Standard Classification Algorithms Research Articles

Application based on Hybrid CNN-SVM and PCA-SVM Approaches for Classification of Cocoa Beans

In our study, we propose a hybrid Convolutional Neural Network with Support Vector Machine (CNN-SVM) and Principal Component Analysis with support vector machine (PCA-SVM) methods for the classification of cocoa beans obtained by the fermentation of beans collected from cocoa pods after harvest. We also use a convolutional neural network (CNN) and support vector machine (SVM) for the classification operation. In the case of the hybrid model, we use a convolutional network as a feature extractor and the SVM is used to perform the classification operation. The use of PCA-SVM allowed for a reduction in image size while maintaining the main features still using the SVM classifier. Radial, linear and polynomial basis function kernels were used with various control parameters for the SVM, and optimizers such as the Stochastic Gradient Descent (SGD) algorithm, Adam, and RMSprop were used for the CNN softmax classifier. The results showed the robustness of the hybrid CNN-SVM model which obtained the best score with a value of 98.32% then the PCA-SVM based model had a score of 97.65% outperforming the standard CNN and SVM classification algorithms. Metrics such as accuracy, recall, F1 score, mean squared error (MSE), and MCC have allowed us to consolidate the results obtained from our different experiments.

Domain-Driven Actionable Knowledge Discovery for Traffic Accidents Using Rules Induction

Due to the limitation of the methodologies of traditional data mining to satisfy business expectations, the shift from mining data-centered hidden patterns to domain-driven actionable knowledge discovery has become a significant direction of KDD research [22]. Traditional data mining algorithms and tools face major obstacles and challenges to solve real-life business problems and issues as they fail to provide actions that can be taken by people in business based on generated rules [22]. A small set of rules are generated by standard classification algorithms to form a classifier, but these classification algorithms use domain independent biases and heuristics [2]. This research aimed to propose a new approach to find actionable rules from sets of discovered rules. It focused on how a combination of traditional classification data mining and domain-driven data mining approach could be applied in solving real-life problems related to the field of traffic accidents in UAE. Real-life data were collected and pre-processed using the user’s existing knowledge and needs. Classification using Rules Induction was applied on the domain-driven dataset. The discovered rules from this technique were then summarized, combined, and analyzed. The final set of actionable rules from Classification technique for each class was then generated using a proposed interestingness method. To support such a process, the domain driven in-depth pattern discovery (DDID-PK) framework was followed [9]. Based on experimental results, the extracted domain-driven rules were more interesting and actionable than those produced by the traditional classification technique of data mining. In addition, the integration of data-centered classification technique of data mining to domain-driven approach of data mining and actionable knowledge discovery could help the Dubai police authority to reduce traffic accident severity by formulating new policies and traffic rules based on the domain-driven knowledge extracted from some hidden patterns from real data.

End-to-End Neural Network for Feature Extraction and Cancer Diagnosis of In Vivo Fluorescence Lifetime Images of Oral Lesions.

In contrast to previous studies that focused on classical machine learning algorithms and hand-crafted features, we present an end-to-end neural network classification method able to accommodate lesion heterogeneity for improved oral cancer diagnosis using multispectral autofluorescence lifetime imaging (maFLIM) endoscopy. Our method uses an autoencoder framework jointly trained with a classifier designed to handle overfitting problems with reduced databases, which is often the case in healthcare applications. The autoencoder guides the feature extraction process through the reconstruction loss and enables the potential use of unsupervised data for domain adaptation and improved generalization. The classifier ensures the features extracted are task-specific, providing discriminative information for the classification task. The data-driven feature extraction method automatically generates task-specific features directly from fluorescence decays, eliminating the need for iterative signal reconstruction. We validate our proposed neural network method against support vector machine (SVM) baselines, with our method showing a 6.5%-8.3% increase in sensitivity. Our results show that neural networks that implement data-driven feature extraction provide superior results and enable the capacity needed to target specific issues, such as inter-patient variability and the heterogeneity of oral lesions.Clinical relevance- We improve standard classification algorithms for in vivo diagnosis of oral cancer lesions from maFLIm for clinical use in cancer screening, reducing unnecessary biopsies and facilitating early detection of oral cancer.

