Showing papers in "Expert Systems With Applications in 2019"

A Comparison of Random Forest Variable Selection Methods for Classification Prediction Modeling.

Abstract: Random forest classification is a popular machine learning method for developing prediction models in many research settings. Often in prediction modeling, a goal is to reduce the number of variables needed to obtain a prediction in order to reduce the burden of data collection and improve efficiency. Several variable selection methods exist for the setting of random forest classification; however, there is a paucity of literature to guide users as to which method may be preferable for different types of datasets. Using 311 classification datasets freely available online, we evaluate the prediction error rates, number of variables, computation times and area under the receiver operating curve for many random forest variable selection methods. We compare random forest variable selection methods for different types of datasets (datasets with binary outcomes, datasets with many predictors, and datasets with imbalanced outcomes) and for different types of methods (standard random forest versus conditional random forest methods and test based versus performance based methods). Based on our study, the best variable selection methods for most datasets are Jiang's method and the method implemented in the VSURF R package. For datasets with many predictors, the methods implemented in the R packages varSelRF and Boruta are preferable due to computational efficiency. A significant contribution of this study is the ability to assess different variable selection techniques in the setting of random forest classification in order to identify preferable methods based on applications in expert and intelligent systems.

Deep Learning for Aspect-Based Sentiment Analysis: A Comparative Review

Abstract: The increasing volume of user-generated content on the web has made sentiment analysis an important tool for the extraction of information about the human emotional state. A current research focus for sentiment analysis is the improvement of granularity at aspect level, representing two distinct aims: aspect extraction and sentiment classification of product reviews and sentiment classification of target-dependent tweets. Deep learning approaches have emerged as a prospect for achieving these aims with their ability to capture both syntactic and semantic features of text without requirements for high-level feature engineering, as is the case in earlier methods. In this article, we aim to provide a comparative review of deep learning for aspect-based sentiment analysis to place different approaches in context.

Machine learning based phishing detection from URLs

Abstract: Due to the rapid growth of the Internet, users change their preference from traditional shopping to the electronic commerce. Instead of bank/shop robbery, nowadays, criminals try to find their victims in the cyberspace with some specific tricks. By using the anonymous structure of the Internet, attackers set out new techniques, such as phishing, to deceive victims with the use of false websites to collect their sensitive information such as account IDs, usernames, passwords, etc. Understanding whether a web page is legitimate or phishing is a very challenging problem, due to its semantics-based attack structure, which mainly exploits the computer users’ vulnerabilities. Although software companies launch new anti-phishing products, which use blacklists, heuristics, visual and machine learning-based approaches, these products cannot prevent all of the phishing attacks. In this paper, a real-time anti-phishing system, which uses seven different classification algorithms and natural language processing (NLP) based features, is proposed. The system has the following distinguishing properties from other studies in the literature: language independence, use of a huge size of phishing and legitimate data, real-time execution, detection of new websites, independence from third-party services and use of feature-rich classifiers. For measuring the performance of the system, a new dataset is constructed, and the experimental results are tested on it. According to the experimental and comparative results from the implemented classification algorithms, Random Forest algorithm with only NLP based features gives the best performance with the 97.98% accuracy rate for detection of phishing URLs.

Binary grasshopper optimisation algorithm approaches for feature selection problems

Abstract: Feature Selection (FS) is a challenging machine learning-related task that aims at reducing the number of features by removing irrelevant, redundant and noisy data while maintaining an acceptable level of classification accuracy. FS can be considered as an optimisation problem. Due to the difficulty of this problem and having a large number of local solutions, stochastic optimisation algorithms are promising techniques to solve this problem. As a seminal attempt, binary variants of the recent Grasshopper Optimisation Algorithm (GOA) are proposed in this work and employed to select the optimal feature subset for classification purposes within a wrapper-based framework. Two mechanisms are employed to design a binary GOA, the first one is based on Sigmoid and V-shaped transfer functions, and will be indicated by BGOA-S and BGOA-V, respectively. While the second mechanism uses a novel technique that combines the best solution obtained so far. In addition, a mutation operator is employed to enhance the exploration phase in BGOA algorithm (BGOA-M). The proposed methods are evaluated using 25 standard UCI datasets and compared with 8 well-regarded metaheuristic wrapper-based approaches, and six well known filter-based (e.g., correlation FS) approaches. The comparative results show the superior performance of the BGOA and BGOA-M methods compared to other similar techniques in the literature.

