations.     

Sensitivity Analysis for Scheduling Problems

This paper represents a first attempt at a systematic study of sensitivity analysis for scheduling problems. Because schedules contain both combinatorial and temporal structures, scheduling problems present unique issues for sensitivity analysis. Some of the issues that we discuss have not been considered before. Others, while studied before, have not been explored in the context of scheduling. The applicability of these issues is illustrated using well-known scheduling models. We provide fast methods to determine when a previously optimal schedule remains optimal. Other methods restore an optimal schedule after a parameter change. The value of studying the sensitivity of an optimal sequence instead of the sensitivity of an optimal schedule is demonstrated. We show that, for some problems, sensitivity analysis results depend on the positions of jobs with changed parameters. We identify scheduling problems where performing additional or different computations during optimization facilitates sensitivity analysis. To improve the robustness of an optimal schedule, selection among multiple optimal schedules is considered. We discuss which types of sensitivity analysis questions are intractable because the scheduling problem itself is intractable. We also study how heuristic error bounds vary when the data of a scheduling problem is continuously modified. Although we focus on scheduling problems, several of the issues we discuss and our classification scheme can be extended to other optimization problems.     

A Comparision of Starch Granules as Seen by Both Scanning Electron and Ordinary Light Microscopy

Starch being a transparent crystal often give images which are difficult to precisely define with the light microscope due to the diffraction and other effects such as internal structure which may appear as a surface phenomena. The scanning electron microscope (SEM); however, gives only surface detail. In a effort to differentiate between surface and internal details, the same starch granules have been studied by both ordinary light and scanning electron microscopy. In each case the granules were held in the same configuration as was seen with the light microscope when they were studied by SEM. In this way a direct comparison could be made between granules viewed by each microscopy technique. From such comparisons it is possible to determine the starch details that are actually due to internal features. The results for canna, potato and corn starches are given.     

Characterization of the Basal and mTOR-Dependent Acute Pulmonary and Systemic Immune Response in a Murine Model of Combined Burn and Inhalation Injury

Severe burn injury leads to a cascade of local and systemic immune responses that trigger an extreme state of immune dysfunction, leaving the patient highly susceptible to acute and chronic infection. When combined with inhalation injury, burn patients have higher mortality and a greater chance of developing secondary respiratory complications including infection. No animal model of combined burn and inhalation injury (B+I) exists that accurately mirrors the human clinical picture, nor are there any effective immunotherapies or predictive models of the risk of immune dysfunction. Our earlier work showed that the mechanistic/mammalian target of rapamycin (mTOR) pathway is activated early after burn injury, and its chemical blockade at injury reduced subsequent chronic bacterial susceptibility. It is unclear if mTOR plays a role in the exacerbated immune dysfunction seen after B+I injury. We aimed to: (1) characterize a novel murine model of B+I injury, and (2) investigate the role of mTOR in the immune response after B+I injury. Pulmonary and systemic immune responses to B+I were characterized in the absence or presence of mTOR inhibition at the time of injury. Data describe a murine model of B+I with inhalation-specific immune phenotypes and implicate mTOR in the acute immune dysfunction observed.     

Fast accurate fuzzy clustering through data reduction

Clustering is a useful approach in image segmentation, data mining, and other pattern recognition problems for which unlabeled data exist. Fuzzy clustering using fuzzy c-means or variants of it can provide a data partition that is both better and more meaningful than hard clustering approaches. The clustering process can be quite slow when there are many objects or patterns to be clustered. This paper discusses the algorithm brFCM, which is able to reduce the number of distinct patterns which must be clustered without adversely affecting the partition quality. The reduction is done by aggregating similar examples and then using a weighted exemplar in the clustering process. The reduction in the amount of clustering data allows a partition of the data to be produced faster. The algorithm is applied to the problem of segmenting 32 magnetic resonance images into different tissue types and the problem of segmenting 172 infrared images into trees, grass and target. Average speed-ups of as much as 59-290 times a traditional implementation of fuzzy c-means were obtained using brFCM, while producing partitions that are equivalent to those produced by fuzzy c-means.     

Stackmaker: An Environment for Creating Hypermedia Learning Material

Hypermedia has for some time now been proposed as an adjunct to printed material within the educational process. However, creating a highly interconnected hypermedia network is complex and time consuming, with overviews of the content and structure of the information seemingly essential in order to avoid the disorientation and cognitive overload problems often described.

This paper describes an environment designed to remove much of the burden of creating such support facilities, allowing the teacher to concentrate on the content and structure of the information presented.