A Random Forest with Minority Condensation and Decision Trees for Class Imbalanced Problems

Building an effective classifier that could classify a target or class of instances in a dataset from historical data has played an important role in machine learning for a decade. The standard classification algorithm has difficulty generating an appropriate classifier when faced with an imbalanced dataset. In 2019, the efficient splitting measure, minority condensation entropy (MCE) [1] is proposed that could build a decision tree to classify minority instances. The aim of this research is to extend the concept of a random forest to use both decision trees and minority condensation trees. The algorithm will build a minority condensation tree from a bootstrapped dataset maintaining all minorities while it will build a decision tree from a bootstrapped dataset of a balanced dataset. The experimental results on synthetic datasets apparent the results that confirm this proposed algorithm compared with the standard random forest are suitable for dealing with the binary-class imbalanced problem. Furthermore, the experiment on real-world datasets from the UCI repository shows that this proposed algorithm constructs a random forest that outperforms other existing random forest algorithms based on the recall, the precision, the F-measure, and the Geometric mean

Risky Driver Recognition with Class Imbalance Data and Automated Machine Learning Framework.

Identifying high-risk drivers before an accident happens is necessary for traffic accident control and prevention. Due to the class-imbalance nature of driving data, high-risk samples as the minority class are usually ill-treated by standard classification algorithms. Instead of applying preset sampling or cost-sensitive learning, this paper proposes a novel automated machine learning framework that simultaneously and automatically searches for the optimal sampling, cost-sensitive loss function, and probability calibration to handle class-imbalance problem in recognition of risky drivers. The hyperparameters that control sampling ratio and class weight, along with other hyperparameters, are optimized by Bayesian optimization. To demonstrate the performance of the proposed automated learning framework, we establish a risky driver recognition model as a case study, using video-extracted vehicle trajectory data of 2427 private cars on a German highway. Based on rear-end collision risk evaluation, only 4.29% of all drivers are labeled as risky drivers. The inputs of the recognition model are the discrete Fourier transform coefficients of target vehicle’s longitudinal speed, lateral speed, and the gap between the target vehicle and its preceding vehicle. Among 12 sampling methods, 2 cost-sensitive loss functions, and 2 probability calibration methods, the result of automated machine learning is consistent with manual searching but much more computation-efficient. We find that the combination of Support Vector Machine-based Synthetic Minority Oversampling TEchnique (SVMSMOTE) sampling, cost-sensitive cross-entropy loss function, and isotonic regression can significantly improve the recognition ability and reduce the error of predicted probability.

Medical Quora Tagging using MATAR and LDA Algorithm

The success of clustering or classification methods the detection of relevant textual formats is incredibly meaningful. The high dimensionality and irrelevance of textual materials was subjected to text records. Existing methods lack integration and are particularly vulnerable to original value. Metaheuristic algorithms are also applied to solve the challenges of standard classification algorithms. In this paper, an enhanced Latent Dirichlet Assignment clustering method & Inter Modeling for Tag Suggestion rating system is documented to boost correlation - based & identification efficiency to suggest labels with material modern web labels that promotes the exchange of medical information using unmonitored data through question-answering. For accurate tagging, Methods like POS marking, Hopping, Whistles& Stopping words are being used for speech recognition. The efficiency of the evolved architectures is compared to the standard methods, by using specificity of the recommendation, defining features, sensitivity, plain word and speed. The findings reveal that the classification and grouping scheme of the proposed structure succeeds traditional textual record approaches.

