关键词： Kernel methods   Low-dimensional embedding   Performance measurement and business   improvement   Machine learning  

摘要： The aim of this paper is to investigate the impact of strategic planning on service small and medium size enterprises' (SMEs) performance. A machine learning methodology, based on an alternating diffusion process, is applied for organizing the SMEs into a network/graph, and generating management profiles. The method relies on ideas from non-linear dimensionality reduction, where a data-set is parameterized by a small number of intrinsic parameters. Recent advances in this field allow for the construction of a common intrinsic representation for two related dataset by using kernel multiplication. In this work, the method is applied to two related datasets that describe a business from two angles: strategic planning and performance. The goal is to reveal underlying hidden strategic patterns that influence SMEs' performance. The proposed methods are applied to a dataset containing planning and performance measures of 467 businesses. The obtained model yields an SME graph, where SMEs that share similar planning strategies and performances are located close to each other. Thus, the model generates typical strategic planning and performance profiles, describing distinct groups of SMEs. The study reveals two different planning strategies emphasizing different management approaches that lead to successful performance. Thus, in service SMEs there is more than one way to enhance performance via management strategy. Differentiating between various planning strategies was enabled by utilizing kernel based machine learning methods, thereby overcoming limitations of linear methods that cannot provide such sensitive profile configurations. (C) 2017 Elsevier B.V. All rights reserved.

关键词： deep learning   drug discovery   machine learning   pharmaceutics   support vector machine  

摘要： Machine learning methods have been applied to many data sets in pharmaceutical research for several decades. The relative ease and availability of fingerprint type molecular descriptors paired with Bayesian methods resulted in the widespread use of this approach for a diverse array of end points relevant to drug discovery. Deep learning is the latest machine learning algorithm attracting attention for many of pharmaceutical applications from docking to virtual screening. Deep learning is based on an artificial neural network with multiple hidden layers and has found considerable traction for many artificial intelligence applications. We have previously suggested the need for a comparison of different machine learning methods with deep learning across an array of varying data sets that is applicable to pharmaceutical research. End points relevant to pharmaceutical research include absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) properties, as well as activity against pathogens and drug discovery data sets. In this study, we have used data sets for solubility, probe-likeness, hERG, KCNQ1, bubonic plague, Chagas, tuberculosis, and malaria to compare different machine learning methods using FCFP6 fingerprints. These data sets represent whole cell screens, individual proteins, physicochemical properties as well as a data set with a complex end point. Our aim was to assess whether deep learning offered any improvement in testing when assessed using an array of metrics including AUC, F1 score, Cohens kappa, Matthews correlation coefficient and others. Based on ranked normalized scores for the metrics or data sets Deep Neural Networks (DNN) ranked higher than SVM, which in turn was ranked higher than all the other machine learning methods. Visualizing these properties for training and test sets using radar type plots indicates when models are inferior or perhaps over trained. These results also suggest the need for assessing deep learning further us

关键词： Machine learning   Endless learning   Computational model   Classification   Regression   Behavioral human learning  

摘要： Learning is an important talent for understanding the nature and accordingly controlling behavioral characteristics. Behavioral learning theories are one of the popular learning theories which are built on experimental findings. These theories are widely applied in psychotherapy, psychology, neurology as well as in advertisements and robotics. There is an abundant literature associated with understanding learning mechanism, and various models have been proposed for the realization of learning theories. Nevertheless, none of those models are able to satisfactorily simulate the concept of classical conditioning. In this study, popular behavioral learning theories were firstly simplified and the contentious issues with them were clarified by conducting intuitive experiments. The experimental results and information available in the literature were evaluated, and behavioral learning theories were jointly generalized accordingly. The proposed model, to our knowledge, is the first one that possesses not only modeling all features of classical conditioning but also including all features with behavioral theories such as Pavlov, Watson, Guthrie, Thorndike and Skinner. Also, a microcontroller card (Arduino Mega 2560) was used to validate the applicability of the proposed model in robotics. Obtained results showed that this generalized model has a high capacity for modeling human learning. Then, the proposed learning model was further improved to be utilized as a machine learning method that can continuously learn similar to human being. The result obtained from the use of this method, in terms of computational cost and accuracy, showed that the proposed method can be successfully employed in machine learning, especially for time ordered datasets.

