Random Subspace Ensembles of Fully Convolutional Network for Time Series Classification

Time series classification (TSC) task is one of the most significant topics in data mining. Among all methods for this issue, the deep-learning-based shows superior performance for its good adaption to raw series data and automatic extraction of features. However, rare eyes are kept on composing ensembles of these superior individual classifiers to achieve further breakthroughs. The existing deep learning ensembles NNE did a heavy work of combining 60 individuals but did not maximize the deserving improvement, since it merely pays attention to the diversity of individuals but ignores their accuracy. In this paper, we propose to construct an ensemble of Full Convolutional Neural Networks (FCN) by Random Subspace Method (RSM), named RSM-FCN. FCN is a simple but outstanding individual classifier and RSM is suitable for high dimensional data such as time series, but there are few instances. Thus, the combination of these strengths, RSM-FCN provides a highly cost-effective approach to yield promising results. Experiments on the UCR dataset demonstrate the effectiveness and reasonability of the proposed method.

Toward Robust Non-Intrusive Load Monitoring via Probability Model Framed Ensemble Method

As a pivotal technological foundation for smart home implementation, non-intrusive load monitoring is emerging as a widely recognized and popular technology to replace the sensors or sockets networks for the detailed household appliance monitoring. In this paper, a probability model framed ensemble method is proposed for the target of robust appliance monitoring. Firstly, the non-intrusive load disaggregation-oriented ensemble architecture is presented. Then, dictionary learning model is utilized to formulate the individual classifier, while the sparse coding-based approach is capable of providing multiple solutions under greedy mechanism. Furthermore, a fully probabilistic model is established for combined classifier, where the candidate solutions are all labelled with probability scores and evaluated via two-stage decision-making. The proposed method is tested on both low-voltage network simulator platform and field measurement datasets, and the results show that the proposed ensemble method always guarantees an enhancement on the performance of non-intrusive load disaggregation. Besides, the proposed approach shows high flexibility and scalability in classification model selection. Therefore, by initializing the architecture and approach of ensemble method-based NILM, this work plays a pioneer role in using ensemble method to improve the robustness and reliability of non-intrusive appliance monitoring.

Leveraging of Weighted Ensemble Technique for Identifying Medical Concepts from Clinical Texts at Word and Phrase Level

Concept identification from medical texts becomes important due to digitization. However, it is not always feasible to identify all such medical concepts manually. Thus, in the present attempt, we have applied five machine learning classifiers (Support Vector Machine, K-Nearest Neighbours, Logistic Regression, Random Forest and Naïve Bayes) and one deep learning classifier (Long Short Term Memory) to identify medical concepts by training a total of 27.383K sentences. In addition, we have also developed a rule based phrase identification module to help the existing classifiers for identifying multi- word medical concepts. We have employed word2vec technique for feature extraction and PCA and T- SNE for conducting ablation study over various features to select important ones. Finally, we have adopted two different ensemble approaches, stacking and weighted sum to improve the performance of the individual classifier and significant improvements were observed with respect to each of the classifiers. It has been observed that phrase identification module plays an important role when dealing with individual classifier in identifying higher order ngram medical concepts. Finally, the ensemble approach enhances the results over SVM that was showing initial improvement even after the application of phrase based module.

Exploring influential factors of shape classifier comprehension and production in Mandarin-speaking children

Mandarin classifiers are a complex system, but little is known about how Mandarin-speaking children manage to learn the system. Based on the extant literature, we explored potential factors influencing the comprehension and production of Mandarin shape classifiers, including classifier-based semantic categorization and errors pertaining to the semantic strategies, input frequency of classifier-noun combinations, and vocabulary knowledge. In total, 138 typically developing monolingual Mandarin-speaking children between ages 4;1 and 6;5 completed an object categorization task, shape classifier comprehension and production tasks, and a vocabulary test. The results showed that classifier-based categorization did not significantly relate to classifier knowledge, but children’s comprehension errors were mostly selecting an object that is perceptually similar to the target object. Estimated input frequency of classifier-noun combinations was significantly related to classifier comprehension, and there was differential accuracy for different classifier-noun combinations, which may indicate item-by-item learning of individual classifier-noun pairings. Mandarin-speaking children may take a combined approach by sorting semantic features for different classifiers and learning individual classifier-noun combinations. The interplay of the two approaches can be very complex and should be further investigated in future studies. Vocabulary knowledge was significantly related to classifier comprehension and production, indicating common traits between classifier learning and noun learning.

