Abstracts:This paper presents an evolving cloud-based algorithm for the recognition of drivers’ actions. The general idea is to detect different manoeuvres by processing the standard signals that are usually measured in a car, such as the speed, the revolutions, the angle of the steering wheel, the position of the pedals, and others, without additional intelligent sensors. The primary goal of this investigation is to propose a concept that can be used to recognise various driver actions. All experiments are performed on a realistic car simulator. The data acquired from the simulator are pre-processed and then used in the evolving cloud-based algorithm to detect the basic elementary actions, which are then combined in a prescribed sequence to create tasks. Finally, the sequences of different tasks form the most complex action, which is called a manoeuvre. As shown in this paper, the evolving cloud-based algorithm can be very efficiently used to recognise the complex driver’s action from raw signals obtained by typical car sensors.

Abstracts:Conventional data envelopment analysis (DEA) models treat the decision-making units (DMUs) as black-boxes: inputs enter the system and outputs exit the system, with no consideration for the intermediate steps characterizing the DMUs. As a result, intermediate measures are lost in the process of changing the inputs to outputs and it becomes difficult, if not impossible, to provide individual DMU managers with specific information on what part of a DMU is responsible for the overall inefficiency. This study defines a two-stage DEA model, where each DMU is composed of two sub-DMUs in series, the intermediate products by the sub-DMU in Stage 1 are partly consumed by the sub-DMU in Stage 2, and the initial inputs of the DMU can be freely allocated in both stages. Also, there are additional inputs directly consumed in Stage 2 while part of the outputs of Stage 1 are final outputs. We develop four new linear models to determine the upper and lower bounds of the efficiencies of the two sub-DMUs in a non-cooperative setting and a linear model to calculate the overall efficiency of DMU in a cooperative setting. That is, the overall efficiency of a DMU is modelled in a cooperative setting via upper and lower bounds obtained in the non-cooperative one. The proposed two-stage DEA method allows for important applications to several management areas. A case study in the banking industry is presented to demonstrate the applicability and exhibit the efficacy of the proposed models.

Abstracts:MinHash is a widely-used method for efficiently estimating the amount of similarity between documents for Near-Duplicate Detection (NDD). However, it is based on the concept of set resemblance rather than near-duplication. In this study, Sectional MinHash (S-MinHash), specifically designed for the detection of near-duplicate documents, is proposed. The proposed method enhances the MinHash data structure with information about the location of the attributes in the document. The method provides an unbiased estimate of the Jaccard coefficient with a smaller variance as compared to the MinHash for same signature sizes. The experiment results showed that the Mean Squared Error (MSE) of the proposed method was around one eighth of the MSE of the MinHash. Also, document NDD with the proposed method resulted in more accuracy in compare to the MinHash and the recent method, the BitHash. The best-captured F-measure was 87.05%. Setting the number of sections s to 2 gave the best results for the tested dataset.

Abstracts:Document fraud constitutes a growing problem in international shipping. Shipping documentation may be deliberately manipulated to avoid shipping restrictions or customs duties. Well-known examples of such fraud are miscoding and smuggling. These are cases in which the documentation of a shipment does not correctly or entirely describe the goods in transit. In an attempt to reduce the risks of document fraud, shipping companies and customs authorities typically perform random audits to check the accompanying documentation of shipments. Although these audits detect many fraud schemes, they are quite labor intensive and do not scale to the massive amounts of cargo that is shipped each day. This paper investigates whether intelligent fraud detection systems can improve the detection of miscoding and smuggling by analyzing large sets of historical shipment data. We develop a Bayesian network that predicts the presence of goods on the cargo list of shipments. The predictions of the Bayesian network are compared with the accompanying documentation of a shipment to determine whether document fraud is perpetrated. We also show how a set of discriminative models can be derived from the topology of the Bayesian network and perform the same fraud detection task. Our experimental results show that intelligent fraud detection systems can considerably improve the detection of miscoding and smuggling compared to random audits.

Abstracts:Demand Side Management (DSM) has been a subject of much research due to the demand of strategies which enable the development of power generation, transmission and distribution in Smart Grids (SG). Throughout the years, many authors have applied mathematical programming and, or, metaheuristic methods for obtaining a satisfactory solution for load management, mainly considering only one performance function in their respective optimization approaches. However, in addition to maximum load peak minimization (an often used performance function), it may be interesting to deal with some other merit functions, such as energy production cost, costumer’s preferences and constraints, among others. Due to this multi-criteria nature of this problem, in this paper it is proposed an exact multi-objective methodology to the optimal Demand Side Management. Essentially, the DSM problem is addressed by using an ϵ-constraint approach, in which one of the objectives is handled with Quadratic Programming whilst the others are considered as linear constrains. This paper also discusses the behavior of the solutions under the presence of uncertainties on the problem parameters due to possible unpredictable variations of the consumption pattern, energy price and number of appliances. The results, regarding the case studies analyzed, have shown promising solutions with interesting trade-off relations between power consumption and energy production cost. Moreover, the robustness analysis has indicated not sensitive solutions even when some parameters of the problem are varied, such as the energy price and the number of types of appliances, which can be regarded as a useful information for the utility.