An empirical study of data intrinsic characteristics that make learning from imbalanced data difficult

Learning from data stemming from real-world problems is inherently challenging and difficult due to the numerous intrinsic characteristics present in datasets. The problem of class imbalance is known to significantly impair classification performance and has attracted increasing attention from researchers. On the other hand, some studies suggest that the detrimental effects of class imbalance occur only when the dataset encompasses other intrinsic characteristics such as small disjuncts, class overlapping, noise or data rarity. However, the literature is often ambiguous in terms of understanding and distinguishing the influence of these characteristics on the behaviour of standard classification algorithms. This paper provides a contemporary empirical study of the behaviour and performance of five well-known classifiers on a large number of imbalanced datasets exhibiting numerous combinations of the stated characteristics. The aim of the study is to identify and rank difficulty factors when learning from imbalanced data, depending on the type of classification algorithm used. In general, the obtained results suggest that if classifiers conceptually have no problem with class separation into sub-concepts, noise is the characteristic that most impairs their performance, closely followed by class overlapping and class imbalance. To alleviate these problems, oversampling and undersampling procedures were tested and directions are given for selecting appropriate techniques when dealing with the problem of class imbalance.

Comparing the Performance of AdaBoost, XGBoost, and Logistic Regression for Imbalanced Data

An imbalanced data problem occurs in the absence of a good class distribution between classes. Imbalanced data will cause the classifier to be biased to the majority class as the standard classification algorithms are based on the belief that the training set is balanced. Therefore, it is crucial to find a classifier that can deal with imbalanced data for any given classification task. The aim of this research is to find the best method among AdaBoost, XGBoost, and Logistic Regression to deal with imbalanced simulated datasets and real datasets. The performances of these three methods in both simulated and real imbalanced datasets are compared using five performance measures, namely sensitivity, specificity, precision, F1-score, and g-mean. The results of the simulated datasets show that logistic regression performs better than AdaBoost and XGBoost in highly imbalanced datasets, whereas in the real imbalanced datasets, AdaBoost and logistic regression demonstrated similarly good performance. All methods seem to perform well in datasets that are not severely imbalanced. Compared to AdaBoost and XGBoost, logistic regression is found to predict better for datasets with severe imbalanced ratios. However, all three methods perform poorly for data with a 5% minority, with a sample size of n = 100. In this study, it is found that different methods perform the best for data with different minority percentages.

An automatic sampling ratio detection method based on genetic algorithm for imbalanced data classification

Imbalanced data are a common phenomenon in both theoretical research and real-world applications. At a data level, standard classification algorithms cannot effectively learn and make predictions from imbalanced data, and this problem is generally solved by using oversampling, undersampling, or hybrid sampling methods. However, most of the current sampling methods use random sampling ratios, and the resulting classification performance can be undesirable and unstable. To obtain satisfactory and stable classification performance, we proposed three algorithms to automatically determine the sampling ratios for oversampling, undersampling, and hybrid sampling methods, based on a genetic algorithm. Experiments were performed to test the algorithms’ effectiveness by utilizing five widely used standard classification algorithms on 14 different imbalanced datasets using two oversampling, two undersampling, and four hybrid sampling methods. The statistical test results showed that for all five standard classification algorithms, sampling methods that used our proposed algorithms achieved the best classification results. Using area under the receiver operating characteristic curve (AUC) as the evaluation metric, it was demonstrated that the proposed algorithms for automatically determining the sampling ratio outperformed the random sampling ratio.

WITHDRAWN: SMO-RF:A machine learning approach by random forest for predicting class imbalancing followed by SMOTE

In different fields, such as machine learning and data mining, class imbalances have been one of the most complex issues for the past few decades. The unique condition of an imbalanced dataset that distributes each class of a particular dataset unevenly. The positive class is slightly smaller than the negative one. Many standard classification algorithms in this case do not classify instances related to the positive class. Typically the main goal of the classification task is a positive class. To deal with this problem, several approaches were proposed, for example sampling dependent over-sampling, undersampling, classification level enhancements, or the combination of two or more classifiers. The major problem however is that most solutions have a negative class, a computational cost, a storage problem, or a long training period. Data upsampling or downsampling may resolve a possible solution to the issue of skewness of data. In this paper, a hybrid technique is presented, followed by a random forest algorithm (SMO-RF), to categorized binary imbalanced data using the Technique of Synthetic Minority Oversampling. We have tested our model with four standard imbalanced datasets and obtained a higher F-measure, G-mean as well as ROC values for all data sets.