Binary butterfly optimization approaches for feature selection

Abstract: In this paper, binary variants of the Butterfly Optimization Algorithm (BOA) are proposed and used to select the optimal feature subset for classification purposes in a wrapper-mode. BOA is a recently proposed algorithm that has not been systematically applied to feature selection problems yet. BOA can efficiently explore the feature space for optimal or near-optimal feature subset minimizing a given fitness function. The two proposed binary variants of BOA are applied to select the optimal feature combination that maximizes classification accuracy while minimizing the number of selected features. In these variants, the native BOA is utilized while its continuous steps are bounded in a threshold using a suitable threshold function after squashing them. The proposed binary algorithms are compared with five state-of-the-art approaches and four latest high performing optimization algorithms. A number of assessment indicators are utilized to properly assess and compare the performance of these algorithms over 21 datasets from the UCI repository. The experimental results confirm the efficiency of the proposed approaches in improving the classification accuracy compared to other wrapper-based algorithms, which proves the ability of BOA algorithm in searching the feature space and selecting the most informative attributes for classification tasks.

CNNpred: CNN-based stock market prediction using a diverse set of variables

Abstract: Feature extraction from financial data is one of the most important problems in market prediction domain for which many approaches have been suggested. Among other modern tools, convolutional neural networks (CNN) have recently been applied for automatic feature selection and market prediction. However, in experiments reported so far, less attention has been paid to the correlation among different markets as a possible source of information for extracting features. In this paper, we suggest a CNN-based framework, that can be applied on a collection of data from a variety of sources, including different markets, in order to extract features for predicting the future of those markets. The suggested framework has been applied for predicting the next day’s direction of movement for the indices of S&P 500, NASDAQ, DJI, NYSE, and RUSSELL based on various sets of initial variables. The evaluations show a significant improvement in prediction’s performance compared to the state of the art baseline algorithms.

Multiple instance learning for histopathological breast cancer image classification

Abstract: Histopathological images are the gold standard for breast cancer diagnosis. During examination several dozens of them are acquired for a single patient. Conventional, image-based classification systems make the assumption that all the patient’s images have the same label as the patient, which is rarely verified in practice since labeling the data is expensive. We propose a weakly supervised learning framework and investigate the relevance of Multiple Instance Learning (MIL) for computer-aided diagnosis of breast cancer patients, based on the analysis of histopathological images. Multiple instance learning consists in organizing instances (images) into bags (patients), without the need to label all the instances. We compare several state-of-the-art MIL methods including the pioneering ones (APR, Diverse Density, MI-SVM, citation-kNN), and more recent ones such as a non parametric method and a deep learning based approach (MIL-CNN). The experiments are conducted on the public BreaKHis dataset which contains about 8000 microscopic biopsy images of benign and malignant breast tumors, originating from 82 patients. Among the MIL methods the non-parametric approach has the best overall results, and in some cases allows to obtain classification rates never reached by conventional (single instance) classification frameworks. The comparison between MIL and single instance classification reveals the relevance of the MIL paradigm for the task at hand. In particular, the MIL allows to obtain comparable or better results than conventional (single instance) classification without the need to label all the images.

Convolutional neural network improvement for breast cancer classification

Abstract: Traditionally, physicians need to manually delineate the suspected breast cancer area. Numerous studies have mentioned that manual segmentation takes time, and depends on the machine and the operator. The algorithm called Convolutional Neural Network Improvement for Breast Cancer Classification (CNNI-BCC) is presented to assist medical experts in breast cancer diagnosis in timely manner. The CNNI-BCC uses a convolutional neural network that improves the breast cancer lesion classification in order to help experts for breast cancer diagnosis. CNNI-BCC can classify incoming breast cancer medical images into malignant, benign, and healthy patients. The application of present algorithm can assist in classification of mammographic medical images into benign patient, malignant patient and healthy patient without prior information of the presence of a cancerous lesion. The presented method aims to help medical experts for the classification of breast cancer lesion through the implementation of convolutional neural network for the classification of breast cancer. CNNI-BCC can categorize incoming medical images as malignant, benign or normal patient with sensitivity, accuracy, area under the receiver operating characteristic curve (AUC) and specificity of 89.47%, 90.50%, 0.901 ± 0.0314 and 90.71% respectively.

