Lihong Zhang

Dr. Lihong Zhang


Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 September 2023 in Applied Sciences
Verified
Journal Article
Open access
PDF file
LZLihong Zhang
Lihong Zhang
CLChaolong Liu
Chaolong Liu
NJNan Jia
Nan Jia
Multimodal emotion classification (MEC) has been extensively studied in human–computer interaction, healthcare, and other domains. Previous MEC research has utilized identical multimodal annotations (IMAs) to train unimodal models, hindering the learning of effective unimodal representations due to differences between unimodal expressions and multimodal perceptions. Additionally, most MEC fusion techniques fail to consider the unimodal–multimodal inconsistencies. This study addresses two important issues in MEC: learning satisfactory unimodal representations of emotion and accounting for unimodal–multimodal inconsistencies during the fusion process. To tackle these challenges, the authors propose the Two-Stage Conformer-based MEC model (Uni2Mul) with two key innovations: (1) in stage one, unimodal models are trained using independent unimodal annotations (IUAs) to optimize unimodal emotion representations; (2) in stage two, a Conformer-based architecture is employed to fuse the unimodal representations learned in stage one and predict IMAs, accounting for unimodal–multimodal differences. The proposed model is evaluated on the CH-SIMS dataset. The experimental results demonstrate that Uni2Mul outperforms baseline models. This study makes two key contributions: (1) the use of IUAs improves unimodal learning; (2) the two-stage approach addresses unimodal–multimodal inconsistencies during Conformer-based fusion. Uni2Mul advances MEC by enhancing unimodal representation learning and Conformer-based fusion.

13 August 2025 in Symmetry
Verified
Journal Article
Open access
PDF file
Zhengguang GaoZhengguang Gao
Zhengguang Gao
JJJing Shen Jia
Jing Shen Jia
LZLihong Zhang
Lihong Zhang
This study introduces the concept of symmetry as a fundamental theoretical perspective for understanding the linguistic structure of cyberbullying texts. It posits that such texts often exhibit symmetry breaking between surface-level language forms and underlying semantic intent. This structural-semantic asymmetry increases the complexity of the recognition task and places higher demands on the semantic modeling capabilities of detection systems. With the rapid growth of social media, the covert and harmful nature of cyberbullying speech has become increasingly prominent, posing serious challenges to public opinion management and public safety. While mainstream approaches to cyberbullying detection—typically based on traditional deep learning models or pre-trained language models—have achieved some progress, they still struggle with low accuracy, poor generalization, and weak interpretability when handling implicit, semantically complex, or borderline expressions. To address these challenges, this paper proposes a cyberbullying detection method that combines LoRA-based fine-tuning with Small-Scale Hard-Sample Adaptive Training (S-HAT), leveraging a large language model framework based on Meta-Llama-3-8B-Instruct. The method employs prompt-based techniques to identify inference failures and integrates model-generated reasoning paths for lightweight fine-tuning. This enhances the model’s ability to capture and represent semantic asymmetry in cyberbullying texts. Experiments conducted on the ToxiCN dataset demonstrate that the S-HAT approach achieves a precision of 84.1% using only 24 hard samples—significantly outperforming baseline models such as BERT and RoBERTa. The proposed method not only improves recognition accuracy but also enhances model interpretability and deployment efficiency, offering a practical and intelligent solution for cyberbullying mitigation.

5 July 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
CMChen Ma
Chen Ma
LZLihong Zhang
Lihong Zhang
ZXZhi Xing
Zhi Xing
JSJunjing Su
Junjing Su
Credit card fraud detection is a typical rare-event classification problem because fraudulent transactions usually account for only a very small proportion of all transactions. Conventional evaluation on balanced or resampled test data may lead to overly optimistic performance estimates. To address this issue, this study proposes a leakage-safe credit card fraud detection framework integrating Random Forest Gini impurity-based feature selection, resampling strategy evaluation, and Genetic Algorithm (GA)-based hyperparameter optimization. The framework was evaluated on the public European credit card fraud dataset containing 284,807 transactions, of which only 492 were fraudulent. The original dataset was first divided into a stratified training set and an untouched original-distribution test set. Feature selection, standardization, resampling, GA optimization, and threshold tuning were performed only on the training data or training folds. The final test set contained 85,443 transactions, including 148 fraudulent transactions, and was used only once for final evaluation. Experimental results show that GA-XGBoost achieved the best overall balance among the optimized models, with a PR-AUC of 0.798, ROC-AUC of 0.967, MCC of 0.814, balanced accuracy of 0.865, fraud-class precision of 0.908, fraud-class recall of 0.730, and fraud-class F1-score of 0.809. Compared with baseline XGBoost, GA-XGBoost improved PR-AUC from 0.741 to 0.798, MCC from 0.766 to 0.814, and fraud-class F1-score from 0.764 to 0.809, while reducing false positives from 22 to 11 and false negatives from 43 to 40. The ablation results further indicate that resampling strategies are not universally beneficial and should be evaluated under the original test distribution. These findings suggest that leakage-safe evaluation and fraud-class-oriented metrics provide a more reliable basis for practical credit card fraud detection.