A priori prediction of response in multicentre locally advanced breast cancer (LABC) patients using quantitative ultrasound and derivative texture methods.

Purpose: We develop a multi-centric response predictive model using QUS spectral parametric imaging and novel texture-derivate methods for determining tumour responses to neoadjuvant chemotherapy (NAC) prior to therapy initiation.Materials and Methods: QUS Spectroscopy provided parametric images of mid-band-fit (MBF), spectral-slope (SS), spectral-intercept (SI), average-scatterer-diameter (ASD), and average-acoustic-concentration (AAC) in 78 patients with locally advanced breast cancer (LABC) undergoing NAC. Ultrasound radiofrequency data were collected from Sunnybrook Health Sciences Center (SHSC), University of Texas MD Anderson Cancer Center (MD-ACC), and St. Michaels Hospital (SMH) using two different systems. Texture analysis was used to quantify heterogeneities of QUS parametric images. Further, a second-pass texture analysis was applied to obtain texture-derivate features. QUS, texture- and texture-derivate parameters were determined from both tumour core and a 5-mm tumour margin and were used in comparison to histopathological analysis for developing a response predictive model to classify responders versus non-responders. Model performance was assessed using leave-one-out cross-validation. Three standard classification algorithms including a linear discriminant analysis (LDA), k-nearest-neighbors (KNN), and support vector machines-radial basis function (SVM-RBF) were evaluated.Results: A combination of tumour core and margin classification resulted in a peak response prediction performance of 88% sensitivity, 78% specificity, 84% accuracy, 0.86 AUC, 84% PPV, and 83% NPV, achieved using the SVM-RBF classification algorithm. Other parameters and classifiers performed less well running from 66% to 80% accuracy.Conclusions: A QUS-based framework and novel texture-derivative method enabled accurate prediction of responses to NAC. Multi-centric response predictive model provides indications of the robustness of the approach to variations due to different ultrasound systems and acquisition parameters.

Using U-Net-Like Deep Convolutional Neural Networks for Precise Tree Recognition in Very High Resolution RGB (Red, Green, Blue) Satellite Images

Very high resolution satellite imageries provide an excellent foundation for precise mapping of plant communities and even single plants. We aim to perform individual tree recognition on the basis of very high resolution RGB (red, green, blue) satellite images using deep learning approaches for northern temperate mixed forests in the Primorsky Region of the Russian Far East. We used a pansharpened satellite RGB image by GeoEye-1 with a spatial resolution of 0.46 m/pixel, obtained in late April 2019. We parametrized the standard U-Net convolutional neural network (CNN) and trained it in manually delineated satellite images to solve the satellite image segmentation problem. For comparison purposes, we also applied standard pixel-based classification algorithms, such as random forest, k-nearest neighbor classifier, naive Bayes classifier, and quadratic discrimination. Pattern-specific features based on grey level co-occurrence matrices (GLCM) were computed to improve the recognition ability of standard machine learning methods. The U-Net-like CNN allowed us to obtain precise recognition of Mongolian poplar (Populus suaveolens Fisch. ex Loudon s.l.) and evergreen coniferous trees (Abies holophylla Maxim., Pinus koraiensis Siebold & Zucc.). We were able to distinguish species belonging to either poplar or coniferous groups but were unable to separate species within the same group (i.e. A. holophylla and P. koraiensis were not distinguishable). The accuracy of recognition was estimated by several metrics and exceeded values obtained for standard machine learning approaches. In contrast to pixel-based recognition algorithms, the U-Net-like CNN does not lead to an increase in false-positive decisions when facing green-colored objects that are similar to trees. By means of U-Net-like CNN, we obtained a mean accuracy score of up to 0.96 in our computational experiments. The U-Net-like CNN recognizes tree crowns not as a set of pixels with known RGB intensities but as spatial objects with a specific geometry and pattern. This CNN’s specific feature excludes misclassifications related to objects of similar colors as objects of interest. We highlight that utilization of satellite images obtained within the suitable phenological season is of high importance for successful tree recognition. The suitability of the phenological season is conceptualized as a group of conditions providing highlighting objects of interest over other components of vegetation cover. In our case, the use of satellite images captured in mid-spring allowed us to recognize evergreen fir and pine trees as the first class of objects (“conifers”) and poplars as the second class, which were in a leafless state among other deciduous tree species.