Portioning-Out and Individuation in Mandarin Non-interrogative wh-Pronominal Phrases: Experimental Evidence From Child Mandarin

Portioning-out and individuation are two important semantic properties for the characterization of countability. In Mandarin, nouns are not marked with count-mass syntax, and it is controversial whether individuation is encoded in classifiers or in nouns. In the present study, we investigates the interpretation of a minimal pair of non-interrogative wh-pronominal phrases, including duo-shao-N and duo-shao-ge-N. Due to the presence/absence of the individual classifier ge, these two wh-pronominal phrases differ in how they encode portioning-out and individuation. In two experiments, we used a Truth Value Judgment Task to examine the interpretation of these two wh-pronominal phrases by Mandarin-speaking adults and 4-to-6-year-old children. We found that both adults and children are sensitive to their interpretative differences with respect to the portioning-out and individuation properties. They assign either count or mass readings to the bare wh-pronominal phrase duo-shao-N depending on specific contexts, but only count readings to the classifier-bearing wh-pronominal phrase duo-shao-ge-N. Moreover, the portioning-out and individuation properties associated with the individual classifier ge emerge independently in the course of language development, with the portioning-out property taking precedence over the individuation property. Taken together, the present study provides new evidence for the view that the portioning-out and individuation properties in Mandarin are encoded in classifiers rather than in nouns, and these two semantic properties are two distinct components in our grammar.

Segmentation of Medical Image Based on Superpixel Boundary Perceptual Convolutional Network in Cancer Diagnosis and Treatment

In order to reduce the computational complexity of breast tumor segmentation algorithms and improve the accuracy of breast segmentation, this paper proposes a breast tumor segmentation method based on super pixel boundary perceptual convolutional network. This method first uses super pixel segmentation convolutional network algorithm to segment breast medical images, and then uses region growth algorithm to achieve breast tumor segmentation at super pixel level. The research results show that in the classification of breast tumors, the fusion efficiency based on the classifier level is better than the fusion based on the feature set; the index R proposed and adopted in this paper can effectively select the appropriate individual classifier and generate a better performing integration 06%. Classifier, the accuracy of this classifier is 88.73%, the sensitivity is 97.06%. The method can be used to assist doctors in breast cancer diagnosis, improve the efficiency and accuracy of doctors' work diagnosis, and has certain significance for clinical research and large-scale screening of breast cancer.

Image-Level Structure Recognition Using Image Features, Templates, and Ensemble of Classifiers

Image-level structural recognition is an important problem for many applications of computer vision such as autonomous vehicle control, scene understanding, and 3D TV. A novel method, using image features extracted by exploiting predefined templates, each associated with individual classifier, is proposed. The template that reflects the symmetric structure consisting of a number of components represents a stage—a rough structure of an image geometry. The following image features are used: a histogram of oriented gradient (HOG) features showing the overall object shape, colors representing scene information, the parameters of the Weibull distribution features, reflecting relations between image statistics and scene structure, and local binary pattern (LBP) and entropy (E) values representing texture and scene depth information. Each of the individual classifiers learns a discriminative model and their outcomes are fused together using sum rule for recognizing the global structure of an image. The proposed method achieves an 86.25% recognition accuracy on the stage dataset and a 92.58% recognition rate on the 15-scene dataset, both of which are significantly higher than the other state-of-the-art methods.

Does Ensemble Learning Always Lead to Better Forecasts?

Ensemble learning is a common machine learning technique applied to business and economic analysis in which several classifiers are combined using majority voting for better forecasts as compared to those of individual classifier. This study presents a counterexample, which demonstrates that ensemble learning leads to worse classifications than those from individual classifiers, using two events and three classifiers. If there is an outstanding classifier, we should follow its forecast instead of using ensemble learning.

Twitter Sentiment Analysis Using an Ensemble Majority Vote Classifier

Twitter social media data generally uses ambiguous text that can cause difficulty in identifying positive or negative sentiments. There are more than one billion social media messages that need to be stored in a proper database and processed correctly to analyze them. In this paper, an ensemble majority vote classifier to enhance sentiment classification performance and accuracy is proposed. The proposed classification model is combined with four classifiers, using varying techniques—naive Bayes, decision trees, multilayer perceptron and logistic regression—to form a single ensemble classifier. In addition to these, a comparison is drawn among the four classifiers to evaluate the performance of the individual classifiers. The result shows that in terms of an individual classifier, the naive Bayes classifier is optimal as compared to the others. However, for comparing the proposed ensemble majority vote classifier with the four individual classifiers, the result illustrates that the performance of the proposed classifier is better than the independent one.

Multi-Font Arabic Isolated Character Recognition Using Combining Machine Learning Classifiers

Nowadays, optical character recognition is one of the most successful automatic pattern recognition applications. Many works have been done regarding the identification of Latin and Chinese characters. However, the reason for having few investigations for the recognition of Arabic characters is the complexity and difficulty of Arabic characters identification compared to the others. In the current work, we investigate combining multiple machine learning algorithms for multi-font Arabic isolated characters recognition, where imperfect and dimensionally variable input charactersare faced. To the best of our knowledge, there is no such work yet available in this regard. Experimental results show that combined multiple classifiers can outperform each individual classifier produces by itself. The current findings are encouraging and opens the door for further research tasks in this direction.