A deep learning approach for real-time detection of atrial fibrillation

Abstract: Goal: To develop a robust and real-time approach for automatic detection of atrial fibrillation (AF) in long-term electrocardiogram (ECG) recordings using deep learning (DL). Method: An end-to-end model combining the Convolutional- and Recurrent-Neural Networks (CNN and RNN) was proposed to extract high level features from segments of RR intervals (RRIs) in order to classify them as AF or normal sinus rhythm (NSR). Results: The model was trained and validated on three different databases including a total of 89 subjects. It achieved a sensitivity and specificity of 98.98% and 96.95%, respectively, validated through a 5-fold cross-validation. Additionally, the proposed model was found to be computationally efficient and it was capable of analyzing 24 h of ECG recordings in less than one second. The proposed algorithm was also tested on the unseen datasets to examine its robustness in detecting AF for new recordings which resulted in 98.96% and 86.04% for specificity and sensitivity, respectively. Conclusion: Compared to the state-of-the-art models evaluated on standard benchmark ECG datasets, the proposed model produced better performance in detecting AF. Additionally, since the model learns features directly from the data, it avoids the need for clever/cumbersome feature engineering.

The Internet of Things: Review and theoretical framework

Abstract: The Internet of Things (IoT) global arena is massive and growing exponentially. Those in the emerging digital world have recently witnessed the proliferation and impact of IoT-enabled devices. The IoT has provided new opportunities in the technology arena while bringing several challenges to an increased level of concern. This research has both practical and theoretical impetus since IoT is still in its infancy, and yet it is considered by many as the most important technology initiative of today. This study includes a systematic review and synthesis of IoT related literature and the development of a theoretical framework and conceptual model. The review of the literature reveals that the number of applications that make use of the IoT has increased dramatically and spans areas from business and manufacturing to home, health care, and knowledge management. Although IoT can create invaluable data in every industry, it does not occur without its challenges. The theoretical framework developed identifies IoT priority areas and challenges, providing a guide for those leading IoT initiatives and revealing opportunities for future IoT research.

Literature review: Machine learning techniques applied to financial market prediction

Abstract: The search for models to predict the prices of financial markets is still a highly researched topic, despite major related challenges. The prices of financial assets are non-linear, dynamic, and chaotic; thus, they are financial time series that are difficult to predict. Among the latest techniques, machine learning models are some of the most researched, given their capabilities for recognizing complex patterns in various applications. With the high productivity in the machine learning area applied to the prediction of financial market prices, objective methods are required for a consistent analysis of the most relevant bibliography on the subject. This article proposes the use of bibliographic survey techniques that highlight the most important texts for an area of research. Specifically, these techniques are applied to the literature about machine learning for predicting financial market values, resulting in a bibliographical review of the most important studies about this topic. Fifty-seven texts were reviewed, and a classification was proposed for markets, assets, methods, and variables. Among the main results, of particular note is the greater number of studies that use data from the North American market. The most commonly used models for prediction involve support vector machines (SVMs) and neural networks. It was concluded that the research theme is still relevant and that the use of data from developing markets is a research opportunity.

A generalized mean distance-based k-nearest neighbor classifier

Abstract: K-nearest neighbor (KNN) rule is a well-known non-parametric classifier that is widely used in pattern recognition. However, the sensitivity of the neighborhood size k always seriously degrades the KNN-based classification performance, especially in the case of the small sample size with the existing outliers. To overcome this issue, in this article we propose a generalized mean distance-based k-nearest neighbor classifier (GMDKNN) by introducing multi-generalized mean distances and the nested generalized mean distance that are based on the characteristic of the generalized mean. In the proposed method, multi-local mean vectors of the given query sample in each class are calculated by adopting its class-specific k nearest neighbors. Using the achieved k local mean vectors per class, the corresponding k generalized mean distances are calculated and then used to design the categorical nested generalized mean distance. In the classification phase, the categorical nested generalized mean distance is used as the classification decision rule and the query sample is classified into the class with the minimum nested generalized mean distance among all the classes. Extensive experiments on the UCI and KEEL data sets, synthetic data sets, the KEEL noise data sets and the UCR time series data sets are conducted by comparing the proposed method to the state-of-art KNN-based methods. The experimental results demonstrate that the proposed GMDKNN performs better and has the less sensitiveness to k. Thus, our proposed GMDKNN with the robust and effective classification performance could be a promising method for pattern recognition in some expert and intelligence systems.

