Personalized Activity Recognition Using Partially Available Target Data

Fallahzadeh, Ramin; Ashari, Zhila Esna; Alinia, Parastoo; Ghasemzadeh, Hassan

doi:10.1109/tmc.2021.3071434

Cited by 8 publications

(5 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Representative results include: Literature. [16] studied the evaluation method of microblog community influence, proposed an index system based on microblog information dissemination mechanism, combined quantitative indicators and qualitative indicators, and used principal component analysis to connect these the indicators combined into several comprehensive hands, which simplifies the indicator system; Literature [17] proposed a community impact evaluation model framework from the perspective of information dissemination and a set of related definitions of community impact evaluation forms, involving user impact and community impact; Literature [18] proposed a variable influence community detection method based on PageRank, which can adjust the community where a specific node is located and increase its influence. However, for the target community discovery task, in addition to accurately mining the community composed of high-quality nodes similar to the sample nodes given by the user, the ability of the community to spread the internal information of the community to external users, that is, the external influence of the community is also a significant factor.…”

Section: Quantification Methods Of Community External Influencementioning

confidence: 99%

“…In the target subspace set obtained by the remaining maximum k-cluster mining, according to the characteristics that there may be no connection relationship between different nodes, but the attributes may be similar, based on the initial point selection strategy in the improved k-medoid algorithm [17] select the one with the most significant difference d, k-groups are used as the core of each potential target community. Specifically, given the user-related maximal k-cluster set Ĉk…”

Section: Mining K-group Pruning Strategymentioning

confidence: 99%

See 1 more Smart Citation

Social Theory Aware Influence Maximization Approach for Target Community Detection

Scholar

Professor

2023

Preprint

View full text Add to dashboard Cite

Targeted community detection is a network analysis involving identifying specific subsets or clusters of nodes (communities) within a more extensive network that are particularly relevant to a topic or phenomenon. Unlike traditional community detection methods, which aim to identify all communities within a network, targeted community detection methods focus on finding a specific community or set of communities based on some pre-defined criteria. Nevertheless, all previous effort either mainly disregards the external effect of the community or is not ”goal-based,” i.e., inappropriate for the goal request. To address this issue, we present an attribute network-oriented target community discovery technique that blends user interest preferences and community influence to mine high-quality communities associated with user preferences and have the most influence. To capture the attribute subspace weight of the possible target community and mine user preferences, the greatest k-cluster, including sample nodes, is first mined as the core of the potential target community by synthesising the node structure and attribute information. Then the fusion of the maximum k-cluster with the sample nodes is mined as the core of the potential target community. Finally, the fusion of the sample nodes defines the community’s external influence score quantifica-tion technique and combines the community quality function value with the exterior influence score. Finally, all possible target communities are ranked by their influence score. As a result, the communities with the highest overall quality become the target communities. In addition, a 2-fold pruning procedure is intended to increase the method’s performance and efficiency while calculating the attribute subspace weights of the largest k-cluster. Experimental results on synthetic and actual network datasets validate the efficiency and efficacy of the suggested strategy.

show abstract

Section: Quantification Methods Of Community External Influencementioning

confidence: 99%

Section: Mining K-group Pruning Strategymentioning

confidence: 99%

Social Theory Aware Influence Maximization Approach for Target Community Detection

Scholar

Professor

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Problem 2 can be viewed as a combinatorial optimization problem that finds an optimal mapping in a two-tier fashion: (i) it initially performs component-level mappings where vertex-wise and edge-wise mappings are found between source and target dependency graphs; and (ii) it then uses the component-level mappings to reach a consensus about the optimal mapping for the problem as a whole. Such a two-level mapping problem can be represented using the objective in (15).…”

Section: Optimal Label Learningmentioning

confidence: 99%

“…Furthermore, M is a normalization factor that is equal to the total number of component-wise mappings. The objective in (15) attempts to minimize the amount of mapping costs at the graph-level and, therefore, can be viewed as the objective for Problem 2.…”

Section: Optimal Label Learningmentioning

confidence: 99%

“…Another study proposed a feature-level transfer learning approach for activity recognition where

% accuracy for cross-subject knowledge transfer was obtained [ 13 ]. Prior research also developed OptiMapper as a transfer learning framework for the case where the target domain provides data about only a subset of the classes [ 15 ]. However, as the degree of divergence between source and target domains grows, the transfer learning task becomes more challenging.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Model-Agnostic Structural Transfer Learning for Cross-Domain Autonomous Activity Recognition

Alinia

Arefeen

Ashari

et al. 2023

Sensors

Self Cite

View full text Add to dashboard Cite

Activity recognition using data collected with smart devices such as mobile and wearable sensors has become a critical component of many emerging applications ranging from behavioral medicine to gaming. However, an unprecedented increase in the diversity of smart devices in the internet-of-things era has limited the adoption of activity recognition models for use across different devices. This lack of cross-domain adaptation is particularly notable across sensors of different modalities where the mapping of the sensor data in the traditional feature level is highly challenging. To address this challenge, we propose ActiLabel, a combinatorial framework that learns structural similarities among the events that occur in a target domain and those of a source domain and identifies an optimal mapping between the two domains at their structural level. The structural similarities are captured through a graph model, referred to as the dependency graph, which abstracts details of activity patterns in low-level signal and feature space. The activity labels are then autonomously learned in the target domain by finding an optimal tiered mapping between the dependency graphs. We carry out an extensive set of experiments on three large datasets collected with wearable sensors involving human subjects. The results demonstrate the superiority of ActiLabel over state-of-the-art transfer learning and deep learning methods. In particular, ActiLabel outperforms such algorithms by average F1-scores of 36.3%, 32.7%, and 9.1% for cross-modality, cross-location, and cross-subject activity recognition, respectively.

show abstract