CLASNet: A Cognitive Load–Aware CNN-LSTM-Attention Framework for Supply Chain Demand Forecasting and Adaptive Human–Computer Interaction

Zheyu Li; Yue Hao; Peifan Zeng

doi:10.63313/JCSFT.9069

Authors

Zheyu Li Wuhan University, Wuhan 430072, Hubei, China Author
Yue Hao Johns Hopkins University, Baltimore, MD, USA Author
Peifan Zeng New York University, New York, NY, USA Author

DOI:

https://doi.org/10.63313/JCSFT.9069

Keywords:

Supply chain demand forecasting, Cognitive load awareness, CNN-LSTM, Attention mechanism, Adaptive human–computer interaction, Deep learning

Abstract

This paper proposes CLASNet, a Cognitive Load–Aware CNN-LSTM-Attention framework for supply chain demand forecasting and adaptive human–computer interaction. Unlike conventional deep learning forecasting approaches that primarily emphasize predictive accuracy while neglecting human cognitive constraints, CLASNet integrates demand prediction and cognitive load modeling within a unified end-to-end architecture to improve both forecasting performance and decision-support efficiency. The proposed framework combines a CNN-LSTM demand feature encoder for capturing local temporal patterns and long-term dependencies, a Cognitive Load Encoder (CLE) for modeling user interaction behaviors, a Cognitive-Aware Attention Fusion (CAAF) mechanism that dynamically adjusts temporal attention weights according to users’ cognitive states, and an Adaptive Interaction Decision Layer (AIDL) that optimizes visualization complexity and information density. Experiments were conducted on a real-world retail supply chain dataset containing three years of demand records and dashboard interaction logs collected from intelligent decision-support systems. Comparative results against ARIMA, LSTM, CNN-LSTM, and CNN-LSTM-Attention baselines demonstrate that CLASNet achieves superior forecasting accuracy, obtaining an RMSE of 14.55, MAE of 9.22, and MAPE of 10.9%, corresponding to an approximately 9.6% RMSE reduction compared with the strongest baseline. Ablation studies further verify the effectiveness of the cognitive load encoder, cognitive-aware attention mechanism, and adaptive interaction module. In addition, user-centered evaluations indicate that the proposed cognitive-aware adaptive interface significantly reduces task completion time and perceived cognitive load during demand analysis tasks. Overall, the results demonstrate that integrating cognitive awareness into deep learning–based supply chain forecasting systems can substantially enhance both predictive performance and human–AI collaborative decision-making efficiency.

References

[1] Fan P, Li H, Hu M. Profit-Oriented Production and Pricing Optimization for Manufacturing Enterprises Using Proximal Policy Optimization[J]. Economics and Management Innovation, 2026, 3(2): 8-17.

[2] Computation N. Long short-term memory[J]. Neural Comput, 2016, 9: 1735-1780.

[3] Gers F A, Schmidhuber J, Cummins F. Learning to forget: Continual prediction with LSTM[J]. Neural computation, 2000, 12(10): 2451-2471.

[4] Bai S, Kolter J Z, Koltun V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[J]. arXiv preprint arXiv:1803.01271, 2018.

[5] Borovykh A, Bohte S, Oosterlee C W. Conditional time series forecasting with convolutional neural networks[J]. arXiv preprint arXiv:1703.04691, 2017.

[6] Bahdanau D, Cho K, Bengio Y. Neural machine translation by jointly learning to align and translate[J]. arXiv preprint arXiv:1409.0473, 2014.

[7] Qin Y, Song D, Chen H, et al. A dual-stage attention-based recurrent neural network for time series prediction[J]. arXiv preprint arXiv:1704.02971, 2017.

[8] Sweller J. Cognitive load during problem solving: Effects on learning[J]. Cognitive science, 1988, 12(2): 257-285.

[9] Joseph A W, Murugesh R. Potential eye tracking metrics and indicators to measure cognitive load in human-computer interaction research[J]. J. Sci. Res, 2020, 64(1): 168-175.

[10] Guo Q, Agichtein E. Towards predicting web searcher gaze position from mouse movements[M]//CHI'10 extended abstracts on human factors in computing systems. 2010: 3601-3606.

[11] Chen S, Epps J, Chen F. A comparison of four methods for cognitive load measurement[C]//Proceedings of the 23rd Australian computer-human interaction conference. 2011: 76-79.

[12] Heer J, Shneiderman B. Interactive dynamics for visual analysis: A taxonomy of tools that support the fluent and flexible use of visualizations[J]. Queue, 2012, 10(2): 30-55.

[13] Waller M A, Fawcett S E. Data science, predictive analytics, and big data: a revolution that will transform supply chain design and management[J]. Journal of Business logistics, 2013, 34(2): 77-84.

[14] Zhang G, Zeng H, Jiang L. Uni-FinLLM: A Unified Multimodal Large Language Model with Modular Task Heads for Micro-Level Stock Prediction and Macro-Level Systemic Risk Assessment[J]. arXiv preprint arXiv:2601.02677, 2026.

[15] Li X, Li Z, Lin X. Automated Implementation of Machine Learning-Based Causal Inference in Product Operations Decision Making[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 204-209.

[16] Li T, Li H, Zhou Y. E-commerce Sentiment Analysis Using Fine-tuned LLaMA3 Models: A QLoRA-based Approach[J]. Journal of Technology Innovation and Engineering, 2025, 1(4).

[17] Xu S, Jiang L, Gu B. Design and Validation of a Smart Neuromorphic System Architecture for Algorithmic Trading[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 127-136.

[18] Zhang, Shanshan, et al. "Hybrid Beamforming for Subarray-Level Movable Antenna Enhanced MU-MIMO Communications." IEEE Wireless Communications Letters 15 (2026): 2559-2563.