A group decision making sustainable supplier selection approach using extended TOPSIS under interval-valued Pythagorean fuzzy environment

Abstract: Due to the increasing awareness of environmental and social issues, sustainable supplier selection becomes an important problem. The aim of this paper is to develop a novel group decision making sustainable supplier selection approach using extended Techniques for Order Preferences by Similarity to Ideal Solution (TOPSIS) under interval-valued Pythagorean fuzzy environment. Sustainable supplier selection often involves uncertain information due to the subjective nature of human judgments, and the interval-valued Pythagorean fuzzy set (IVPFS) has great ability to address strong fuzziness, ambiguity and inexactness during the decision-making process. The first contribution of this research is to use the IVPFS to capture the uncertain information of decision makers. Moreover, sustainable supplier selection often involves multiple decision makers from different groups. The second contribution of this research is to develop a group decision making approach for sustainable supplier selection. TOPSIS is the most commonly used technique in sustainable supplier selection. The third contribution of this research is to propose an extended TOPSIS method by integrating distance and similarity between alternatives concurrently to evaluate performances of suppliers. In this research, the group decision making approach and extended TOPSIS method is also extended to IVPFSs. Finally, experiments are conducted to verify the feasibility and efficiency of the proposed sustainable supplier selection approach. Experiments results show that the proposed approach is effective and efficient to help decision makers to select optimal sustainable suppliers. Therefore, the proposed approach can be applied by managers to evaluate and determine appropriate suppliers in sustainable supplier selection process.

An efficient chaotic mutative moth-flame-inspired optimizer for global optimization tasks

Abstract: Moth-flame optimization algorithm (MFO) is a new nature-inspired meta-heuristic based on the navigation routine of moths in the environment known as transverse orientation. For some complex optimization tasks, especially high dimensional and multimodal problems, the conventional MFO may face problems in the convergence trends or be trapped into the local and deceptive optima. Therefore, in this study, two strategies have been introduced into the conventional MFO to get a more stable sense of balance between the exploration and exploitation propensities. First, Gaussian mutation is employed to increase the population diversity of MFO. Then, a chaotic local search is applied to the flame updating process of MFO for better exploiting the locality of the solutions. The proposed CLSGMFO approach was compared against a wide range of well-known classical metaheuristic algorithms (MAs) and various advanced MAs using 23 classical benchmark functions. It was shown that the designed CLSGMFO can outperform most of the popular MAs in terms of solution quality and convergence speed. Moreover, based on CLSGMFO, a hybrid kernel extreme learning machine model, which is called CLSGMFO-KELM, is established to deal with financial stress prediction scenarios. To investigate the effectiveness of the CLSGMFO-KELM model, the proposed hybrid system was tested on two widely used financial datasets and compared against a broad array of popular classifiers. The results demonstrate that the proposed learning scheme can offer a superior kernel extreme learning machine model with excellent predictive performance. Accordingly, the proposed CLSGMFO can serve as an effective and efficient computer-aided tool for financial prediction.

Sentiment analysis based on improved pre-trained word embeddings

Abstract: Sentiment analysis is a fast growing area of research in natural language processing (NLP) and text classifications. This technique has become an essential part of a wide range of applications including politics, business, advertising and marketing. There are various techniques for sentiment analysis, but recently word embeddings methods have been widely used in sentiment classification tasks. Word2Vec and GloVe are currently among the most accurate and usable word embedding methods which can convert words into meaningful vectors. However, these methods ignore sentiment information of texts and need a large corpus of texts for training and generating exact vectors. As a result, because of the small size of some corpora, researcher often have to use pre-trained word embeddings which were trained on other large text corpora such as Google News with about 100 billion words. The increasing accuracy of pre-trained word embeddings has a great impact on sentiment analysis research. In this paper, we propose a novel method, Improved Word Vectors (IWV), which increases the accuracy of pre-trained word embeddings in sentiment analysis. Our method is based on Part-of-Speech (POS) tagging techniques, lexicon-based approaches, word position algorithm and Word2Vec/GloVe methods. We tested the accuracy of our method via different deep learning models and benchmark sentiment datasets. Our experiment results show that Improved Word Vectors (IWV) are very effective for sentiment analysis.

