Xin Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Xin Xu has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Minghao Yin
School of Information Science and Technology, Northeast Normal University, Changchun, China
4 shared publications - Jianan Wang
School of Physics, Northeast Normal University, Changchun, China
2 shared publications - Yupeng Zhou
School of Information Science and Technology, Northeast Normal University, Changchun, China
2 shared publications - Dunbo Cai
Center for Technology Research and Innovation, China Mobile (Suzhou) Software Technology Company Ltd., Suzhou, China
1 shared publication Fuyu Yuan
Northeast Normal University, Changchun, China
1 shared publication- Hui Lu
School of Electronic and Information Engineering, Beihang University, Beijing, China
1 shared publication
Latest Publications
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Learning Technologies
- Vol. 19, 433-446
ConfuciusLM: Exploring Confucian Personalized Instruction for Primary Learners With Large Language Models
Large Language Models (LLMs) have emerged as a transformative force in educational informatization, offering unprecedented capabilities to simulate teacher–student interactions and generate adaptive instructional resources. However, general-purpose LLMs often provide direct, outcome-oriented responses that lack pedagogical sensitivity, while existing educational models frequently prioritize rigid logical inquiry (e.g., Socratic questioning) over the affective and cognitive needs of novice learners. Such limitations can induce learning anxiety and impede the development of student confidence, particularly at the primary education level. To bridge this gap, we propose ConfuciusLM, a specialized educational LLM fine tuned on a novel instructional dialogue dataset grounded in Confucian ped agogical philosophy. ConfuciusLM embodies a learner-centered paradigm by implementing a heuristic reasoning framework that integrates cognitive guidance with affective support and motivational scaffolding. Through both automated metrics and human-centered pedagogical evaluations, we demonstrate that ConfuciusLM exhibits superior performance compared to open source LLMs of similar parameter scales in fostering positive learning dispositions and maintaining instructional coherence. This study advances the field of LLM-empowered education by proposing a robust framework that integrates Eastern pedagogical principles with AI-driven instructional design, thereby informing the development of more inclusive, empathetic, and human-centered intelligent tutoring systems. Importantly, the proposed approach offers concrete pedagogical and design implications for primary education, supporting the creation of learning environments that are psychologically safe, conducive to the development of self-efficacy, and capable of sustaining learner engagement under appropriate teacher guidance.
- 29 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Learning Technologies
- Vol. 19, 433-446
ConfuciusLM: Exploring Confucian Personalized Instruction for Primary Learners With Large Language Models
Large Language Models (LLMs) have emerged as a transformative force in educational informatization, offering unprecedented capabilities to simulate teacher–student interactions and generate adaptive instructional resources. However, general-purpose LLMs often provide direct, outcome-oriented responses that lack pedagogical sensitivity, while existing educational models frequently prioritize rigid logical inquiry (e.g., Socratic questioning) over the affective and cognitive needs of novice learners. Such limitations can induce learning anxiety and impede the development of student confidence, particularly at the primary education level. To bridge this gap, we propose ConfuciusLM, a specialized educational LLM fine tuned on a novel instructional dialogue dataset grounded in Confucian ped agogical philosophy. ConfuciusLM embodies a learner-centered paradigm by implementing a heuristic reasoning framework that integrates cognitive guidance with affective support and motivational scaffolding. Through both automated metrics and human-centered pedagogical evaluations, we demonstrate that ConfuciusLM exhibits superior performance compared to open source LLMs of similar parameter scales in fostering positive learning dispositions and maintaining instructional coherence. This study advances the field of LLM-empowered education by proposing a robust framework that integrates Eastern pedagogical principles with AI-driven instructional design, thereby informing the development of more inclusive, empathetic, and human-centered intelligent tutoring systems. Importantly, the proposed approach offers concrete pedagogical and design implications for primary education, supporting the creation of learning environments that are psychologically safe, conducive to the development of self-efficacy, and capable of sustaining learner engagement under appropriate teacher guidance.
- Open Access
DeeWaNA: An Unsupervised Network Representation Learning Framework Integrating Deepwalk and Neighborhood Aggregation for Node Classification
This paper introduces DeeWaNA, an unsupervised network representation learning framework that unifies random walk strategies and neighborhood aggregation mechanisms to improve node classification performance. Unlike existing methods that treat these two paradigms separately, our approach integrates them into a cohesive model, addressing limitations in structural feature extraction and neighborhood relationship modeling. DeeWaNA first leverages DeepWalk to capture global structural information and then employs an attention-based weighting mechanism to refine neighborhood relationships through a novel distance metric. Finally, a weighted aggregation operator fuses these representations into a unified low-dimensional space. By bridging the gap between random-walk-based and neural-network-based techniques, our framework enhances representation quality and improves classification accuracy. Extensive evaluations on real-world networks demonstrate that DeeWaNA outperforms four widely used unsupervised network representation learning methods, underscoring its effectiveness and broader applicability.