New adaptive intelligent grey wolf optimizer based multi-objective quantitative classification rules mining approaches

The classification rule mining problem is one of the most important tasks of data mining. A constructed classification model is needed to have high accuracy, comprehensiveness, interestingness, etc. Furthermore, when the data composed of quantitative, numerical, or mixed data types, automatically discovering the appropriate intervals at the time of the mining process is a hard problem. The standard classification algorithms in the literature do not find the intervals of the quantitative attributes in the rules and this is performed a priori that causes the modification of datasets. Automatically constructing a successful classification model consisting of an explainable rule set without changing or modifying the data in artificial intelligence and machine learning is a very hot topic. Due to the philosophy of constantly researching to discover more accurate, surprising, and comprehensible rule sets and the absence of the most effective algorithm for all kinds of data sets, new methods or new versions of existing methods are proposed. In this study, automatic mining of high-quality rule set is handled as a multi-objective optimization problem due to the nature of the necessities and novel adaptive multi-objective intelligent search and optimization algorithms based on Grey Wolf Optimizer are proposed for this task. The datasets are considered as search spaces and the proposed adaptive Pareto based multi-objective Grey Wolf Optimizer algorithms are designed and applied as search methods for automatically discovering the high-quality rules. The proposed intelligent methods and successful classification algorithms such as naive Bayes, k-NN, support vector machines, decision trees, and RIPPER are tested in five real-world data sets with different characteristics. The obtained results show the efficiency of the proposed intelligent search and optimization methods.

SDSV: Angle Measurement for Supervised Classification

Data classification is one of the most fundamental tasks that can be accomplished by supervised machine learning. There exists a lot of algorithms, and they have the specific case of uses. Different classification methods follow different techniques to map the relationship between input and output. This article proposes an angle measurement-based classification technique called Support Directional Shifting Vectors (SDSV) to segment a spectral domain into regions with a very effective solution for classification problems. This method introduces two shifting vectors, named Support Direction Vector (SDV) and Support Origin Vector (SOV). These vectors are formed as a linear function to measure cosine-angle by the dot product of two separated data classes, named target data points and non-target data points. Considering the target class samples, the vectors get aligned in a way that the angle with the target class gets minimized, while the angle with the non-target class gets maximized. The error in the position of the linear function has been modeled as the loss function. Then, the vector position is updated iteratively by optimizing this loss function using a gradient descent algorithm. We have used this model to classify data from two different machine learning datasets to evaluate the performance of the proposed method. Finally, we have carefully examined the results and compared them with the other standard classification algorithms. In summary, the proposed SDSV algorithm has shown a comparable accuracy compared to different standard algorithms.