A survey on wearable sensor modality centred human activity recognition in health care

Abstract: Increased life expectancy coupled with declining birth rates is leading to an aging population structure. Aging-caused changes, such as physical or cognitive decline, could affect people's quality of life, result in injuries, mental health or the lack of physical activity. Sensor-based human activity recognition (HAR) is one of the most promising assistive technologies to support older people's daily life, which has enabled enormous potential in human-centred applications. Recent surveys in HAR either only focus on the deep learning approaches or one specific sensor modality. This survey aims to provide a more comprehensive introduction for newcomers and researchers to HAR. We first introduce the state-of-art sensor modalities in HAR. We look more into the techniques involved in each step of wearable sensor modality centred HAR in terms of sensors, activities, data pre-processing, feature learning and classification, including both conventional approaches and deep learning methods. In the feature learning section, we focus on both hand-crafted features and automatically learned features using deep networks. We also present the ambient-sensor-based HAR, including camera-based systems, and the systems which combine the wearable and ambient sensors. Finally, we identify the corresponding challenges in HAR to pose research problems for further improvement in HAR.

Multi-output bus travel time prediction with convolutional LSTM neural network

Abstract: Accurate and reliable travel time predictions in public transport networks are essential for delivering an attractive service that is able to compete with other modes of transport in urban areas. The traditional application of this information, where arrival and departure predictions are displayed on digital boards, is highly visible in the city landscape of most modern metropolises. More recently, the same information has become critical as input for smart-phone trip planners in order to alert passengers about unreachable connections, alternative route choices and prolonged travel times. More sophisticated Intelligent Transport Systems (ITS) include the predictions of connection assurance, i.e. an expert system that will decide to hold services to enable passenger exchange, in case one of the services is delayed up to a certain level. In order to operate such systems, and to ensure the confidence of passengers in the systems, the information provided must be accurate and reliable. Traditional methods have trouble with this as congestion, and thus travel time variability, increases in cities, consequently making travel time predictions in urban areas a non-trivial task. This paper presents a system for bus travel time prediction that leverages the non-static spatio-temporal correlations present in urban bus networks, allowing the discovery of complex patterns not captured by traditional methods. The underlying model is a multi-output, multi-time-step, deep neural network that uses a combination of convolutional and long short-term memory (LSTM) layers. The method is empirically evaluated and compared to other popular approaches for link travel time prediction and currently available services, including the currently deployed model at Movia, the regional public transport authority in Greater Copenhagen. We find that the proposed model significantly outperforms all the other methods we compare with, and is able to detect small irregular peaks in bus travel times very quickly.

A hybrid self-adaptive sine cosine algorithm with opposition based learning

Abstract: Real-world optimization problems demand an efficient meta-heuristic algorithm which maintains the diversity of solutions and properly exploits the search space of the problem to find the global optimal solution. Sine Cosine Algorithm (SCA) is a recently developed population-based meta-heuristic algorithm for solving global optimization problems. SCA uses the characteristics of sine and cosine trigonometric functions to update the solutions. But, like other population-based optimization algorithms, SCA also suffers the problem of low diversity, stagnation in local optima and skipping of true solutions. Therefore, in the present work, an attempt has been made towards the eradication of these issues, by proposing a modified version of SCA. The proposed algorithm is named as modified Sine Cosine Algorithm (m-SCA). In m-SCA, the opposite population is generated using opposite numbers based on perturbation rate to jump out from the local optima. Secondly, in the search equations of SCA self-adaptive component is added to exploit all the promising search regions which are pre-visited. To evaluate the effectiveness in solving the global optimization problems, m-SCA has been tested on two sets of benchmark problems – classical set of 23 well-known benchmark problems and standard IEEE CEC 2014 benchmark test problems. In the paper, the performance of proposed algorithm m-SCA is also tested on five engineering optimization problems. The conducted statistical, convergence and average distance analysis demonstrate the efficacy of the proposed algorithm to determine the efficient solution of real-life global optimization problems.

Survey of conversational agents in health

Abstract: Artificial intelligence (AI) has transformed the world and the relationships among humans as the learning capabilities of machines have allowed for a new means of communication between humans and machines. In the field of health, there is much interest in new technologies that help to improve and automate services in hospitals. This article aims to explore the literature related to conversational agents applied to health care, searching for definitions, patterns, methods, architectures, and data types. Furthermore, this work identifies an agent application taxonomy, current challenges, and research gaps. In this work, we use a systematic literature review approach. We guide and refine this study and the research questions by applying Population, Intervention, Comparison, Outcome, and Context (PICOC) criteria. The present study investigated approximately 4145 articles involving conversational agents in health published over the last ten years. In this context, we finally selected 40 articles based on their approaches and objectives as related to our main subject. As a result, we developed a taxonomy, identified the main challenges in the field, and defined the main types of dialog and contexts related to conversational agents in health. These results contributed to discussions regarding conversational health agents, and highlighted some research gaps for future study.