Open AccessDeeWaNA: An Unsupervised Network Representation Learning Framework Integrating Deepwalk and Neighborhood Aggregation for Node Classification
This paper introduces DeeWaNA, an unsupervised network representation learning framework that unifies random walk strategies and neighborhood aggregation mechanisms to improve node classification performance. Unlike existing methods that treat these two paradigms separately, our approach integrates them into a cohesive model, addressing limitations in structural feature extraction and neighborhood relationship modeling. DeeWaNA first leverages DeepWalk to capture global structural information and then employs an attention-based weighting mechanism to refine neighborhood relationships through a novel distance metric. Finally, a weighted aggregation operator fuses these representations into a unified low-dimensional space. By bridging the gap between random-walk-based and neural-network-based techniques, our framework enhances representation quality and improves classification accuracy. Extensive evaluations on real-world networks demonstrate that DeeWaNA outperforms four widely used unsupervised network representation learning methods, underscoring its effectiveness and broader applicability.
- Top Cited Papers
- 16 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Learning Technologies
- Vol. 17, 629-641
Unleashing ChatGPT's Power: A Case Study on Optimizing Information Retrieval in Flipped Classrooms via Prompt Engineering
This research project investigates the impact of Prompt Engineering, a key aspect of ChatGPT, on college students' information retrieval in flipped classrooms. In recent years, an increasing number of students have been using AI-based tools such as ChatGPT rather than traditional research engines to learn and to complete course assignments. Despite this growing trend, previous research has largely overlooked the influence of prompt engineering on students' use of ChatGPT and effective strategies for improving the quality of information retrieval in learning settings. To address this research gap, this study examines the information quality obtained from ChatGPT in a flipped classroom by evaluating its effectiveness in task completion among 26 novice undergraduates from the same major and cohort. The experimental results provide evidence that proficient mastery of prompt engineering improves the quality of information obtained by students using ChatGPT. Consequently, by acquiring proficiency in prompt engineering, students can maximize the positive impact of ChatGPT, obtain high-quality information, and enhance their learning efficiency in flipped classrooms.
- 16 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Learning Technologies
- Vol. 17, 629-641
Top Cited PapersUnleashing ChatGPT's Power: A Case Study on Optimizing Information Retrieval in Flipped Classrooms via Prompt Engineering
This research project investigates the impact of Prompt Engineering, a key aspect of ChatGPT, on college students' information retrieval in flipped classrooms. In recent years, an increasing number of students have been using AI-based tools such as ChatGPT rather than traditional research engines to learn and to complete course assignments. Despite this growing trend, previous research has largely overlooked the influence of prompt engineering on students' use of ChatGPT and effective strategies for improving the quality of information retrieval in learning settings. To address this research gap, this study examines the information quality obtained from ChatGPT in a flipped classroom by evaluating its effectiveness in task completion among 26 novice undergraduates from the same major and cohort. The experimental results provide evidence that proficient mastery of prompt engineering improves the quality of information obtained by students using ChatGPT. Consequently, by acquiring proficiency in prompt engineering, students can maximize the positive impact of ChatGPT, obtain high-quality information, and enhance their learning efficiency in flipped classrooms.
- 4 May 2022
- Published by Springer Nature in Applied Intelligence
- Vol. 53 (2) , 1912-1944
A master-apprentice evolutionary algorithm for maximum weighted set K-covering problem
- 4 May 2022
- Published by Springer Nature in Applied Intelligence
- Vol. 53 (2) , 1912-1944
A master-apprentice evolutionary algorithm for maximum weighted set K-covering problem
- Open Access
- 27 July 2021
- Published by MDPI AG in Mathematics
- Vol. 9 (15) , 1767
An Information-Explainable Random Walk Based Unsupervised Network Representation Learning Framework on Node Classification Tasks
Network representation learning aims to learn low-dimensional, compressible, and distributed representational vectors of nodes in networks. Due to the expensive costs of obtaining label information of nodes in networks, many unsupervised network representation learning methods have been proposed, where random walk strategy is one of the wildly utilized approaches. However, the existing random walk based methods have some challenges, including: 1. The insufficiency of explaining what network knowledge in the walking path-samplings; 2. The adverse effects caused by the mixture of different information in networks; 3. The poor generality of the methods with hyper-parameters on different networks. This paper proposes an information-explainable random walk based unsupervised network representation learning framework named Probabilistic Accepted Walk (PAW) to obtain network representation from the perspective of the stationary distribution of networks. In the framework, we design two stationary distributions based on nodes’ self-information and local-information of networks to guide our proposed random walk strategy to learn representational vectors of networks through sampling paths of nodes. Numerous experimental results demonstrated that the PAW could obtain more expressive representation than the other six widely used unsupervised network representation learning baselines on four real-world networks in single-label and multi-label node classification tasks.
