Xinye Sha

Mr. Xinye Sha

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
17 February 2025 in Scientific Reports
Verified
Journal Article
Open access
PDF file
DWDedai Wei
Dedai Wei
ZWZimo Wang
Zimo Wang
MQMinyu Qiu
Minyu Qiu
JYJuntao Yu
Juntao Yu
JYJiaquan Yu
Jiaquan Yu
KOKaichen Ouyang
Kaichen Ouyang
Stock market prediction has long attracted the attention of academia and industry due to its potential for substantial financial returns. Despite the availability of various forecasting methods, such as CNN, LSTM, BiLSTM, GRU, and Transformer, the hyperparameter optimization of these models often faces limitations, particularly in single-objective optimization, where they can easily fall into local optima. To address this issue, this paper proposes an innovative multi-objective optimization algorithm—the Multi-Objective Escape Bird Algorithm (MOEBS)—and introduces the MOEBS-Transformer architecture to enhance the efficiency and effectiveness of hyper-parameter optimization for Transformer models. This study first validates the performance of MOEBS through a series of multi-objective benchmark tests on standard problem sets such as ZDT, DTLZ, and WFG, comparing it with other multi-objective optimization algorithms (e.g., MOMVO, MSSA, and MOEAD) using evaluation metrics such as GD, Spacing, IGD, and HV for comprehensive analysis. In the context of stock price prediction, we select the closing price datasets of Amazon, Google, and Uniqlo, using MOEBS to optimize the core hyper parameters of the Transformer while considering multiple objectives, including training set RMSE, testing set RMSE, and testing set error variance. In the experiments, this paper first compares CNN, LSTM, BiLSTM, GRU, and traditional Transformer models to establish the Transformer as the optimal model for stock market prediction. Subsequently, the study compares the MOEBS-Transformer with Transformer models optimized using various hyperparameter optimization methods, including MOMVO-Transformer, MSSA-Transformer, and MOEAD-Transformer. Additionally, it evaluates Transformer models optimized through conventional methods: Random Search (RS-Transformer), Grid Search (GS-Transformer), and Bayesian Optimization (BO-Transformer). By assessing the performance of these models using R2, RMSE, and RPD metrics on both training and testing sets, the results demonstrate that the Transformer model optimized by MOEBS significantly outperforms the other methods in terms of prediction accuracy and prediction stability. This research offers a new solution for complex optimization scenarios and lays a foundation for advancements in stock market prediction technologies.

4 September 2025 in Journal of Advanced Research
Verified
Journal Article
Open access
KOKaichen Ouyang
Kaichen Ouyang
DWDedai Wei
Dedai Wei
XSXinye Sha
Xinye Sha
JYJuntao Yu
Juntao Yu
YZYuanli Zhao
Yuanli Zhao
Huiling ChenHuiling Chen
Huiling Chen
Numerical optimization plays a key role in improving the efficiency of solar photovoltaic (PV) systems and solving complex engineering problems. Traditional optimization methods often struggle with finding optimal solutions within a reasonable timeframe due to high-dimensional and non-linear problem landscapes. This study aims to introduce a novel swarm intelligence algorithm, the Beaver Behavior Optimizer (BBO), inspired by the cooperative behaviors of beavers during dam construction. The goal is to validate BBO's performance on both benchmark test functions and real-world engineering problems, particularly in solar PV parameter optimization. The BBO was modeled based on beaver behaviors of material gathering and dam maintenance, integrating exploration and exploitation phases. To assess its performance, experiments were conducted using CEC 2017 and CEC 2022 benchmark functions with varying dimensions (10, 20, 30, 50, 100). Statistical significance was verified using Wilcoxon signed-rank and Friedman mean rank tests. Furthermore, BBO was applied to solve three solar PV parameter identification problems and four real-world engineering problems, comparing its performance with 11 other algorithms. BBO demonstrated superior performance across all benchmark functions and ranked first in tackling solar PV and engineering design problems. It outperformed other state-of-the-art algorithms in most test scenarios, showcasing robust convergence, quick optimization, and minimal variance in results. The results validate BBO as a powerful optimization tool, particularly for solar PV parameter identification and engineering challenges. Its bio-inspired approach effectively balances exploration and exploitation, making it a competitive algorithm for solving complex optimization tasks.The source code of BBO is available at: https://github.com/oykc1234/BBO.

24 May 2024
Verified
Conference Paper
Open access
HSHongyi Shui
Hongyi Shui
XSXinye Sha
Xinye Sha
BCBaizheng Chen
Baizheng Chen
JWJiajie Wu
Jiajie Wu
Stock price prediction is essential yet challenging in financial markets, guiding investment decisions for stakeholders. While traditional methods rely on technical and fundamental analysis, the integration of big data and machine learning presents new opportunities for enhanced accuracy. This paper introduces a novel approach for stock price prediction, utilizing feature engineering and the LightGBM model. Feature engineering, a crucial step in machine learning, is employed to select or create relevant features, focusing on capturing the weighted average price from historical trading data. The LightGBM model, known for its efficiency and accuracy, is harnessed to improve prediction accuracy. Our study contributes by offering a comprehensive framework that integrates advanced machine learning techniques with domain-specific feature engineering. Real-world stock market data validates the effectiveness of our approach, outperforming traditional methods. Our contributions include leveraging LightGBM with Optuna fine-tuning for enhanced prediction accuracy, introducing a novel feature engineering approach for WAP incorporation, and demonstrating superior performance through extensive experiments, marking a significant advancement in stock price prediction methodologies.

