Multi-task learning for intelligent data processing in granular computing context

Algebraic Structure Based Clustering Method from Granular Computing Prospective

Clustering, as one of the main tasks of machine learning, is also the core work of granular computing, namely granulation. Most of the recent granular computing based clustering algorithms only utilize the plain granule features without taking the granule structure into account, especially in information area with widespread application of algebraic structure. This paper aims at proposing an algebraic structure based clustering method from granular computing prospective. Specifically, the algebraic structure based granularity is firstly formulated based on the granule structure of an algebraic binary operator. An algebraic structure based clustering method is then proposed by incorporating congruence partitioning granules and homomorphically projecting granule structure. Finally, proof of the lattice at multiple hierarchical levels and comparative analysis of experimental cases validate the effectiveness of the proposed clustering method. The algebraic structure based clustering method can provide a general framework to perform granularity clustering using the algebraic granule structure information. It meanwhile advances the granular computing methods by combing the granular computing theory and the clustering theory.

Human posture recognition based on multiple features and rule learning

The use of skeleton data for human posture recognition is a key research topic in the human-computer interaction field. To improve the accuracy of human posture recognition, a new algorithm based on multiple features and rule learning is proposed in this paper. Firstly, a 219-dimensional vector that includes angle features and distance features is defined. Specifically, the angle and distance features are defined in terms of the local relationship between joints and the global spatial location of joints. Then, during human posture classification, the rule learning method is used together with the Bagging and random subspace methods to create different samples and features for improved classification performance of sub-classifiers for different samples. Finally, the performance of our proposed algorithm is evaluated on four human posture datasets. The experimental results show that our algorithm can recognize many kinds of human postures effectively, and the results obtained by the rule-based learning method are of higher interpretability than those by traditional machine learning methods and CNNs.

Event abstraction in process mining: literature review and taxonomy

The execution of processes in companies generates traces of event data, stored in the underlying information system(s), capturing the actual execution of the process. Analyzing event data, i.e., the focus of process mining, yields a detailed understanding of the process, e.g., we are able to discover the control flow of the process and detect compliance and performance issues. Most process mining techniques assume that the event data are of the same and/or appropriate level of granularity. However, in practice, the data are extracted from different systems, e.g., systems for customer relationship management, Enterprise Resource Planning, etc., record the events at different granularity levels. Hence, pre-processing techniques that allow us to abstract event data into the right level of granularity are vital for the successful application of process mining. In this paper, we present a literature study, in which we assess the state-of-the-art in the application of such event abstraction techniques in the field of process mining. The survey is accompanied by a taxonomy of the existing approaches, which we exploit to highlight interesting novel directions.