Examining Swarm Intelligence-based Feature Selection for Multi-Label Classification

2021 ◽  
Vol 2 (2) ◽  
Author(s):  
Awder Mohammed Ahmed ◽  
◽  
Adnan Mohsin Abdulazeez ◽  
Keyword(s):  
Feature Selection ◽  
Swarm Intelligence ◽  
Real World ◽  
Computational Cost ◽  
Computing Methods ◽  
Open Problems ◽  
Label Data ◽  
Filter Methods ◽  
Wide Range ◽  
Classification Tasks

Multi-label classification addresses the issues that more than one class label assigns to each instance. Many real-world multi-label classification tasks are high-dimensional due to digital technologies, leading to reduced performance of traditional multi-label classifiers. Feature selection is a common and successful approach to tackling this problem by retaining relevant features and eliminating redundant ones to reduce dimensionality. There is several feature selection that is successfully applied in multi-label learning. Most of those features are wrapper methods that employ a multi-label classifier in their processes. They run a classifier in each step, which requires a high computational cost, and thus they suffer from scalability issues. Filter methods are introduced to evaluate the feature subsets using information-theoretic mechanisms instead of running classifiers to deal with this issue. Most of the existing researches and review papers dealing with feature selection in single-label data. While, recently multi-label classification has a wide range of real-world applications such as image classification, emotion analysis, text mining, and bioinformatics. Moreover, researchers have recently focused on applying swarm intelligence methods in selecting prominent features of multi-label data. To the best of our knowledge, there is no review paper that reviews swarm intelligence-based methods for multi-label feature selection. Thus, in this paper, we provide a comprehensive review of different swarm intelligence and evolutionary computing methods of feature selection presented for multi-label classification tasks. To this end, in this review, we have investigated most of the well-known and state-of-the-art methods and categorize them based on different perspectives. We then provided the main characteristics of the existing multi-label feature selection techniques and compared them analytically. We also introduce benchmarks, evaluation measures, and standard datasets to facilitate research in this field. Moreover, we performed some experiments to compare existing works, and at the end of this survey, some challenges, issues, and open problems of this field are introduced to be considered by researchers in the future.

scholarly journals Swarm Intelligence-Based Feature Selection for Multi-Label Classification: A Review

2021 ◽  
pp. 65-78
Author(s):  
Adnan Mohsin Abdulazeez ◽  
Dathar A. Hasan ◽  
Awder Mohammed Ahmed ◽  
Omar S. Kareem
Keyword(s):  
Feature Selection ◽  
Swarm Intelligence ◽  
State Of The Art ◽  
Computing Methods ◽  
Open Problems ◽  
Wide Range ◽  
Real World Applications ◽  
Classification Tasks ◽  
Selection For ◽  
Feature Selection Techniques

Multi-label classification is the process of specifying more than one class label for each instance. The high-dimensional data in various multi-label classification tasks have a direct impact on reducing the efficiency of traditional multi-label classifiers. To tackle this problem, feature selection is used as an effective approach to retain relevant features and eliminating redundant ones to reduce dimensionality. Multi-label classification has a wide range of real-world applications such as image classification, emotion analysis, text mining and bioinformatics. Moreover, in recent years researchers have focused on applying swarm intelligence methods in selecting prominent features of multi-label data. After reviewing various researches, it seems there are no researches that provide a review of swarm intelligence-based methods for multi-label feature selection. Thus, in this study, a comprehensive review of different swarm intelligence and evolutionary computing methods of feature selection presented for the tasks of multi-label classification. To this end, in this review, we have investigated most of the well-known and state-of-the-art methods and categorize them based on different perspectives. We then provided the main characteristics of the existing multi-label feature selection techniques and compared them analytically. We also introduce benchmarks, evaluation measures and standard datasets to facilitate research in this field. Moreover, we performed some experiments to compare existing works and at the end of this survey, some challenges, issues and open problems of this field are introduced to be considered by researchers in future.

scholarly journals Performance Evaluation of Feature Selection Algorithms Applied to Online Learning in Concept Drift Environments

2018 ◽  
Author(s):  
Matheus B. De Moraes ◽  
André L. S. Gradvohl
Keyword(s):  
Feature Selection ◽  
Information Gain ◽  
Concept Drift ◽  
Computational Cost ◽  
Selection Algorithm ◽  
Information Need ◽  
High Speeds ◽  
Online Feature Selection ◽  
Classification Tasks ◽  
Selection Algorithms

Data streams are transmitted at high speeds with huge volume and may contain critical information need processing in real-time. Hence, to reduce computational cost and time, the system may apply a feature selection algorithm. However, this is not a trivial task due to the concept drift. In this work, we show that two feature selection algorithms, Information Gain and Online Feature Selection, present lower performance when compared to classification tasks without feature selection. Both algorithms presented more relevant results in one distinct scenario each, showing final accuracies up to 14% higher. The experiments using both real and artificial datasets present a potential for using these methods due to their better adaptability in some concept drift situations.