- 27 July 2021
- Published by MDPI AG in Mathematics
- Vol. 9 (15) , 1767
Open AccessAn Information-Explainable Random Walk Based Unsupervised Network Representation Learning Framework on Node Classification Tasks
Network representation learning aims to learn low-dimensional, compressible, and distributed representational vectors of nodes in networks. Due to the expensive costs of obtaining label information of nodes in networks, many unsupervised network representation learning methods have been proposed, where random walk strategy is one of the wildly utilized approaches. However, the existing random walk based methods have some challenges, including: 1. The insufficiency of explaining what network knowledge in the walking path-samplings; 2. The adverse effects caused by the mixture of different information in networks; 3. The poor generality of the methods with hyper-parameters on different networks. This paper proposes an information-explainable random walk based unsupervised network representation learning framework named Probabilistic Accepted Walk (PAW) to obtain network representation from the perspective of the stationary distribution of networks. In the framework, we design two stationary distributions based on nodes’ self-information and local-information of networks to guide our proposed random walk strategy to learn representational vectors of networks through sampling paths of nodes. Numerous experimental results demonstrated that the PAW could obtain more expressive representation than the other six widely used unsupervised network representation learning baselines on four real-world networks in single-label and multi-label node classification tasks.
- Open Access
- 31 August 2020
- Published by Association for Computing Machinery (ACM) in ACM/IMS Transactions on Data Science
- Vol. 1 (3) , 1-16
Adaptive Weighted Finite Mixture Model
It is significant for real estate investors to understand how the construction environments and building characteristics impact the housing unit price. However, it is challenging for identifying the complex feature-influence from construction environments and building characteristics. It is also hard to alleviate the heterogeneity of real estate. In this article, we propose a framework named Adaptive Weighted Finite Mixture Model to identify the feature-influence and simultaneously alleviate the ill effect of heterogeneity. Applying this framework, we can predict the housing unit price based on the corresponding features. Besides, we discover that the feature-influence exists in the dissimilarity among similar cities. Specifically, we adaptively learn the weights of features to identify the feature-influence, and we model the estimation of the housing unit price with the feature-influence into a finite mixture model. We utilize the Principle Component Analysis algorithm to obtain a low-dimensional representation of housing features. The low-dimensional representation reduces the computational cost of model learning, and it avoids a potential catastrophe of the singular matrix inversion during the process of learning model parameters, which are estimated by the Expectation Maximization algorithm. To avoid the blind search for the latent group number used in the proposed framework, we employ the pre-clustering result as a guide of the searching range of the latent group numbers. We conduct numerous experiments on three real datasets from Shenyang, Changchun, and Harbin, which are the three provincial capital cities that have similar geography, economics, and cultures. The experimental results demonstrate the effectiveness of the proposed framework.
- 31 August 2020
- Published by Association for Computing Machinery (ACM) in ACM/IMS Transactions on Data Science
- Vol. 1 (3) , 1-16
Open AccessAdaptive Weighted Finite Mixture Model
It is significant for real estate investors to understand how the construction environments and building characteristics impact the housing unit price. However, it is challenging for identifying the complex feature-influence from construction environments and building characteristics. It is also hard to alleviate the heterogeneity of real estate. In this article, we propose a framework named Adaptive Weighted Finite Mixture Model to identify the feature-influence and simultaneously alleviate the ill effect of heterogeneity. Applying this framework, we can predict the housing unit price based on the corresponding features. Besides, we discover that the feature-influence exists in the dissimilarity among similar cities. Specifically, we adaptively learn the weights of features to identify the feature-influence, and we model the estimation of the housing unit price with the feature-influence into a finite mixture model. We utilize the Principle Component Analysis algorithm to obtain a low-dimensional representation of housing features. The low-dimensional representation reduces the computational cost of model learning, and it avoids a potential catastrophe of the singular matrix inversion during the process of learning model parameters, which are estimated by the Expectation Maximization algorithm. To avoid the blind search for the latent group number used in the proposed framework, we employ the pre-clustering result as a guide of the searching range of the latent group numbers. We conduct numerous experiments on three real datasets from Shenyang, Changchun, and Harbin, which are the three provincial capital cities that have similar geography, economics, and cultures. The experimental results demonstrate the effectiveness of the proposed framework.