Graph Synthesis for Pig Breed Classification From Muzzle Images

Non-intrusive and automated detection of pig breeds, particularly from visual standpoint, is important from a food quality tracking perspective. In this work, colour as well as texture based visual descriptors from muzzle images have been identified, which, serve as breed-identifiers to separate four common pig-breeds: Duroc, Ghungroo, Hampshire and Yorkshire. While these handcrafted visual descriptors by themselves are fairly robust and discriminative, it is recognized that by controlling the decision space by choosing the feature-type based on colour or texture or both and the order in which particular breeds are siphoned, classification accuracy can be improved considerably. In that light, a stable, relatively data-independent, breed-specific, hierarchical tree synthesis and feature selection procedure is proposed based on a breed-pair cluster separation table. The proposed approach has been compared with the state of the art Phylogenetic distance based Hierarchical Agglomerative Clustering algorithm (AGNES) and also with the standard decision tree classification algorithm. On cross-validation, When completely different sets of pigs were used for training and testing (50-50 split), the proposed algorithm reported relatively high mean classification accuracies of 86.45% for Duroc, 93.02% for Ghungroo, 86.91% for Hampshire and 98.54% for Yorkshire, respectively.

Decomposition-Fusion for Label Distribution Learning

Label Distribution Learning (LDL) is a general learning framework that assigns an instance to a distribution over a set of labels rather than to a single label or multiple labels. Current LDL methods have proven their effectiveness in many real-life machine learning applications. However, LDL is a generalization of the classification task and as such it is exposed to the same problems as standard classification algorithms, including class-imbalanced, noise, overlapping or irregularities. The purpose of this paper is to mitigate these effects by using decomposition strategies. The technique devised, called Decomposition-Fusion for LDL (DF-LDL), is based on one of the most renowned strategy in decomposition: the One-vs-One scheme, which we adapt to be able to deal with LDL datasets. In addition, we propose a competent fusion method that allows us to discard non-competent classifiers when their output is probably not of interest. The effectiveness of the proposed DF-LDL method is verified on several real-world LDL datasets on which we have carried out two types of experiments. First, comparing our proposal with the base learners and, second, comparing our proposal with the state-of-the-art LDL algorithms. DF-LDL shows significant improvements in both experiments.

Validation of low-cost system for gait assessment in children with ataxia

BackgroundAtaxic syndromes include several rare, inherited and acquired conditions. One of the main issues is the absence of specific, and sensitive automatic evaluation tools and digital outcome measures to obtain a continuous monitoring of subjects' motor ability. ObjectivesThis study aims to test the usability of the Kinect system for assessing ataxia severity, exploring the potentiality of clustering algorithms and validating this system with a standard motion capture system. MethodsGait evaluation was performed by standardized gait analysis and by Kinect v2 during the same day in a cohort of young patient (mean age of 13.8±7.2). We analyzed the gait spatio-temporal parameters and we looked at the differences between the two systems through correlation and agreement tests. As well, we tested for possible correlations with the SARA scale as well. Finally, standard classification algorithm and principal components analysis were used to discern disease severity and groups. ResultsWe found biases and linear relationships between all the parameters. Significant correlations emerged between the SARA and the Speed, the Stride Length and the Step Length. PCA results, highlighting that a machine learning approach combined with Kinect-based evaluation shows great potential to automatically assess disease severity and diagnosis. ConclusionsThe spatio-temporal parameters measured by Kinect cannot be used interchangeably with those parameters acquired with standard motion capture system in clinical practice but can still provide fundamental information. Specifically, these results might bring to the development of a novel system to perform easy and quick evaluation of gait in young patients with ataxia, useful for patients stratification in terms of clinical severity and diagnosis.

Breast lesion characterization using Quantitative Ultrasound (QUS) and derivative texture methods.