Multi-objective multi-robot path planning in continuous environment using an enhanced genetic algorithm

Abstract: This paper presents a hybrid approach for path planning of multiple mobile robots in continuous environments. For this purpose, first, an innovative Artificial Potential Field (APF) algorithm is presented to find all feasible paths between the start and destination locations in a discrete gridded environment. Next, an enhanced Genetic Algorithm (EGA) is developed to improve the initial paths in continuous space and find the optimal path between start and destination locations. The proposed APF works based on a time-efficient deterministic scheme to find a set of feasible initial paths and is guaranteed to find a feasible path if one exists. The EGA utilizes five customized crossover and mutation operators to improve the initial paths. In this paper, path length, smoothness, and safety are combined to form a multi-objective path planning problem. In addition, the proposed method is extended to deal with multiple mobile robot path planning problem. For this purpose, a new term is added to the objective function which measures the distance between robots and a collision removal operator is added to the EGA to remove possible collision between paths. To assess the efficiency of the proposed algorithm, 12 planar environments with different sizes and complexities were examined. Evaluations showed that the control parameters of the proposed algorithm do not affect the performance of the EGA considerably. Moreover, a comparative study has been made between the proposed algorithm, A*, PRM, B-RRT and Particle Swarm Optimization (PSO). The comparative study showed that the proposed algorithm outperforms PSO as well as well-recognized deterministic (A*) and probabilistic (PRM and B-RRT) path planning algorithms in terms of path length, run time, and success rate. Finally, simulations proved the efficiency of the proposed algorithm for a four-robot path planning problem. In this case, not only the proposed algorithm determined collision-free paths, but also it found near optimal solution for all robots.

End-to-End Environmental Sound Classification using a 1D Convolutional Neural Network

Abstract: In this paper, we present an end-to-end approach for environmental sound classification based on a 1D Convolution Neural Network (CNN) that learns a representation directly from the audio signal. Several convolutional layers are used to capture the signal’s fine time structure and learn diverse filters that are relevant to the classification task. The proposed approach can deal with audio signals of any length as it splits the signal into overlapped frames using a sliding window. Different architectures considering several input sizes are evaluated, including the initialization of the first convolutional layer with a Gammatone filterbank that models the human auditory filter response in the cochlea. The performance of the proposed end-to-end approach in classifying environmental sounds was assessed on the UrbanSound8k dataset and the experimental results have shown that it achieves 89% of mean accuracy. Therefore, the proposed approach outperforms most of the state-of-the-art approaches that use handcrafted features or 2D representations as input. Moreover, the proposed approach outperforms all approaches that use raw audio signal as input to the classifier. Furthermore, the proposed approach has a small number of parameters compared to other architectures found in the literature, which reduces the amount of data required for training.

Process mining techniques and applications – A systematic mapping study

Abstract: Process mining is a growing and promising study area focused on understanding processes and to help capture the more significant findings during real execution rather than, those methods that, only observed idealized process model. The objective of this article is to map the active research topics of process mining and their main publishers by country, periodicals, and conferences. We also extract the reported application studies and classify these by exploration domains or industry segments that are taking advantage of this technique. The applied research method was systematic mapping, which began with 3713 articles. After applying the exclusion criteria, 1278 articles were selected for review. In this article, an overview regarding process mining is presented, the main research topics are identified, followed by identification of the most applied process mining algorithms, and finally application domains among different business segments are reported on. It is possible to observe that the most active research topics are associated with the process discovery algorithms, followed by conformance checking, and architecture and tools improvements. In application domains, the segments with major case studies are healthcare followed by information and communication technology, manufacturing, education, finance, and logistics.

Effective heuristics and metaheuristics to minimize total flowtime for the distributed permutation flowshop problem

Abstract: Distributed permutation flowshop scheduling problem (DPFSP) has become a very active research area in recent years. However, minimizing total flowtime in DPFSP, a very relevant and meaningful objective for today's dynamic manufacturing environment, has not captured much attention so far. In this paper, we address the DPFSP with total flowtime criterion. To suit the needs of different CPU time demands and solution quality, we present three constructive heuristics and four metaheuristics. The constructive heuristics are based on the well-known LR and NEH heuristics. The metaheuristics are based on the high-performing frameworks of discrete artificial bee colony, scatter search, iterated local search, and iterated greedy, which have been applied with great success to closely related scheduling problems. We explore the problem-specific knowledge and accelerations to evaluate neighboring solutions for the considered problem. We introduce advanced and effective technologies like a referenced local search, a strategy to escape from local optima, and an enhanced intensive search method for the presented metaheuristics. A comprehensive computational campaign against the closely related and well performing algorithms in the literature is carried out. The results show that both the presented constructive heuristics and metaheuristics are very effective for solving the DPFSP with total flowtime criterion.