A VARIATIONAL BAYES BETA MIXTURE MODEL FOR FEATURE SELECTION IN DNA METHYLATION STUDIES

2013 ◽  
Vol 11 (04) ◽  
pp. 1350005 ◽  
Author(s):  
ZHANYU MA ◽  
ANDREW E. TESCHENDORFF
Keyword(s):  
Dna Methylation ◽  
Feature Selection ◽  
Mixture Model ◽  
Computational Cost ◽  
Methylation Data ◽  
Variational Bayesian ◽  
A Genome ◽  
Wide Range ◽  
Novel Biomarkers ◽  
Beta Mixture Model

An increasing number of studies are using beadarrays to measure DNA methylation on a genome-wide basis. The purpose is to identify novel biomarkers in a wide range of complex genetic diseases including cancer. A common difficulty encountered in these studies is distinguishing true biomarkers from false positives. While statistical methods aimed at improving the feature selection step have been developed for gene expression, relatively few methods have been adapted to DNA methylation data, which is naturally beta-distributed. Here we explore and propose an innovative application of a recently developed variational Bayesian beta-mixture model (VBBMM) to the feature selection problem in the context of DNA methylation data generated from a highly popular beadarray technology. We demonstrate that VBBMM offers significant improvements in inference and feature selection in this type of data compared to an Expectation-Maximization (EM) algorithm, at a significantly reduced computational cost. We further demonstrate the added value of VBBMM as a feature selection and prioritization step in the context of identifying prognostic markers in breast cancer. A variational Bayesian approach to feature selection of DNA methylation profiles should thus be of value to any study undergoing large-scale DNA methylation profiling in search of novel biomarkers.

scholarly journals Feature Selection Based on Divergence Functions: A Comparative Classiffication Study

2021 ◽  
Vol 9 (3) ◽  
pp. 587-606
Author(s):  
Saeid Pourmand ◽  
Ashkan Shabbak ◽  
Mojtaba Ganjali
Keyword(s):  
Feature Selection ◽  
Information Gain ◽  
Learning Model ◽  
Distribution Functions ◽  
Selection Methods ◽  
Accuracy Rate ◽  
Probability Distribution Functions ◽  
Filter Methods ◽  
Wide Range ◽  
Chi Squared

Due to the extensive use of high-dimensional data and its application in a wide range of scientifc felds of research, dimensionality reduction has become a major part of the preprocessing step in machine learning. Feature selection is one procedure for reducing dimensionality. In this process, instead of using the whole set of features, a subset is selected to be used in the learning model. Feature selection (FS) methods are divided into three main categories: flters, wrappers, and embedded approaches. Filter methods only depend on the characteristics of the data, and do not rely on the learning model at hand. Divergence functions as measures of evaluating the differences between probability distribution functions can be used as flter methods of feature selection. In this paper, the performances of a few divergence functions such as Jensen-Shannon (JS) divergence and Exponential divergence (EXP) are compared with those of some of the most-known flter feature selection methods such as Information Gain (IG) and Chi-Squared (CHI). This comparison was made through accuracy rate and F1-score of classifcation models after implementing these feature selection methods.

scholarly journals Estimation of camera-space manipulation parameters by means of an extended Kalman filter: Applications to parallel robots

2019 ◽  
Vol 16 (2) ◽  
pp. 172988141984298
Author(s):  
Alejandro González ◽  
Emilio J Gonzalez-Galvan ◽  
Mauro Maya ◽  
Antonio Cardenas ◽  
Davide Piovesan
Keyword(s):  
Kalman Filter ◽  
Extended Kalman Filter ◽  
High Speed ◽  
Computational Cost ◽  
Time Estimation ◽  
Industrial Applications ◽  
Parallel Robots ◽  
Filter Method ◽  
Open Problems ◽  
Wide Range