关键词： Confidence machine   Ensemble learning   Reject option   One class classifier   Machine learning  

摘要： Confidence machine learning method is an important research content in machine learning area and of significant importance to high-risk machine learning application area. In confidence machine learning, design and calculation of confidence level is a difficulty;nevertheless, the algorithm in this paper enables confidence classification even when specific threshold setting is omitted and confidence level calculation for each example and unknown sample is neglected. Based on ensemble learning structure, this algorithm employs twice one-class classifier to classify binary problems, and with reject option being set, confidence learning of binary classification is performed by means of multi-layer ensemble learning. The algorithm has been validated on eight experimental datasets, such as heart disease and diabetes mellitus, and good results have been achieved. The learning method proposed in this paper can make the confidence machine learning easier and more efficient. In the era of big data, the research of pattern recognition algorithm based on small sample is still of great significance.

关键词： gamma camera   position sensitive gamma detectors   scintillator characterization   maximum-likelihood estimation   depth-of-interaction (DOI)   manifold learning   locally linear embedding  

摘要： Measuring the depth-of-interaction (DOI) of gamma photons enables increasing the resolution of emission imaging systems. Several design variants of DOI-sensitive detectors have been recently introduced to improve the performance of scanners for positron emission tomography (PET). However, the accurate characterization of the response of DOI detectors, necessary to accurately measure the DOI, remains an unsolved problem. Numerical simulations are, at the state of the art, imprecise, while measuring directly the characteristics of DOI detectors experimentally is hindered by the impossibility to impose the depth-of-interaction in an experimental set-up. In this article we introduce a machine learning approach for extracting accurate forward models of gamma imaging devices from simple pencil-beam measurements, using a nonlinear dimensionality reduction technique in combination with a finite mixture model. The method is purely data-driven, not requiring simulations, and is applicable to a wide range of detector types. The proposed method was evaluated both in a simulation study and with data acquired using a monolithic gamma camera designed for PET (the cMiCE detector), demonstrating the accurate recovery of the DOI characteristics. The combination of the proposed calibration technique with maximum-a posteriori estimation of the coordinates of interaction provided a depth resolution of approximate to 1.14 mm for the simulated PET detector and approximate to 1.74 mm for the cMiCE detector. The software and experimental data are made available at http://***/depthembedding/.

关键词： Weather derivatives   Rainfall   Machine learning  

摘要： Regression problems provide some of the most challenging research opportunities in the area of machine learning, and more broadly intelligent systems, where the predictions of some target variables are critical to a specific application. Rainfall is a prime example, as it exhibits unique characteristics of high volatility and chaotic patterns that do not exist in other time series data. This work's main impact is to show the benefit machine learning algorithms, and more broadly intelligent systems have over the current state-of-the-art techniques for rainfall prediction within rainfall derivatives. We apply and compare the predictive performance of the current state-of-the-art (Markov chain extended with rainfall prediction) and six other popular machine learning algorithms, namely: Genetic Programming, Support Vector Regression, Radial Basis Neural Networks, M5 Rules, M5 Model trees, and k-Nearest Neighbours. To assist in the extensive evaluation, we run tests using the rainfall time series across data sets for 42 cities, with very diverse climatic features. This thorough examination shows that the machine learning methods are able to outperform the current state-of-the-art. Another contribution of this work is to detect correlations between different climates and predictive accuracy. Thus, these results show the positive effect that machine learning-based intelligent systems have for predicting rainfall based on predictive accuracy and with minimal correlations existing across climates. (C) 2017 Elsevier Ltd. All rights reserved.

