Search

Effectively grouping learners in an online environment is a highly useful task. However, datasets used in this task often have large numbers of attributes of disparate types and different scales, which traditional clustering approaches cannot handle effectively. Here, a unique dissimilarity measure based on the random forest, which handles the stated drawbacks of more traditional clustering approaches, is presented. Additionally, a rule-based method is proposed for interpreting the resulting learner segmentations. The approach was implemented on a real dataset of users of the "Career"WISE online educational environment, designed to provide resilience training for women STEM doctoral students, and was shown to find stable and meaningful groups of users.

This paper proposes a new feature selection methodology. The methodology is based on the stepwise variable selection procedure, but, instead of using the traditional discriminant metrics such as Wilks' Lambda, it uses an estimation of the misclassification error as the figure of merit to evaluate the introduction of new features. The expected misclassification error rate (MER) is obtained by using the densities of a constructed function of random variables, which is the stochastic representation of the conditional distribution of the quadratic discriminant function estimate. The application of the proposed methodology results in significant savings of computational time in the estimation of classification error over the traditional simulation and cross-validation methods. One of the main advantages of the proposed method is that it provides a direct estimation of the expected misclassification error at the time of feature selection, which provides an immediate assessment of the benefits of introducing an additional feature into an inspection/classification algorithm. Index Terms--Feature selection, misclassification error rate, quadratic discriminant function.

Process knowledge can be exploited to improve the performance of control charts and it is not unusual to know that a specific variable shifts above or below its mean under an assignable cause. In such a case, a one-sided control chart is common. The available statistical theory for the one-sided tests is used to provide a reasonable compromise for a numerical procedure to design and implement multivariate solutions. Although simulation is used in the analysis, it is not a direct estimate of performance through simulation. Instead, weights are estimated and these are used to easily set a desired on-target average run length. Furthermore, an interesting quadratic programming solution is incorporated into the analysis. Then the statistical results are extended to a partial one-sided case where only some (not all) variables are known to shift in one direction and the numerical procedure is extended to design and implement the charts. A modern method can blend theory and algorithms into a practical solution. We conclude that modern computer software permits an efficient solution to this problem with meaningful performance advantages over traditional multivariate control charts., 1. Introduction Multivariate control charts such as chi-squared, Multivariate Exponentially Weighted Moving Average (MEWMA) and multivariate CUmulative SUM (CUSUM) charts are sensitive to shifts of the process mean vector in [...]

An estimate of cycle time for a product in a factory is critical to semiconductor manufacturers (and in other industries) to assess customer due dates, schedule resources and actions for anticipated job completions, and to monitor the operation. Historical data can be used to learn a predictive model for cycle time based on measured and calculated process metrics (such as work-in-progress at specific operations, lot priority, product type, and so forth). Such a method is relatively easy to develop and maintain. Modern data mining algorithms are used to develop nonlinear predictors applicable to the majority of process lots, and three methods are compared here. They are compared with respect to performance in actual manufacturing data (to predict times for both final and intermediate steps) and for the feasibility to maintain and rebuild the model. Index Terms--Due date, scheduling, statistical models, work-in-progress (WIP).

In the fields of statistics and computer science a wide variety of methodologies exist for solving the traditional classification problem. This study will compare the error rates for various methods under a variety of conditions when the explanatory variables are continuous. The methods under considerations are neural networks, classical discriminant analysis, and two different approaches to decision trees. Training and testing sets are utilized to estimate the error rates of these methods for different numbers of sample sizes, number of explanatory variables, and the number of classes in the dependent variable. These error rates will be used to draw generalized conclusions about the relative efficiencies of the techniques., 1. Introduction The classification of objects into groups or supervised learning is a common decision-making requirement. In general, there are two primary approaches to this process of classification. The traditional [...]

Probe testing following wafer fabrication can produce extremely large amounts of data, which is often used to inspect a final product to determine if the product meets specifications. This data can be further utilized in studying the effects of the wafer fabrication process on the quality or yield of the wafers. Relationships among the parameters may provide valuable process information that can improve future production. This paper compares many methods of using the probe test data to determine the cause of low yield wafers. The methods discussed include two classes of traditional multivariate statistical methods, clustering and principal component methods and regression-based methods. These traditional methods are compared to a classification and regression tree (CART) method. The results for each method are presented. CART adequately fits the data and provides a 'recipe' for avoiding low yield wafers and because CART is distribution-free there are no assumptions about the distributional properties of the data. CART is strongly recommended for analyzing wafer probe data. Index Terms--CART, multivariate statistical methods, tree regression, yield analysis.

In this paper, we introduce the concept of fine-tuned learning which relies on the notion of data approximation followed by sequential data refinement. We seek to determine whether fine-tuned learning is a viable approach to use when trying to solve combinatorial optimization problems. In particular, we conduct an extensive computational experiment to study the performance of fine-tuned-learning-based heuristics for the traveling salesman problem (TSP). We provide important insight that reveals how fine-tuned learning works and why it works well, and conclude that it is a meritorious concept that deserves serious consideration by researchers solving difficult problems.

Modern statistical process control must often cope with large quantities of highly autocorrelated data. Alwan and Radson (1992) proposed the monitoring of autocorrelated processes by plotting the averages of small batches of data separated by skipping observations. Using results for the AR(1) process, we show that generally better performance can be achieved with no skipping and much larger batch sizes. The resulting batch-means charts derive from methods used in simulation output analysis and can be implemented easily with common digital control systems., 1. Introduction Increased use of online automated data acquisition systems is producing unprecedented volumes of data with high autocorrelation (Hahn, 1989). This trend is most pronounced in the chemical and [...]