Parallel robots have a growing range of applications due to their appealing characteristics (high speed and acceleration, increased rigidity, etc.). However, several open problems make it difficult to model and control them. Low computational-cost algorithms are needed for high speed tasks where high accelerations are required. This article develops the nonlinear camera-space manipulation method and makes use of an extended Kalman filter (EKF) for the estimation of the camera-space manipulation parameters. This is presented as an alternative to the traditional method which can be time consuming while reaching convergence. The proposed camera-space manipulation parameter identification was performed in positioning tasks for a parallel manipulator and the experimental results are reported. Results show that it is possible to estimate the set of camera-space manipulation parameters by means of an extended Kalman filter. Using the proposed Kalman filter method we observed a significant reduction of the computational effort when estimating the camera-space manipulation parameters. However, there was no significant reduction of the robot’s positioning error. The proposed extended Kalman filter implementation requires only 2 ms to update the camera-space manipulation parameters compared to the 85 ms required by the traditional camera-space manipulation algorithm. Such time reduction is beneficial for the implementation of the method for a wide range of high speed and industrial applications. This article presents a novel use of an extended Kalman filter for the real-time estimation of the camera-space manipulation parameters and shows that it can be used to increase the positioning accuracy of a parallel robot.

scholarly journals Genetic Algorithm Ensemble Filter Methods on Kidney Disease Classification

2021 ◽  
Vol 11 (2) ◽  
pp. 73-80
Author(s):  
Sharin Hazlin Huspi ◽  
Chong Ke Ting
Keyword(s):  
Genetic Algorithm ◽  
Feature Selection ◽  
Nearest Neighbor ◽  
Information Gain ◽  
Computational Cost ◽  
Disease Classification ◽  
Support Vector ◽  
K Nearest Neighbor ◽  
Fisher Score ◽  
Filter Methods

Kidney failure will give effect to the human body, and it can lead to a series of seriously illness and even causing death. Machine learning plays important role in disease classification with high accuracy and shorter processing time as compared to clinical lab test. There are 24 attributes in the Chronic K idney Disease (CKD) clinical dataset, which is considered as too much of attributes. To improve the performance of the classification, filter feature selection methods used to reduce the dimensions of the feature and then the ensemble algorithm is used to identify the union features that selected from each filter feature selection. The filter feature selection that implemented in this research are Information Gain (IG), Chi-Squares, ReliefF and Fisher Score. Genetic Algorithm (GA) is used to select the best subset from the ensemble result of the filter feature selection. In this research, Random Forest (RF), XGBoost, Support Vector Machine (SVM), K-Nearest Neighbor (KNN) and Naïve Bayes classification techniques were used to diagnose the CKD. The features subset that selected are different and specialised for each classifier. By implementing the proposed method irrelevant features through filter feature selection able to reduce the burden and computational cost for the genetic algorithm. Then, the genetic algorithm able to perform better and select the best subset that able to improve the performance of the classifier with less attributes. The proposed genetic algorithm union filter feature selections improve the performance of the classification algorithm. The accuracy of RF, XGBoost, KNN and SVM can achieve to 100% and NB can achieve to 99.17%. The proposed method successfully improves the performance of the classifier by using less features as compared to other previous work.

scholarly journals Goal-driven active learning

2021 ◽  
Vol 35 (2) ◽  
Author(s):  
Nicolas Bougie ◽  
Ryutaro Ichise
Keyword(s):  
Decision Making ◽  
Reinforcement Learning ◽  
Learning Process ◽  
Real World ◽  
Imitation Learning ◽  
Learning Approaches ◽  
Wide Range ◽  
Fixed Set ◽  
Complex Decision Making ◽  
Complex Decision

AbstractDeep reinforcement learning methods have achieved significant successes in complex decision-making problems. In fact, they traditionally rely on well-designed extrinsic rewards, which limits their applicability to many real-world tasks where rewards are naturally sparse. While cloning behaviors provided by an expert is a promising approach to the exploration problem, learning from a fixed set of demonstrations may be impracticable due to lack of state coverage or distribution mismatch—when the learner’s goal deviates from the demonstrated behaviors. Besides, we are interested in learning how to reach a wide range of goals from the same set of demonstrations. In this work we propose a novel goal-conditioned method that leverages very small sets of goal-driven demonstrations to massively accelerate the learning process. Crucially, we introduce the concept of active goal-driven demonstrations to query the demonstrator only in hard-to-learn and uncertain regions of the state space. We further present a strategy for prioritizing sampling of goals where the disagreement between the expert and the policy is maximized. We evaluate our method on a variety of benchmark environments from the Mujoco domain. Experimental results show that our method outperforms prior imitation learning approaches in most of the tasks in terms of exploration efficiency and average scores.