26 July 2024
Verified
Conference Paper
Open access
XSXinye Sha
Xinye Sha
In order to understand the application of computer technology in financial investment, the author proposes a research on the application of computer technology in financial investment. The author used user transaction data from a certain online payment platform as a sample, with a total of 284908 sample records, including 593 positive samples (fraud samples) and 285214 negative samples (normal samples), to conduct an empirical study on user fraud detection based on data mining. In this process, facing the problem of imbalanced positive and negative samples, the author proposes to use the Under Sampling method to construct sub samples, and then perform feature scaling, outlier detection, feature screening and other processing on the sub samples. Then, four classification models, logistic regression, K-nearest neighbor algorithm, decision tree, and support vector machine, are trained on the processed sub samples. The prediction results of the four models are evaluated, and the results show that the recall rate, Fl score, and AUC value of the logistic regression model are the highest, indicating that the detection method based on computer data mining is practical and feasible.

10 September 2024 in IEEE Access
Verified
Journal Article
Open access
YGYujia Gu
Yujia Gu
BLBrian Liu
Brian Liu
TZTianlong Zhang
Tianlong Zhang
XSXinye Sha
Xinye Sha
SCShiyong Chen
Shiyong Chen
The integration of visual and linguistic elements within artificial intelligence research is increasingly emphasized, spurred by advancements in pre-trained model technologies. Traditionally, such models have been developed independently, using methods like contrastive learning and image-captioning to boost their analytical and creative outputs. This paper introduces an innovative architecture known as the Zero-shot Unified Image-Text (ZsU-IT) framework, which synthesizes pre-training objectives into a cohesive Unicode-decoder structure. The ZsU-IT is intricately designed with distinct components for image and text processing, coupled with a bi-modal decoder, which seamlessly manages both encoding and decoding tasks across various functions. This dual functionality promotes an effective knowledge transfer between the visual and linguistic modalities, thereby enhancing the system’s adaptability and efficiency in tasks like image-to-text translation and vice versa. Rigorous empirical studies reveal that ZsU-IT outstrips prevailing models across multiple applications, including image and text retrieval, image captioning, Visual Question Answering (VQA), and Stanford Natural Language Inference - Visual Entailment (SNLI-VE). This is particularly notable in complex settings involving sophisticated datasets such as medical texts and CT images. In zero-shot environments, ZsU-IT excels, displaying exceptional generalization capabilities. This prowess is highlighted by its significant achievements. The ZsU-IT framework not only sets a newbenchmark in the fusion of vision and language technologies but also fosters novel opportunities for both ongoing research and practical implementations. This advancement marks a crucial step forward in the application of integrated multimodal data for complex problem-solving within the artificial intelligence landscape, paving the way for future breakthroughs.

17 August 2025 in Scientific Reports
Verified
Journal Article
Open access
DWDedai Wei
Dedai Wei
ZWZimo Wang
Zimo Wang
HKHanfu Kang
Hanfu Kang
XSXinye Sha
Xinye Sha
YXYiran Xie
Yiran Xie
KOKaichen Ouyang
Kaichen Ouyang
In the context of global economic transformation, high-quality enterprise development (HQED) is crucial for driving economic growth, particularly through enhancing Total Factor Productivity (TFPLP). Digital Inclusive Finance (DIF), as a classical financial model, plays an important role in promoting high-quality enterprise development. To explore the relationship between TFP and DIF, we first applied traditional double fixed-effects models, along with robustness and heterogeneity tests, for modeling experiments. This series of tests effectively revealed the theoretical linear relationships between economic variables. However, the double fixed-effects model has limitations in capturing nonlinear relationships and making predictions. Given the growing body of research on existing hybrid models, we acknowledge the importance of exploring and contributing to this evolving area. To address this issue, based on the results of traditional economic analysis, we introduced improved time series models. These advanced deep learning models allow us to better capture the complex nonlinear relationship between DIF and TFP. The experiment initially explored the preliminary structural relationship between DIF and TFP using double fixed-effects models combined with robustness and heterogeneity tests. Then, based on the results of these tests, we selected deep learning features and combined Kolmogorov-Arnold Neural Network (KAN), Graph Neural Network (GNN) models with classic time series deep learning models (Transformer, LSTM, BiLSTM, GRU) to capture the latent nonlinear features in the data for prediction. The results show that, compared to traditional time series forecasting methods, the improved deep learning models perform better in capturing the nonlinear relationships of economic variables, improving prediction accuracy, and reducing prediction errors. Finally, paired t-tests and Cohen's d effect size tests were used to evaluate error metrics, and the results indicate that the introduction of KAN and GNN models significantly improved the performance of time series forecasting models.