Adaptive sampling is an enhancement to a classical statistical process control scheme in which the time between samples from the process is varied based on the available information. Unlike the classical Shewhart control schemes, the performance of an adaptive sampling scheme for detecting a process which is off-target before the first observation is selected can differ from the steady-state performance, which is the performance of the scheme for detecting a process shift after the control scheme has been operating for some time. Often steady-state performance is the more important criterion for evaluating statistical process control schemes. Previous work has considered initial performance and steady-state performance, but the charts examined were optimized for initial performance. The development of charts optimized for steady-state performance and a comparison to previously developed schemes are the topics of this paper. Tables providing performance information on adaptive schemes are included.

Gradient boosting method has been successfully used for single target prediction problems. In real world applications, however, problems involving the prediction of multiple target attributes are often of interest. In this paper, a multi-target boosting method for regression problems, named as MTBR, is proposed. Although MTBR builds one model for each target attribute separately, all the target attributes are utilized when building each model. In each boosting iteration, the base learner, the regression tree in particular, is learned by selecting the best models from all the target attributes. We also introduce a novel knowledge transfer approach. That is, the tree structure learned from one target attribute, representing a way to partition the feature space, is used to predict another target attribute. Experiments with six data sets compare MTBR to other ensemble regression methods, and prove the effectiveness of MTBR in leveraging the knowledge of multiple target attributes and improving the model accuracy. [ABSTRACT FROM AUTHOR]

Motivation Matched case–control analysis is widely used in biomedical studies to identify exposure variables associated with health conditions. The matching is used to improve the efficiency. Existing variable selection methods for matched case–control studies are challenged in high-dimensional settings where interactions among variables are also important. We describe a quite different method for high-dimensional matched case–control data, based on the potential outcome model, which is not only flexible regarding the number of matching and exposure variables but also able to detect interaction effects. Results We present Matched Forest (MF), an algorithm for variable selection in matched case–control data. The method preserves the case and control values in each instance but transforms the matched case–control data with added counterfactuals. A modified variable importance score from a supervised learner is used to detect important variables. The method is conceptually simple and can be applied with widely available software tools. Simulation studies show the effectiveness of MF in identifying important variables. MF is also applied to data from the biomedical domain and its performance is compared with alternative approaches. Availability and implementation R code for implementing MF is available at https://github.com/NooshinSh/Matched%5fForest. Supplementary information Supplementary data are available at Bioinformatics online. [ABSTRACT FROM AUTHOR]

The detection of changes in the distribution of process variables is referred to as the change-point problem. Existing methods focus on detecting a single (or few) change point in a univariate (or low-dimensional) process. We consider the important high-dimensional multivariate case with multiple change points and without an assumed distribution. In this work the problem is transformed into a supervised learning problem with time as the output response and the process variables as inputs. Our focus is to identify the subset of variables that change. This important, practical scenario is analysed through a supervised learner with a variable importance measure that is used to identify the variables that change among hundreds of variables. Simulated cases are discussed in the paper to verify the proposed method. Moreover, the same data sets are compared with a multivariate exponentially weighted moving average control chart and the advantages of the supervised learner are illustrated. [ABSTRACT FROM AUTHOR]

The 2N problem is defined by the function h, which maps N to N on the set of positive integers, h(N) = (2N), where N is a 2-digit positive integer. A 2-digit positive integer is chosen at random from 10-99 and multiplied by 2. The last two digits of the resulting product are then isolated and multiplied by 2, and the process is repeated until the starting integer is obtained. For even integers, this process results in either of three outcomes, a repetitive loop without returning to the starting integer, or 4 or 20 repetitions to return to the starting integer.

A procedure for monitoring multiple discrete counts is proposed. This procedure is based on the likelihood ratio statistic for Poisson counts when input variables are measurable. This statistic is the deviance residual resulting from a generalized linear model. The likelihood ratio statistic is derived and shown to be more effective than using a C chart on the raw counts. [ABSTRACT FROM AUTHOR]

A particle count data stream is examined. The data are shown to consist of two mixed process distributions, a base Poisson count process and an outlier process. A model is developed that describes the mixed data stream. A smoothing and filtering method, using an exponentially weighted moving average (EWMA) and Poisson probabilities, is developed that separates the two process distributions into a base process and an outlier process. By separating the two distributions, statistical monitoring schemes can be applied to each. A Bernoulli EWMA is introduced for monitoring the outlier process. Average run length (ARL) evaluation is performed using Markov chain methodology, and suggestions for implementation are provided. [ABSTRACT FROM AUTHOR]

A comprehensive performance study and comparison of several adaptive statistical process control procedures is presented. These adaptive control chart procedures are modifications to standard Shewhart control charts that include changing the sampling interval, the sample size or both according to rules based on the value of the sample statistic. Adaptive control techniques are known to improve the performance of the standard Shewhart control charts. In this paper we develop a four-state adaptive sample size control chart and several variations of a three-state combined adaptive sample size and sampling interval control chart. We then compare these new schemes with the previously developed schemes, the two-state adaptive sampling interval, the two-state adaptive sample size and two-state combined adaptive sample size and sampling interval control chart, three-state adaptive sample size control chart and non-adaptive Shewhart control charts. These results show that the addition of the third and fourth states on the adaptive control chart schemes improve the control chart performance; however, the improvement is relatively modest. [ABSTRACT FROM AUTHOR]

A computationally efficient and flexible method for mapping categorical variables to numeric scores in heterogeneous (mixed-type) data is discussed. A novel scoring objective attempts to preserve the mutual information between all the variables using a nontraditional clustering approach for mixed data.

Douglas M. HAWKINS and David H. OLWELL. New York: Springer-Verlag, 1998. ISBN 0-387-98365-1. xii + 247 pp. $54.95. A book devoted to this topic is an excellent contribution. Over the [...]