Semi-Supervised Ensemble Classification Method Based on Near Neighbor and Its Application

Abstract

Semi-supervised learning can be used to solve the problem of insufficient labeled samples in the process industry. However, in an actual scenario, traditional semi-supervised learning methods usually do not achieve satisfactory performance when the small number of labeled samples is subjective and inaccurate and some do not consider how to develop a strategy to expand the training set. In this paper, a new algorithm is proposed to alleviate the above two problems, and consequently, the information contained in unlabeled samples can be fully mined. First, the multivariate adaptive regression splines (MARS) and adaptive boosting (Adaboost) algorithms are adopted for co-training to make the most of the deep connection between samples and features. In addition, the strategies, pseudo-labeled dataset selection algorithm based on near neighbor degree (DSSA) and pseudo-labeled sample detection algorithm based on near neighbor degree selection (SPDA) are adopted to enlarge the dataset of labeled samples. When we select the samples from the pseudo-labeled data to join the training set, the confidence degree and the spatial relationship with labeled samples are considered, which are able to improve classifier accuracy. The results of tests on multiple University of California Irvine (UCI) datasets and an actual dataset in the aluminum electrolysis industry demonstrate the effectiveness of the proposed algorithm.

Missing label imputation through inception-based semi-supervised ensemble learning

“…Although in different scenarios semi-supervised learning methods have been successfully applied during the training process, the randomness in the selection of unlabeled samples through some methods has led to unstable and less robust models (Zhu and Goldberg 2009). The above issue can be solved using ensemble learning approaches (Li et al 2020). In Zhou (2009) a theoretical analysis shows the benefits of combining these two types of learning approaches (semisupervised learning and ensemble learning) in the context of disagreement-based learning.…”

“…Since the development of an ensemble of classifiers has become a necessary direction for improving the classification accuracy and also it shows a potential for reducing the cost of data labeling, ensemble methods have been used in different directions (e.g., medical, industrial, etc. ) to get more accurate outputs (Livieris et al 2018;Li et al 2020).…”

“…In addition, a semi-supervised ensemble classification method based on nearest neighbors is proposed in Li et al (2020), which is used to solve the problem on the presence of missing labels. This method involves using the multivariate adaptive regression splines (MARS) and adaptive boosting (Adaboost) algorithms as sub-models for co-training, and combining pseudo-labeled data set selection algorithm based on nearest neighbor degree (DSSA) and pseudo-labeled sample detection algorithm based on nearest neighbor degree selection (SPDA) to enlarge the number of labeled samples in the field of process industry.…”

“…However, at the same time, when semi-supervised ensemble learning is applied in industrial applications, there are still some limitations, e.g., in industry, a set of collected data is mostly labeled by people through their subjective and uncertain empirical knowledge (Yue et al 2019). In this case, when the total number of training instances is limited, then the impact of wrongly labeled samples on the learning performance becomes much larger (Li et al 2020). To overcome the limitations of traditional semi-supervised ensemble learning approaches, a novel Semi-Supervised Inception Neural Network Ensemble (SS-INNE)-based architecture is proposed.…”

An Improved Semi-Supervised Gaussian Mixture Model (I-SGMM)

“…They have been shown to outperform traditional supervised and unsupervised learning methods in various domains by utilizing the information from both labeled and unlabeled data [6]. In semi-supervised learning, the incorporation of unlabeled data samples into the training process has been found to enhance the predictive performance of the models and improve the generalization ability [7]. The research aims to advance the field of semi-supervised learning by enhancing the efficacy and applicability of Gaussian Mixture Models (GMMs) in the context of Improved Semi-Supervised Gaussian Mixture Models.…”