- Open Access
- 19 April 2019
- Published by SAGE Publications in Advances in Mechanical Engineering
- Vol. 11 (4)
A novel fuzzy model for multi-objective permutation flow shop scheduling problem with fuzzy processing time
This article proposes a novel fuzzy model for solving fuzzy multi-objective permutation flow shop scheduling problem with fuzzy processing time. Specifically, two fuzzy objectives, that is, the fuzzy makespan and the fuzzy total flow time, are taken into account in this model simultaneously. In addition, to solve fuzzy multi-objective permutation flow shop scheduling problem, an efficient algorithm called fuzzy multi-objective local search-based decomposition is proposed. In order to generate a high quality and diverse set of initial solutions, a problem-specific Nawaz–Enscore–Ham heuristic approach is incorporated into the framework as an initialization. Then, two perturbation strategies with different strength are adopted to find better solutions and to avoid local optimum as well. Besides, a ranking concept based on fuzzy number centroid is provided to compare with fuzzy solutions. After that, a restart strategy is employed to change the searching space when the best solution has not been improved for a certain number of iterations. Finally, we conduct an extensive computational study on Taillard benchmarks to compare the proposed fuzzy multi-objective local search-based decomposition with the fuzzy NSGAII. Experimental results demonstrate that the fuzzy multi-objective local search-based decomposition algorithm is both effective and efficient in solving the fuzzy multi-objective permutation flow shop scheduling problem.
- 19 April 2019
- Published by SAGE Publications in Advances in Mechanical Engineering
- Vol. 11 (4)
Open AccessA novel fuzzy model for multi-objective permutation flow shop scheduling problem with fuzzy processing time
This article proposes a novel fuzzy model for solving fuzzy multi-objective permutation flow shop scheduling problem with fuzzy processing time. Specifically, two fuzzy objectives, that is, the fuzzy makespan and the fuzzy total flow time, are taken into account in this model simultaneously. In addition, to solve fuzzy multi-objective permutation flow shop scheduling problem, an efficient algorithm called fuzzy multi-objective local search-based decomposition is proposed. In order to generate a high quality and diverse set of initial solutions, a problem-specific Nawaz–Enscore–Ham heuristic approach is incorporated into the framework as an initialization. Then, two perturbation strategies with different strength are adopted to find better solutions and to avoid local optimum as well. Besides, a ranking concept based on fuzzy number centroid is provided to compare with fuzzy solutions. After that, a restart strategy is employed to change the searching space when the best solution has not been improved for a certain number of iterations. Finally, we conduct an extensive computational study on Taillard benchmarks to compare the proposed fuzzy multi-objective local search-based decomposition with the fuzzy NSGAII. Experimental results demonstrate that the fuzzy multi-objective local search-based decomposition algorithm is both effective and efficient in solving the fuzzy multi-objective permutation flow shop scheduling problem.
- 1 November 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2014, 11-18
Dynamic multi-objective evolutionary algorithm based on decomposition for test task scheduling problem
Test task scheduling problem in the dynamic environment (DTTSP) is an important issue in automatic test system. In this paper, a dynamic multi-objective evolutionary algorithm based on decomposition (DMOEA/D) is proposed to improve the adaptability of the environment changes in test process. The mathematical model considering the arrival of dynamic tasks is proposed based on the Markov decision process. Three standard test functions and two DTTSP examples are used in experiment for illustrating the performance of the proposed algorithm. The results show that the proposed algorithm has good performance in convergence and diversity. Almost all the performance metrics of convergence and diversity obtain stable statistical results. The result of convergence ratio of an algorithm is not good as other metrics because of the slow convergence rate. The results also show that the solutions obtained by DMOEA/D have better Pareto front than the dynamic multi-objective particle swarm optimization algorithm (DMOPSO).
- 1 November 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 2014, 11-18
Dynamic multi-objective evolutionary algorithm based on decomposition for test task scheduling problem
Test task scheduling problem in the dynamic environment (DTTSP) is an important issue in automatic test system. In this paper, a dynamic multi-objective evolutionary algorithm based on decomposition (DMOEA/D) is proposed to improve the adaptability of the environment changes in test process. The mathematical model considering the arrival of dynamic tasks is proposed based on the Markov decision process. Three standard test functions and two DTTSP examples are used in experiment for illustrating the performance of the proposed algorithm. The results show that the proposed algorithm has good performance in convergence and diversity. Almost all the performance metrics of convergence and diversity obtain stable statistical results. The result of convergence ratio of an algorithm is not good as other metrics because of the slow convergence rate. The results also show that the solutions obtained by DMOEA/D have better Pareto front than the dynamic multi-objective particle swarm optimization algorithm (DMOPSO).