A new reverse reduce-error ensemble pruning algorithm

Although greedy algorithms possess high efficiency, they often receive suboptimal solutions of the ensemble pruning problem, since their exploration areas are limited in large extent. And another marked defect of almost all the currently existing ensemble pruning algorithms, including greedy ones, consists in: they simply abandon all of the classifiers which fail in the competition of ensemble selection, causing a considerable waste of useful resources and information. Inspired by these observations, an interesting greedy Reverse Reduce-Error (RRE) pruning algorithm incorporated with the operation of subtraction is proposed in this work. The RRE algorithm makes the best of the defeated candidate networks in a way that, the Worst Single Model (WSM) is chosen, and then, its votes are subtracted from the votes made by those selected components within the pruned ensemble. The reason is because, for most cases, the WSM might make mistakes in its estimation for the test samples. And, different from the classical RE, the near-optimal solution is produced based on the pruned error of all the available sequential subensembles. Besides, the backfitting step of RE algorithm is replaced with the selection step of a WSM in RRE. Moreover, the problem of ties might be solved more naturally with RRE. Finally, soft voting approach is employed in the testing to RRE algorithm. The performances of RE and RRE algorithms, and two baseline methods, i.e., the method which selects the Best Single Model (BSM) in the initial ensemble, and the method which retains all member networks of the initial ensemble (ALL), are evaluated on seven benchmark classification tasks under different initial ensemble setups. The results of the empirical investigation show the superiority of RRE over the other three ensemble pruning algorithms.     

An integrated alcoholic index using tunable-Q wavelet transform based features extracted from EEG signals for diagnosis of alcoholism

Alcoholism affects the structure and functioning of brain. Electroencephalogram (EEG) signals can depict the state of brain. The EEG signals are ensemble of various neuronal activity recorded from different scalp regions having different characteristics and very low magnitude in microvolts. These factors make human interpretation difficult and time consuming to analyze these signals. Moreover, these highly varying EEG signals are susceptible to inter/intra variability errors. So, a Computer-Aided Diagnosis (CAD) can be used to identify the alcoholic and normal subjects accurately. However, these EEG signals exhibit nonlinear and non-stationary properties. Therefore, it needs much effort in deciphering the diagnostic evidence from them using linear time and frequency-domain methods. The nonlinear parameters together with time-frequency/scale domain methods can help to detect tiny changes in these signals. The correntropy is nonlinear indicator which characterizes the dynamic behavior of EEG signals in time-scale domain. In this paper, we present a new way for diagnosis of alcoholism using Tunable-Q Wavelet Transform (TQWT) based features derived from EEG signals. The feature extraction is performed using TQWT based decomposition and extracted Centered Correntropy (CC) from the forth decomposed detail sub-band. The Principal Component Analysis (PCA) is used for feature reduction followed by Least Squares-Support Vector Machine (LS-SVM) for classifying normal and alcoholic EEG signals. In order to make sure reliable classification performance, 10-fold cross-validation scheme is adopted. Our proposed system is able to diagnose the alcoholic and normal EEG signals, with an average accuracy of 97.02%, sensitivity of 96.53%, specificity of 97.50% and Matthews correlation coefficient of 0.9494 for Q-factor (Q) varying between 3 and 8 using Radial Basis Function (RBF) kernel function. Also, we have established a novel Alcoholism Risk Index (ARI) using three clinically significant features to discriminate the given classes by means of a single number. This system can be used for automated diagnosis and monitoring of alcoholic subjects to evaluate the effect of treatment.     

Using self-organizing maps to model turnover of sales agents in a call center

This paper proposes an approach for modeling employee turnover in a call center using the versatility of supervised self-organizing maps. Two main distinct problems exist for the modeling employee turnover: first, to predict the employee turnover at a given point in the sales agent's trial period, and second to analyze the turnover behavior under different performance scenarios by using psychometric information about the sales agents. Identifying subjects susceptible to not performing well early on, or identifying personality traits in an individual that does not fit with the work style is essential to the call center industry, particularly when this industry suffers from high employee turnover rates. Self-organizing maps can model non-linear relations between different attributes and ultimately find conditions between an individual's performance and personality attributes that make him more predisposed to not remain long in an organization. Unlike other models that only consider performance attributes, this work successfully uses psychometric information that describes a sales agent's personality, which enables a better performance in predicting turnover and analyzing potential personality profiles that can identify agents with better prospects of a successful career in a call center. The application of our model is illustrated and real data are analyzed from an outbound call center.     