Combining Principal Component Analysis, Discrete Wavelet Transform and XGBoost to trade in the financial markets

Abstract: When investing in financial markets it is crucial to determine a trading signal that can provide the investor with the best entry and exit points of the financial market, however this is a difficult task and has become a very popular research topic in the financial area. This paper presents an expert system in the financial area that combines Principal Component Analysis (PCA), Discrete Wavelet Transform (DWT), Extreme Gradient Boosting (XGBoost) and a Multi-Objective Optimization Genetic Algorithm (MOO-GA) in order to achieve high returns with a low level of risk. PCA is used to reduce the dimensionality of the financial input data set and the DWT is used to perform a noise reduction to every feature. The resultant data set is then fed to an XGBoost binary classifier that has its hyperparameters optimized by a MOO-GA. The importance of the PCA is analyzed and the results obtained show that it greatly improves the performance of the system. In order to improve even more the results obtained in the system using PCA, the PCA and the DWT are then applied together in one system and the results obtained show that this system is capable of outperforming the Buy and Hold (B&H) strategy in three of the five analyzed financial markets, achieving an average rate of return of 49.26% in the portfolio, while the B&H achieves on average 32.41%.

A comparison of fuzzy DEA and fuzzy TOPSIS in sustainable supplier selection: Implications for sourcing strategy

Abstract: This paper presents a comparative analysis of the outcomes achieved when two widely applied methods for supplier selection—the ‘technique for order of preference by similarity to ideal solution’ (TOPSIS) and ‘data envelopment analysis’—are applied to the problem of identifying the most preferred sustainable suppliers. Both fuzzy DEA and fuzzy TOPSIS are applied to a common dataset of logistics service providers in Sweden. The results reveal that TOPSIS outperforms DEA in terms of both calculation complexity and sensitivity to changes in the number of suppliers. However, output rankings from the two models are found to be less than perfectly correlated. The paper concludes that utilizing both methods, as applied to just a small number of evaluation criteria and a relatively low level of detail in the data, produces a useful pooled shortlist of potential sustainable suppliers. This can then form the basis for a second stage application where either of the methods may be applied to a greater number of criteria that are specified to a higher level of detail. Even more critically, the results also have the potential to point to specific aspects for discussion when negotiating price and service quality commitments with potential sustainable suppliers.

Technical analysis and sentiment embeddings for market trend prediction

Abstract: Stock market prediction is one of the most challenging problems which has been distressing both researchers and financial analysts for more than half a century. To tackle this problem, two completely opposite approaches, namely technical and fundamental analysis, emerged. Technical analysis bases its predictions on mathematical indicators constructed on the stocks price, while fundamental analysis exploits the information retrieved from news, profitability, and macroeconomic factors. The competition between these schools of thought has led to many interesting achievements, however, to date, no satisfactory solution has been found. Our work aims to combine both technical and fundamental analysis through the application of data science and machine learning techniques. In this paper, the stock market prediction problem is mapped in a classification task of time series data. Indicators of technical analysis and the sentiment of news articles are both exploited as input. The outcome is a robust predictive model able to forecast the trend of a portfolio composed by the twenty most capitalized companies listed in the NASDAQ100 index. As a proof of real effectiveness of our approach, we exploit the predictions to run a high frequency trading simulation reaching more than 80% of annualized return. This project represents a step forward to combine technical and fundamental analysis and provides a starting point for developing new trading strategies.

Effects of normalization on the entropy-based TOPSIS method

Abstract: In this paper, the frequently used normalization methods for the entropy method (EM) and the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) when these two methods are used in combination with each other are summarized. Taking information entropy (IE) as an indicator to measure the diversity of attribute data (DAD), the effects of normalization on the entropy-based TOPSIS method are analyzed. It is found that normalization can affect the decision result by affecting the DAD, while the DAD affects the contribution of attributes to the distance of each alternative from the ideal solution and the negative ideal solution, manifested in that the higher the DAD is, the bigger the contribution of the attribute will be. It is proved that vector normalization (VN) and sum normalization (SN) will not change the DAD, whereas min-max normalization (MMN) not only will change the DAD, but also may cause the appearance of numerous zero values, finally resulting in the fact that the calculated results of IE cannot represent the diversity of raw data. Therefore, normalization is not suggested for the EM, and VN is suggested for TOPSIS method. Some discussions on the combinability between the EM and TOPSIS method are also given as well as the applicability of different normalization methods in TOPSIS method.

