Search Results (4)

The generation and analysis of vast amounts of data have become increasingly prevalent in diverse applications. In this study, we propose a novel approach to address the challenge of rule explosion in association rule mining by utilizing the coverage-based representations of clusters determined by K-modes. We utilize the FP-Growth algorithm to generate class association rules (CARs). To further enhance the interpretability and compactness of the rule set, we employ the K-modes clustering algorithm with a distance metric that binarizes the rules. The optimal number of clusters is determined using the silhouette score. Representative rules are then selected based on their coverage within each cluster. To evaluate the effectiveness of our approach, we conducted experimental evaluations on both UCI and Kaggle datasets. The results demonstrate a significant reduction in the rule space (71 rules on average, which is the best result among all state-of-the-art rule-learning algorithms), aligning with our goal of producing compact classifiers. Our approach offers a promising solution for managing rule complexity in association rule mining, thereby facilitating improved rule interpretation and analysis, while maintaining a significantly similar classification accuracy (ACMKC: 80.0% on average) to other rule learners on most of the datasets. Full article

Classification is a very important task in data mining and pattern analysis, which have been widely used to solve various real-world problems. To obtain better classification performance, in this paper, we propose a novel classification framework based on multiple weighted class association rules (C-MWCAR), whose key idea is to transform the association among features into a set of class association rules (CARs), then classify unknown instances based on the CARs obtained. Concretely, C-MWCAR consists of a dictionary order-based CAR mining algorithm (DOCMA), a branch-based CAR selection algorithm (BCSA), and a multiple weighted CARs-based classifier (MWCC). Specifically, DOCMA mines the complete set of CARs, from which BCSA further selects a representative and concise set of CARs based on the distribution, coverage, and redundancy of the mined CARs. When classifying an unknown instance, MWCC picks out a set of CARs that are most similar to the given instance and computes the weighted importance of those CARs. Finally, the class label of the given instance will be determined by the similarities between the instance and the CARs and the weighted importance of the CARs. Furthermore, we apply the proposed C-MWCAR to a real-world classification task, i.e., hypertension diagnosis, based on a real dataset of 128 subjects. Experimental results indicate that C-MWCAR outperforms four baseline methods and achieves 93.3%, 93.8%, and 92.7% in terms of accuracy, sensitivity, and specificity, respectively. Full article

Machine learning techniques are ever prevalent as datasets continue to grow daily. Associative classification (AC), which combines classification and association rule mining algorithms, plays an important role in understanding big datasets that generate a large number of rules. Clustering, on the other hand, can contribute by reducing the rule space to produce compact models. The above-mentioned facts were the main motivation for this research work. We propose a new distance (similarity) metric based on “direct” and “indirect” measures and explain the overall importance of this method, which can produce compact and accurate models. Specifically, we aim to employ agglomerative hierarchical clustering to develop new associative classification models that contain a lower number of rules. Furthermore, a new strategy (based on the cluster center) is presented to extract the representative rule for each cluster. Twelve real-world datasets were evaluated experimentally for accuracy and compactness, and the results were compared to those of previously established associative classifiers. The results show that our method outperformed the other algorithms in terms of classifier size on most of the datasets, while still being as accurate in classification. Full article

Hypertension is one of the most common cardiovascular diseases, which will cause severe complications if not treated in a timely way. Early and accurate identification of hypertension is essential to prevent the condition from deteriorating further. As a kind of complex physiological state, hypertension is hard to characterize accurately. However, most existing hypertension identification methods usually extract features only from limited aspects such as the time-frequency domain or non-linear domain. It is difficult for them to characterize hypertension patterns comprehensively, which results in limited identification performance. Furthermore, existing methods can only determine whether the subjects suffer from hypertension, but they cannot give additional useful information about the patients’ condition. For example, their classification results cannot explain why the subjects are hypertensive, which is not conducive to further analyzing the patient’s condition. To this end, this paper proposes a novel hypertension identification method by integrating classification and association rule mining. Its core idea is to exploit the association relationship among multi-dimension features to distinguish hypertensive patients from normotensive subjects. In particular, the proposed method can not only identify hypertension accurately, but also generate a set of class association rules (CARs). The CARs are proved to be able to reflect the subject’s physiological status. Experimental results based on a real dataset indicate that the proposed method outperforms two state-of-the-art methods and three common classifiers, and achieves 84.4%, 82.5% and 85.3% in terms of accuracy, precision and recall, respectively. Full article