Combination of multiple diverse classifiers using belief functions for handling data with imperfect labels

This paper addresses the supervised learning in which the class memberships of training data are subject to ambiguity. This problem is tackled in the ensemble learning and the Dempster-Shafer theory of evidence frameworks. The initial labels of the training data are ignored and by utilizing the main classes’ prototypes, each training pattern is reassigned to one class or a subset of the main classes based on the level of ambiguity concerning its class label. Multilayer perceptron neural network is employed to learn the characteristics of the data with new labels and for a given test pattern its outputs are considered as basic belief assignment. Experiments with artificial and real data demonstrate that taking into account the ambiguity in labels of the learning data can provide better classification results than single and ensemble classifiers that solve the classification problem using data with initial imperfect labels.     

Tree ensemble construction using a GRASP-based heuristic and annealed randomness

Two new methods for tree ensemble construction are presented: G-Forest and GAR-Forest. In a similar way to Random Forest, the tree construction process entails a degree of randomness.The same strategy used in the GRASP metaheuristic for generating random and adaptive solutions is used at each node of the trees. The source of diversity of the ensemble is the randomness of the solution generation method of GRASP. A further key feature of the tree construction method for GAR-Forest is a decreasing level of randomness during the process of constructing the tree: maximum randomness at the root and minimum randomness at the leaves. The method is therefore named “GAR”, GRASP with annealed randomness.The results conclusively demonstrate that G-Forest and GAR-Forest outperform Bagging, AdaBoost, MultiBoost, Random Forest and Random Subspaces. The results are even more convincing in the presence of noise, demonstrating the robustness of the method.The relationship between base classifier accuracy and their diversity is analysed by application of kappa-error diagrams and a variant of these called kappa-error relative movement diagrams.     

斜窄上升流过程与单元集成式分离设备

拉美拉斜板以及早年的Lavodune/Ladoflux分离设备,应用的都是斜窄上升流过程,但前者是单元集成结构,而后者是单个设备。介绍了在拉美拉的基础上,对斜窄流过程及其单元集成设备结构特征的系统研究以及斜窄流浓缩、分级、沉渣-除油等分离设备及其在矿业多种领域的实际应用。     

Deriving reservoir operating policies using classifier systems

An application of classifier systems to time variant, water resources allocation problems is described, the aim of the work being to assess the suitability of the technique for deriving control strategies. The difficulties of allocating credit to classifiers where co-operating sequences of rules have to be developed are discussed and methods for overcoming some of the difficulties are covered. The training of the classifier on two problems, the first containing a single, surface water reservoir and the second two reservoirs is used to develop techniques and the results show that classifiers possess an ability to learn about the domain. However the resulting operating strategies are not appropriate for the operation of water resources systems. The work indicates that in their current format, classifier systems cannot learn to operate systems where long, interdependent chains of decisions are involved.     

Combining predictions in pairwise classification: An optimal adaptive voting strategy and its relation to weighted voting

Weighted voting is the commonly used strategy for combining predictions in pairwise classification. Even though it shows good classification performance in practice, it is often criticized for lacking a sound theoretical justification. In this paper, we study the problem of combining predictions within a formal framework of label ranking and, under some model assumptions, derive a generalized voting strategy in which predictions are properly adapted according to the strengths of the corresponding base classifiers. We call this strategy adaptive voting and show that it is optimal in the sense of yielding a MAP prediction of the class label of a test instance. Moreover, we offer a theoretical justification for weighted voting by showing that it yields a good approximation of the optimal adaptive voting prediction. This result is further corroborated by empirical evidence from experiments with real and synthetic data sets showing that, even though adaptive voting is sometimes able to achieve consistent improvements, weighted voting is in general quite competitive, all the more in cases where the aforementioned model assumptions underlying adaptive voting are not met. In this sense, weighted voting appears to be a more robust aggregation strategy.     

Supervised data analysis and reliability estimation with exemplary application for spectral data

The analysis and classification of data is a common task in multiple fields of experimental research such as bioinformatics, medicine, satellite remote sensing or chemometrics leading to new challenges for an appropriate analysis. For this purpose different machine learning methods have been proposed. These methods usually do not provide information about the reliability of the classification. This, however, is a common requirement in, e.g. medicine and biology. In this line the present contribution offers an approach to enhance classifiers with reliability estimates in the context of prototype vector quantization. This extension can also be used to optimize precision or recall of the classifier system and to determine items which are not classifiable. This can lead to significantly improved classification results. The method is exemplarily presented on satellite remote spectral data but is applicable to a wider range of data sets.     

一种网络数据包的有效分类方法

终端用户的各种网络行为都会产生大量的数据包，在没有其它任何先验知识的情况下，如何把数据包分成网络游戏包和非网络游戏包是一件困难的工作。通过对大量的数据包样本进行分析，发现用地址偶对和端口偶对的线性拟合斜率作为模式的两个特征，再构造出一个线性分类器，采用增量校正算法求解该线性分类器的权向量，进一步利用该线性分类器，可以对各种环境下新的数据包准确划分为网络游戏包和非网络游戏包，据此可以控制终端用户网络游戏行为。