International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems

The uncertain vector autoregressive model is able to model the interrelationships between different variables, which is more advantageous compared to the traditional autoregressive model, when modeling real-life objects and where the observed values are imprecise. In this paper, the parameters of the uncertain vector autoregressive model are estimated by using least absolute deviation estimation (LAD) to obtain a fitted uncertain vector autoregressive model, and residual analysis is performed to obtain estimates of expected values and variances of the residuals. In addition, future values are modeled by using forecasting methods, i.e., point estimation and interval estimation. The order of the uncertain vector autoregressive model is also determined by the indicator summation of test errors (STE) in the cross-validation, and we also analyze that the least absolute deviation estimation outperforms the least squares estimation method in the presence of outliers.

The idea of nondeterministic typical hesitant fuzzy automata is a generalization of the fuzzy automata presented by Costa and Bedregal. This paper, presents the sufficient and necessary conditions for a typical hesitant fuzzy language to be computed by nondeterministic typical hesitant fuzzy automata. Besides, the paper introduces a new class of typical hesitant fuzzy automata with crisp transitions, and we will show that this new class is equivalent to the original class introduced by Costa and Bedregal.

The breast cancer diagnosis is one of the challenging tasks of medical field. Especially, the breast cancer diagnosis among younger women (under 40 years old) is more complicated, because their breast tissue is generally denser than the older women. The Breast Cancer Wisconsin image dataset contains two classes: (i) Benign (Minority class), (ii) Malignant (Majority class). The imbalanced class distribution leads to a deterioration in the classifier model performance owing to the biased classification towards the majority class. Therefore, in this article, an improved generative adversarial network (I-GAN) is proposed to overcome the class imbalance problem. Here, the proposed method is the consolidation of deep convolutional generative adversarial network (DCIGAN) and modified convolutional neural network, (MCNN), therefore it is known as DCIGAN-MCNN method. First, the DCIGAN is utilized for balancing the dataset by generating more samples in the training dataset. Then, this training dataset based the classification of Breast cancer is developed using the modified convolutional neural network. The proposed method is executed in MATLAB. The performance analysis are carried out in Breast Cancer Wisconsin (Prognostic) Data Set provides Maximum Geometry Mean (MGM) as 24.058%, 9.582%, Matthews’s correlation coefficient (MCC) as 78.623%, 30.357% higher than the existing methods, like CI-BC-RK-SVM, CI-BC-GA, CI-BC-DC-CNN, CI-BC-RF and CI-BC-BMIC-Net respectively. Finally, the simulation results prove that the proposed method can be able to find the optimal solutions efficiently and accurately.

Web page recommendation system has attracted more attention in recent decades. The web page recommendation has various characteristics than the classical recommenders. It is the process of predicting the request of the next web page that users are significantly interested while searching the web. It helps the users to find relevant pages in the field of web mining. In particular, web user may spend more time to identify expected information. To understand behavior of users and to visit the page based on their interest at a specific time, an effective web page recommendation method is developed by developed Multi-Verse Sailfish Optimization (MVSFO)-based Deep Residual network. Accordingly, proposed MVSFO is derived by the integration of Multi-Verse Optimizer (MVO) and Sailfish Optimizer (SFO), respectively. Here, the process of recommendation is carried out using weblog data and the web page image. The sequential patterns are acquired from weblog data, and the patterns are grouped with Deep fuzzy clustering based on cosine similarity. The matching process among test pattern and sequential patterns are made using Canberra distance. Here, the recommended web pages obtained from the weblog data and pages obtained from web pages image using the Deep Residual network are enable to generate the output using fractional order-based ranking. The developed scheme attained more effectiveness by the measures, such as F-measure, precision, and recall as 85.30%, 86.59%, and 86.04%, respectively for MSNBC dataset.

The imbalanced data classification is a challenging issue in many domains including medical intelligent diagnosis and fraudulent transaction analysis. The performance of the conventional classifier degrades due to the imbalanced class distribution of the training data set. Recently, machine learning and deep learning techniques are used for imbalanced data classification. Data preprocessing approaches are also suitable for handling class imbalance problem. Data augmentation is one of the preprocessing techniques used to handle skewed class distribution. Synthetic Minority Oversampling Technique (SMOTE) is a promising class balancing approach and it generates noise during the process of creation of synthetic samples. In this paper, AutoEncoder is used as a noise reduction technique and it reduces the noise generated by SMOTE. Further, Deep one-dimensional Convolutional Neural Network is used for classification. The performance of the proposed method is evaluated and compared with existing approaches using different metrics such as Precision, Recall, Accuracy, Area Under the Curve and Geometric Mean. Ten data sets with imbalance ratio ranging from 1.17 to 577.87 and data set size ranging from 303 to 284807 instances are used in the experiments. The different imbalanced data sets used are Heart-Disease, Mammography, Pima Indian diabetes, Adult, Oil-Spill, Phoneme, Creditcard, BankNoteAuthentication, Balance scale weight & distance database and Yeast data sets. The proposed method shows an accuracy of 96.1%, 96.5%, 87.7%, 87.3%, 95%, 92.4%, 98.4%, 86.1%, 94% and 95.9% respectively. The results suggest that this method outperforms other deep learning methods and machine learning methods with respect to G-mean and other performance metrics.

Uncertainty estimation methods using deep learning approaches strive against separating how uncertain the state of the world manifests to us via measurement (objective end) from the way this gets scrambled with the model specification and training procedure used to predict such state (subjective means) — e.g., number of neurons, depth, connections, priors (if the model is bayesian), weight initialization, etc. This poses the question of the extent to which one can eliminate the degrees of freedom associated with these specifications and still being able to capture the objective end. Here, a novel non-parametric quantile estimation method for continuous random variables is introduced, based on the simplest neural network architecture with one degree of freedom: a single neuron. Its advantage is first shown in synthetic experiments comparing with the quantile estimation achieved from ranking the order statistics (specifically for small sample size) and with quantile regression. In real-world applications, the method can be used to quantify predictive uncertainty under the split conformal prediction setting, whereby prediction intervals are estimated from the residuals of a pre-trained model on a held-out validation set and then used to quantify the uncertainty in future predictions — the single neuron used here as a structureless “thermometer” that measures how uncertain the pre-trained model is. Benchmarking regression and classification experiments demonstrate that the method is competitive in quality and coverage with state-of-the-art solutions, with the added benefit of being more computationally efficient.

This research proposes a new wrapper model based on chaos theory and nature-inspired pelican optimization algorithm (POA) for feature selection. The base algorithm is converted into a binary one and a chaotic search to augment POA’s exploration and exploitation process, denoted as chaotic binary pelican optimization algorithm (CBPOA). The main focus of chaos theory is to resolve the slow convergence rate as well as entrapment in local optimal issues of classical POA. Therefore, ten dissimilar chaotic maps are entrenched in POA to tackle these issues and attain a more robust and effective search mechanism. CBPOA executes on continuous search; thus, the continuous search is reformed to a discrete one by adapting transfer functions. In CBPOA, eight transfer functions are used to find the best one and inspect CBPOA. Consequently, the performance of the CBPOA has been investigated by targeting several metrics under 18 UCI datasets. The best variant is nominated and explored the performance with classical wrapper-based and filter-based schemes. Furthermore, the proposed CBPOA is evaluated using 23 functions from CEC-2017, 2018 and 2020 benchmarks. As an outcome, CBPOA has accomplished better outcomes than existing schemes and is superior in handling feature selection problems.