关键词： 直流电弧   故障检测   特征提取   机器学习   支持向量机  

摘要： 为了解决传统分析方法在直流供电系统中电弧故障检测的精确度不足及过程繁琐的问题,将直流电弧故障检测归为二分类问题,引入机器学习方法,通过直流电弧实验得到正常状态和电弧状态的数据,从时域中提取电流均值等4个特征,从频域中提取高频分量标准差等3个特征。利用提取到的特征对支持向量机(SVM)进行训练,利用求解得到的模型对测试数据集进行分类,分类准确率为94.483%。结果证明:所提方法能有效检测直流电弧故障,提高故障检测精度,且步骤精简,易于推广。

关键词： automatic classification   Parkinsonian tremor severity   machine learning   clinical decision tool  

摘要： Motivation: Although clinical aspirations for new technology to accurately measure and diagnose Parkinsonian tremors exist, automatic scoring of tremor severity using machine learning approaches has not yet been employed. Objective: This study aims to maximize the scientific validity of automatic tremor-severity classification using machine learning algorithms to score Parkinsonian tremor severity in the same manner as the unified Parkinson's disease rating scale (UPDRS) used to rate scores in real clinical practice. Approach: Eighty-five PD patients perform four tasks for severity assessment of their resting, resting with mental stress, postural, and intention tremors. The tremor signals are measured using a wristwatch-type wearable device with an accelerometer and gyroscope. Displacement and angle signals are obtained by integrating the acceleration and angular-velocity signals. Nineteen features are extracted from each of the four tremor signals. The optimal feature configuration is decided using the wrapper feature selection algorithm or principal component analysis, and decision tree, support vector machine, discriminant analysis, and k-nearest neighbour algorithms are considered to develop an automatic scoring system for UPDRS prediction. The results are compared to UPDRS ratings assigned by two neurologists. Main results: The highest accuracies are 92.3%, 86.2%, 92.1%, and 89.2% for resting, resting with mental stress, postural, and intention tremors, respectively. The weighted Cohen's kappa values are 0.745, 0.635 and 0.633 for resting, resting with mental stress, and postural tremors (almost perfect agreement), and 0.570 for intention tremors (moderate). Significance: These results indicate the feasibility of the proposed system as a clinical decision tool for Parkinsonian tremor-severity automatic scoring.

关键词： Traffic crash severity prediction   Multinomial logit   Nearest neighbor classification   Support vector machines   Random forests   Crash costs  

摘要： Crash severity prediction models enable different agencies to predict the severity of a reported crash with unknown severity or the severity of crashes that may be expected to occur sometime in the future. This paper had three main objectives: comparison of the performance of four statistical and machine learning methods including Multinomial Logit (MNL), Nearest Neighbor Classification (NNC), Support Vector Machines (SVM) and Random Forests (RF), in predicting traffic crash severity;developing a crash costs-based approach for comparison of crash severity prediction methods;and investigating the effects of data clustering methods comprising K-means Clustering (KC) and Latent Class Clustering (LCC), on the performance of crash severity prediction models. The 2012-2015 reported crash data from Nebraska, United States was obtained and two-vehicle crashes were extracted as the analysis data. The dataset was split into training/estimation (2012-2014) and validation (2015) subsets. The four prediction methods were trained/estimated using the training/estimation dataset and the correct prediction rates for each crash severity level, overall correct prediction rate and a proposed crash costs based accuracy measure were obtained for the validation dataset. The correct prediction rates and the proposed approach showed NNC had the best prediction performance in overall and in more severe crashes. RF and SVM had the next two sufficient performances and MNL was the weakest method. Data clustering did not affect the prediction results of SVM, but KC improved the prediction performance of MNL, NNC and RF, while LCC caused improvement in MNL and RF but weakened the performance of NNC. Overall correct prediction rate had almost the exact opposite results compared to the proposed approach, showing that neglecting the crash costs can lead to misjudgment in choosing the right prediction method.