Zijian Zhou

Zijian Zhou

Auto-generated profile, based on publicly available information

Zijian Zhou


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

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

Representative Publications

View All Publications
27 September 2026 in Sensors
Verified
Journal Article
Open access
PDF file
ZZZijian Zhou
Zijian Zhou
RLRuijia Liu
Ruijia Liu
XLXiangchen Long
Xiangchen Long
YHYongbiao Hu
Yongbiao Hu
FXFei Xia
Fei Xia
YSYihong Song
Yihong Song
Extreme climate events increasingly amplify agricultural trade security risks across interconnected stages: agricultural production, commodity conditions, cold-chain storage, logistics distribution, and trade fulfillment. However, existing approaches predominantly focus on isolated stages and fail to model cross-stage risk propagation, enable real-time edge inference, or provide well-calibrated risk warnings. To address these challenges, a multimodal edge-intelligence framework, termed AgriClimate-EdgeNet, is proposed to jointly model climatic conditions, agricultural production, commodity imagery, cold-chain states, logistics trajectories, and trade records. An extreme-climate-aware cross-modal consistency mechanism is developed to capture normal inter-stage correspondence and identify abnormal information conflicts. Depthwise separable temporal convolutions, gated temporal units, lightweight attention, and Teacher–Student distillation are incorporated for efficient edge inference. Furthermore, dynamic modality reliability estimation and dual predictive uncertainty modeling (decoupling epistemic and heteroscedastic aleatoric uncertainties) are integrated to ensure decision trustworthiness under severe sensory noise and missing observations. On a 38,400-window agricultural trade dataset, AgriClimate-EdgeNet achieves an Accuracy of 0.914, Recall of 0.896, Macro-F1 of 0.902, and AUC of 0.949, while reducing expected calibration error to 0.028 in routine single-pass Streaming Mode and 0.021 under multi-sample Deep Audit Mode. In operational edge deployment on NVIDIA Jetson AGX Orin, the model requires 3.96 M on-device parameters and 1.21 G FLOPs, achieving an inference latency of 8.6 ms (116.3 samples/s) in Streaming Mode and 38.4 ms in Deep Audit Mode. These results demonstrate that AgriClimate-EdgeNet provides an accurate, robust, and low-latency solution for full-chain agricultural trade risk sensing under extreme climate shocks.

22 August 2026 in Sensors
Verified
Journal Article
Open access
PDF file
ZZZijian Zhou
Zijian Zhou
NWNuo Wang
Nuo Wang
SXShengzhe Xu
Shengzhe Xu
SHSurui Hua
Surui Hua
HWHanyang Wang
Hanyang Wang
MLManzhou Li
Manzhou Li
Short-term financial risk in intelligent trading systems is reflected not only in prices, trading volumes, and textual sentiment but also in infrastructure operating states, including server workload, device power consumption, network latency, and packet loss rate. We propose HSF-LLMNet, a hardware sensor and large language model semantic fusion network for jointly modeling external information shocks and infrastructure responses. A large language model extracts event category, sentiment polarity, risk intensity, and semantic uncertainty from financial texts. Reliability-aware temporal modeling handles sensor missingness, drift, and abnormal noise, while asynchronous soft alignment, bidirectional cross-attention, and reliability-aware gated fusion integrate irregular textual events with continuous hardware signals. The model jointly predicts market direction, realized volatility, and three-level risk over the subsequent 30 min. Experiments were conducted on eight Chinese A-share indices: the SSE Composite Index (000001.SH), SSE 50 Index (000016.SH), CSI 300 Index (000300.SH), STAR 50 Index (000688.SH), CSI 500 Index (000905.SH), CSI 1000 Index (000852.SH), Shenzhen Component Index (399001.SZ), and ChiNext Index (399006.SZ). The common observation period for market, textual, and hardware data extended from 1 March 2024 to 30 June 2025. After data cleaning, timestamp matching, and multimodal temporal alignment, 169,208 aligned asset–time prediction windows were retained for the 30 min forecasting task. Realized volatility was defined as the square root of the sum of squared one-minute log returns over the future 30 min interval. The three-level risk label was constructed from future realized volatility, absolute 30 min return, and liquidity stress, with all thresholds estimated exclusively from the training portion of each fold. A sample was labeled high risk when at least two of the three indicators exceeded their 85th-percentile thresholds or when any indicator exceeded its 95th-percentile threshold. It was labeled medium risk when, after excluding high-risk samples, at least two indicators exceeded their 60th-percentile thresholds or any indicator exceeded its 85th-percentile threshold; all remaining samples were labeled low risk. Results showed that HSF-LLMNet achieved an accuracy of 78.62%, a precision of 78.14%, a recall of 77.83%, an F1-score of 77.98%, an area under the receiver operating characteristic curve of 84.91%, and a Matthews correlation coefficient of 57.36% for directional prediction. For realized-volatility regression, the MAE, RMSE, MAPE, and R2 were 0.0089, 0.0135, 9.21%, and 0.812, respectively. For high-risk-event warning, the mean effective warning time, defined as the interval between the first valid alarm and the corresponding event, was 15.37 min; the false-alarm rate and missed-alarm rate were 6.82% and 8.14%, respectively. Ablation experiments showed performance reductions after removing semantic encoding, sensor-reliability estimation, asynchronous alignment, bidirectional cross-attention, gated fusion, or multi-task learning. These results indicate that textual events and infrastructure operating states provide complementary information for quantitative risk analytics and fintech applications.