An Interpretable Deep Hierarchical Semantic Convolutional Neural Network for Lung Nodule Malignancy Classification.

Abstract: While deep learning methods have demonstrated performance comparable to human readers in tasks such as computer-aided diagnosis, these models are difficult to interpret, do not incorporate prior domain knowledge, and are often considered as a "black-box." The lack of model interpretability hinders them from being fully understood by end users such as radiologists. In this paper, we present a novel interpretable deep hierarchical semantic convolutional neural network (HSCNN) to predict whether a given pulmonary nodule observed on a computed tomography (CT) scan is malignant. Our network provides two levels of output: 1) low-level semantic features; and 2) a high-level prediction of nodule malignancy. The low-level outputs reflect diagnostic features often reported by radiologists and serve to explain how the model interprets the images in an expert-interpretable manner. The information from these low-level outputs, along with the representations learned by the convolutional layers, are then combined and used to infer the high-level output. This unified architecture is trained by optimizing a global loss function including both low- and high-level tasks, thereby learning all the parameters within a joint framework. Our experimental results using the Lung Image Database Consortium (LIDC) show that the proposed method not only produces interpretable lung cancer predictions but also achieves significantly better results compared to using a 3D CNN alone.

A systematic survey of computer-aided diagnosis in medicine: Past and present developments

Abstract: Computer-aided diagnosis (CAD) in medicine is the result of a large amount of effort expended in the interface of medicine and computer science. As some CAD systems in medicine try to emulate the diagnostic decision-making process of medical experts, they can be considered as expert systems in medicine. Furthermore, CAD systems in medicine may process clinical data that can be complex and/or massive in size. They do so in order to infer new knowledge from data and use that knowledge to improve their diagnostic performance over time. Therefore, such systems can also be viewed as intelligent systems because they use a feedback mechanism to improve their performance over time. The main aim of the literature survey described in this paper is to provide a comprehensive overview of past and current CAD developments. This survey/review can be of significant value to researchers and professionals in medicine and computer science. There are already some reviews about specific aspects of CAD in medicine. However, this paper focuses on the entire spectrum of the capabilities of CAD systems in medicine. It also identifies the key developments that have led to today's state-of-the-art in this area. It presents an extensive and systematic literature review of CAD in medicine, based on 251 carefully selected publications. While medicine and computer science have advanced dramatically in recent years, each area has also become profoundly more complex. This paper advocates that in order to further develop and improve CAD, it is required to have well-coordinated work among researchers and professionals in these two constituent fields. Finally, this survey helps to highlight areas where there are opportunities to make significant new contributions. This may profoundly impact future research in medicine and in select areas of computer science.

Bankruptcy prediction using imaged financial ratios and convolutional neural networks

Abstract: Convolutional neural networks are being applied to identification problems in a variety of fields, and in some areas are showing higher discrimination accuracies than conventional methods However, applications of convolutional neural networks to financial analyses have only been reported in a small number of studies on the prediction of stock price movements The reason for this seems to be that convolutional neural networks are more suitable for application to images and less suitable for general numerical data including financial statements Hence, in this research, an attempt is made to apply a convolutional neural network to the prediction of corporate bankruptcy, which in most cases is treated as a two-class classification problem We use the financial statements (balance sheets and profit-and-loss statements) of 102 companies that have been delisted from the Japanese stock market due to de facto bankruptcy as well as the financial statements of 2062 currently listed companies over four financial periods In our proposed method, a set of financial ratios are derived from the financial statements and represented as a grayscale image The image generated by this process is utilized for training and testing a convolutional neural network Moreover, the size of the dataset is increased using the weighted averages to create synthetic data points A total of 7520 images for the bankrupt and continuing enterprises classes are used for training the convolutional neural network based on GoogLeNet Bankruptcy predictions through the trained network are shown to have a higher performance compared to methods using decision trees, linear discriminant analysis, support vector machines, multi-layer perceptron, AdaBoost, or Altman’s Z′′-score