A CFD Study of Focused Extreme Wave Impact on Decks of Offshore Structures

2014 ◽  
Author(s):  
Xin Lu ◽  
Pankaj Kumar ◽  
Anand Bahuguni ◽  
Yanling Wu
Keyword(s):  
Real World ◽  
Offshore Structures ◽  
Air Gap ◽  
Irregular Waves ◽  
Linear Wave ◽  
Extreme Wave ◽  
Wave Impact ◽  
Loading Test ◽  
Wide Range ◽  
Wave Maker

The design of offshore structures for extreme/abnormal waves assumes that there is sufficient air gap such that waves will not hit the platform deck. Due to inaccuracies in the predictions of extreme wave crests in addition to settlement or sea-level increases, the required air gap between the crest of the extreme wave and the deck is often inadequate in existing platforms and therefore wave-in-deck loads need to be considered when assessing the integrity of such platforms. The problem of wave-in-deck loading involves very complex physics and demands intensive study. In the Computational Fluid Mechanics (CFD) approach, two critical issues must be addressed, namely the efficient, realistic numerical wave maker and the accurate free surface capturing methodology. Most reported CFD research on wave-in-deck loads consider regular waves only, for instance the Stokes fifth-order waves. They are, however, recognized by designers as approximate approaches since “real world” sea states consist of random irregular waves. In our work, we report a recently developed focused extreme wave maker based on the NewWave theory. This model can better approximate the “real world” conditions, and is more efficient than conventional random wave makers. It is able to efficiently generate targeted waves at a prescribed time and location. The work is implemented and integrated with OpenFOAM, an open source platform that receives more and more attention in a wide range of industrial applications. We will describe the developed numerical method of predicting highly non-linear wave-in-deck loads in the time domain. The model’s capability is firstly demonstrated against 3D model testing experiments on a fixed block with various deck orientations under random waves. A detailed loading analysis is conducted and compared with available numerical and measurement data. It is then applied to an extreme wave loading test on a selected bridge with multiple under-deck girders. The waves are focused extreme irregular waves derived from NewWave theory and JONSWAP spectra.

scholarly journals Classification of unlabeled online media

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Sakthi Kumar Arul Prakash ◽  
Conrad Tucker
Keyword(s):  
Social Media ◽  
Real World ◽  
Graphical Model ◽  
Ground Truth ◽  
Classification Problem ◽  
Machine Learning Algorithms ◽  
Social Media Networks ◽  
Online Social Media ◽  
Wide Range

AbstractThis work investigates the ability to classify misinformation in online social media networks in a manner that avoids the need for ground truth labels. Rather than approach the classification problem as a task for humans or machine learning algorithms, this work leverages user–user and user–media (i.e.,media likes) interactions to infer the type of information (fake vs. authentic) being spread, without needing to know the actual details of the information itself. To study the inception and evolution of user–user and user–media interactions over time, we create an experimental platform that mimics the functionality of real-world social media networks. We develop a graphical model that considers the evolution of this network topology to model the uncertainty (entropy) propagation when fake and authentic media disseminates across the network. The creation of a real-world social media network enables a wide range of hypotheses to be tested pertaining to users, their interactions with other users, and with media content. The discovery that the entropy of user–user and user–media interactions approximate fake and authentic media likes, enables us to classify fake media in an unsupervised learning manner.

scholarly journals An information theoretic approach to link prediction in multiplex networks

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Seyed Hossein Jafari ◽  
Amir Mahdi Abdolhosseini-Qomi ◽  
Masoud Asadpour ◽  
Maseud Rahgozar ◽  
Naser Yazdani
Keyword(s):  
Real World ◽  
Link Prediction ◽  
Large Scale ◽  
Similarity Measures ◽  
Prediction Method ◽  
General Purpose ◽  
Fast Method ◽  
Theoretic Approach ◽  
Multiplex Networks ◽  
Wide Range

AbstractThe entities of real-world networks are connected via different types of connections (i.e., layers). The task of link prediction in multiplex networks is about finding missing connections based on both intra-layer and inter-layer correlations. Our observations confirm that in a wide range of real-world multiplex networks, from social to biological and technological, a positive correlation exists between connection probability in one layer and similarity in other layers. Accordingly, a similarity-based automatic general-purpose multiplex link prediction method—SimBins—is devised that quantifies the amount of connection uncertainty based on observed inter-layer correlations in a multiplex network. Moreover, SimBins enhances the prediction quality in the target layer by incorporating the effect of link overlap across layers. Applying SimBins to various datasets from diverse domains, our findings indicate that SimBins outperforms the compared methods (both baseline and state-of-the-art methods) in most instances when predicting links. Furthermore, it is discussed that SimBins imposes minor computational overhead to the base similarity measures making it a potentially fast method, suitable for large-scale multiplex networks.

Sign in / Sign up

Export Citation Format

Share Document