PurposeAccurate and timely diagnosis of breast cancer is extremely important because of its high incidence and high morbidity. Early diagnosis of breast cancer through screening can improve overall prognosis. Currently, biopsy remains as the gold standard for tumor pathological confirmation. Development of diagnostic imaging techniques for rapid and accurate characterization of breast lesions is required. We aim to evaluate the usefulness of texture-derivate features of QUS spectral parametric images for non-invasive characterization of breast lesions.MethodsQUS Spectroscopy was used to determine parametric images of mid-band fit (MBF), spectral slope (SS), spectral intercept (SI), average scatterer diameter (ASD), and average acoustic concentration (AAC) in 204 patients with suspicious breast lesions. Subsequently, texture analysis techniques were used to generate texture maps from parametric images to quantify heterogeneities of QUS parametric images. Further, a second-pass texture analysis was applied to obtain texture-derivate features. QUS parameters, texture-parameters and texture-derivate parameters were determined from both tumor core and a 5-mm tumor margin and were used in comparison to histopathological analysis in order to develop a diagnostic model for classifying breast lesions as either benign or malignant. Both leave-one-out and hold-out cross-validations were used to evaluate the performance of the diagnostic model. Three standard classification algorithms including a linear discriminant analysis (LDA), k-nearest neighbors (KNN), and support vector machines-radial basis function (SVM-RBF) were evaluated.ResultsCore and margin information using the SVM-RBF attained the best classification performance of 90% sensitivity, 92% specificity, 91% accuracy, and 0.93 AUC utilizing QUS parameters and their texture derivatives, evaluated using leave-one-out cross-validation. Implementation of hold-out cross-validation using combination of both core and margin information and SVM-RBF achieved average accuracy and AUC of 88% and 0.92, respectively.ConclusionsQUS-based framework and derivative texture methods enable accurate classification of breast lesions. Evaluation of the proposed technique on a large cohort using hold-out cross-validation demonstrates its robustness and its generalization.

A Hybrid Evolutionary Computation Approach to Inducing Transfer Classifiers for Domain Adaptation.

Domain adaptation utilizes learned knowledge from an existing domain (source domain) to improve the classification performance of another related, but not identical, domain (target domain). Most existing domain adaptation methods first perform domain alignment, then apply standard classification algorithms. Transfer classifier induction is an emerging domain adaptation approach that incorporates the domain alignment into the process of building an adaptive classifier instead of using a standard classifier. Although transfer classifier induction approaches have achieved promising performance, they are mainly gradient-based approaches which can be trapped at local optima. In this article, we propose a transfer classifier induction algorithm based on evolutionary computation to address the above limitation. Specifically, a novel representation of the transfer classifier is proposed which has much lower dimensionality than the standard representation in existing transfer classifier induction approaches. We also propose a hybrid process to optimize two essential objectives in domain adaptation: 1) the manifold consistency and 2) the domain difference. Particularly, the manifold consistency is used in the main fitness function of the evolutionary search to preserve the intrinsic manifold structure of the data. The domain difference is reduced via a gradient-based local search applied to the top individuals generated by the evolutionary search. The experimental results show that the proposed algorithm can achieve better performance than seven state-of-the-art traditional domain adaptation algorithms and four state-of-the-art deep domain adaptation algorithms.

Light Gradient Boosting Machine for General Sentiment Classification on Short Texts: A Comparative Evaluation

Recently, the focus on sentiment analysis has been domain dependent even though the expressions used by the public are unsophisticatedly familiar regardless of the topics or domains. Online social media (OSNs) has been a daily venue for informal conversational contents from various domains ranging from sports and cooking to politics and human rights. Generating specific resources for every domain independently requires high cost and extensive efforts. In response, we propose to build a general multi-class sentiment classifier using our Domain-Free Sentiment Multimedia Dataset (DFSMD). Based on the proven capabilities of Light Gradient Boosting Machine (LGBM) in dealing with high dimensional and imbalance data, we have trained an LGBM model to recognize one of three sentiments of tweets: positive, negative, or neutral. We have conducted extensive comparisons and evaluations for six other standard sentiment classification algorithms and different sets of features including OSNs-specific ones. Our results have shown that LGBM model is the winner among the other six algorithms. It has been also shown that our dataset contains distinguishing characteristics in the three classes. Moreover, hashtag words are shown to be significantly important in capturing the sentiments of tweets. In addition, our findings have revealed the effectiveness of our approach in adapting general-domain sentiment to domain-specific sentiment analysis.