Abstracts:In this article, a novel approach is proposed for automatic detection and diagnosis of mammographic masses, one of the common signs of non-palpable breast cancer. However, detection and diagnosis of mass are difficult due to its irregular shape, variability in size, and occlusion within breast tissue. The main aim of this study is to classify masses into benign and malignant after detecting them automatically. We propose an iterative method of high-to-low intensity thresholding controlled by radial region growing for the detection of masses. Based on the observation that in presence of mass orientation of tissue patterns changes, which may differ from benign to malignant, a multi resolution analysis of orientation of tissue patterns is then performed to categorize them. The performance of the proposed algorithm is evaluated with images from the digital database for screening mammography (DDSM), containing 450 benign masses, 440 malignant masses, and 410 normal images. A sensitivity of 85.0% is achieved at 1.4 false positives per image in mass detection, whereas an area under the receiver operating characteristic curve of 0.92 with an accuracy of 83.30% is achieved for the diagnosis of malignant masses.

Abstracts:Classification of gender from images of handwriting is an interesting research problem in computerized analysis of handwriting. The correlation between handwriting and gender of writer can be exploited to develop intelligent systems to facilitate forensic experts, document examiners, paleographers, psychologists and neurologists. We propose a handwriting based gender recognition system that exploits texture as the discriminative attribute between male and female handwriting. The textural information in handwriting is captured using combinations of different configurations of oriented Basic Image Features (oBIFs). oBIFs histograms and oBIFs columns histograms extracted from writing samples of male and female handwriting are used to train a Support Vector Machine classifier (SVM). The system is evaluated on three subsets of the QUWI database of Arabic and English writing samples using the experimental protocols of the ICDAR 2013, ICDAR 2015 and ICFHR 2016 gender classification competitions reporting classification rates of 71%, 76% and 68% respectively; outperforming the participating systems of these competitions. While textural measures like local binary patterns, histogram of oriented gradients and Gabor filters etc. have remained a popular choice for many expert systems targeting recognition problems, the present study demonstrates the effectiveness of relatively less investigated oBIFs as a robust textual descriptor.

Abstracts:During the last decade, sampling based methods for motion and path planning have gained more interest. Specifically, in the field of robotics, approaches based on the Rapidly-exploring Random Tree (RRT) algorithm have become the customary technique for solving the single-query motion planning problem. However, dynamic large maps still represent a challenging scenario for these methods to produce fast enough results. Taking advantage of an NVidia CUDA-enabled Graphic Processing Unit (GPU), we present quad-RRT, an extension of the bi-directional strategy to speed up the RRT when dealing with large-scale, bidimensional (2D) maps. Designed for modern GPUs, quad-RRT computes four trees instead of the two ones built by the bidirectional approaches. This modification aims balancing the direct searching ability of these methods with the parallel exploration of those parts of the map at both sides of the path joining the initial and goal poses. Experimental results demonstrate that the proposed algorithm provides a significant speedup dealing with large-scale maps densely populated by obstacles, when compared to other implementations of the RRT. Hence, the algorithm can have a high impact in the field of inspection path planning for distributed infrastructure. It is also a promising approach to allow new generation robots, designed to work in unconstrained environments, dynamically plan large-scale paths.

Abstracts:ε-support vector regression (ε-SVR), as a direct implementation of the structural risk minimization principle rather than empirical risk minimization principle, is a new regression method with good generalization ability and can efficiently solve small-sample learning problems. In this work, through incorporating gradient information into the traditional ε-SVR, the gradient-enhanced ε-SVR (GESVR) is developed. The efficiency of GESVR is compared with the traditional ε-SVR by employing analytical function fitting, compared with the gradient-enhanced least square support vector regression (GELSSVR) by using two real-life examples, and tested in a scenario where the exact gradient information is unknown. The results show that GESVR provides more accurate prediction results than the traditional ε-SVR model, and outperforms GELSSVR in some real-life cases.

Abstracts:Classification procedure is an important task of expert and intelligent systems. Developing new algorithms of classification which improve accuracy or true positive rates could have an influence on some life problems such as diagnosis prediction in medical domain. Multi-criteria decision making (MCDM) methods are expected to search the best alternative according to some criteria. Each criterion has a value relative to each alternative. There are only two sets: a set of criteria and a set of alternatives. This work merges MCDM methods TOPSIS and VIKOR and modifies them to be used for classification where the used sets are three: the classes, the objects and the attributes (features) describing the objects. Hence, ATOVIC, a new classification algorithm is proposed. In ATOVIC, criteria are replaced by features and alternatives are replaced by objects. The latter belong to corresponding classes. Two sets are employed one serves as reference and second serves as test. An object from test set will be classified to the relative class based on the reference set. ATOVIC is applied on a benchmark (UCI) CLEVELAND data set to predict heart disease. Following the complexity of the data set and its importance, ATOVIC application is done on different test sets of CLEVELAND using binary classification and multi-classification. Moreover, ATOVIC is applied to thyroid data set to detect hyperthyroidism and hypothyroidism diseases. The obtained results show the efficiency of ATOVIC in medical domain. In addition, ATOVIC is applied to three other data sets: chess, nursery and titanic, from UCI and KEEL websites. The obtained results are compared to those of some classifiers from literature. The experimental results demonstrate that ATOVIC method improves accuracy and true positive rates comparing to most classifiers considered from literature. Hence ATOVIC is promising for use in prediction or classification.