Objective We propose a multimodal model integrating social media text and image data for automated assessment of psychological stress in college students to support the development of intelligent mental health services in higher education institutions. Methods Based on deep learning technology, we designed an evaluation framework comprising a text sentiment modeling module, an image sentiment modeling module, and a multimodal fusion prediction module. Text sentiment features were extracted using Bi-LSTM, and image semantic cues were extracted via U-Net. A feature concatenation strategy was used to enable cross-modal semantic collaboration to achieve automatic identification of 3 psychological stress levels: mild, moderate, and severe. We constructed a multimodal annotated dataset using social platform data from 1577 students across multiple universities in Guangdong Province. After data cleaning, 252 samples were randomly selected for model training and testing. Results In the 3-classification task, the model demonstrated outstanding performance on the test set, and achieved an accuracy of 92.86% and an F1 score of 0.9276, exhibiting excellent stability and consistency. Confusion matrix analysis further revealed the model's ability to effectively distinguish between different pressure levels. Conclusion The multimodal psychological stress assessment model developed in this study effectively integrates unstructured social behavior data to enhance the scientific rigor and practical applicability of psychological state recognition, and thus provides support for developing intelligent psychological service systems.
FanYY, LiuJ, ZengYY, et al. Factors associated with non-suicidal self-injury in Chinese adolescents: a meta-analysis[J]. Front Psychiatry, 2021, 12: 747031. doi:10.3389/fpsyt.2021.747031
[2]
HuangJP, NigatuYT, Smail-CrevierR, et al. Interventions for common mental health problems among university and college students: a systematic review and meta-analysis of randomized controlled trials[J]. J Psychiatr Res, 2018, 107: 1-10. doi:10.1016/j.jpsychires.2018.09.018
[3]
BruffaertsR, MortierP, KiekensG, et al. Mental health problems in college freshmen: Prevalence and academic functioning[J]. J Affect Disord, 2018, 225: 97-103. doi:10.1016/j.jad.2017.07.044
[4]
CageE, JonesE, RyanG, et al. Student mental health and transitions into, through and out of university: student and staff perspectives[J]. J Furth High Educ, 2021, 45(8): 1076-89. doi:10.1080/0309877x.2021.1875203
[5]
PedrelliP, NyerM, YeungA, et al. College students: mental health problems and treatment considerations[J]. Acad Psychiatry, 2015, 39(5): 503-11. doi:10.1007/s40596-014-0205-9
[6]
HickeyBA, ChalmersT, NewtonP, et al. Smart devices and wearable technologies to detect and monitor mental health conditions and stress: a systematic review[J]. Sensors (Basel), 2021, 21(10): 3461. doi:10.3390/s21103461
[7]
OphirY, TikochinskiR, AsterhanCSC, et al. Deep neural networks detect suicide risk from textual facebook posts[J]. Sci Rep, 2020, 10(1): 16685. doi:10.1038/s41598-020-73917-0
[8]
夏先益. 基于文本挖掘的在线论坛用户心理健康自动评估[D]. 南昌: 江西财经大学, 2019.
[9]
张凤云. 基于ON-LSTM的文本情绪分析方法研究[D]. 郑州: 郑州大学, 2020.
[10]
ColeEJ, PhillipsAL, BentzleyBS, et al. Stanford neuromodulation therapy (SNT): a double-blind randomized controlled trial[J]. Am J Psychiatry, 2022, 179(2): 132-41. doi:10.1176/appi.ajp.2021.20101429
[11]
BakerJ, NgoH, EfthimiouTN, et al. Electrical stimulation of smiling muscles reduces visual processing load and enhances happiness perception in neutral faces[J]. Commun Psychol, 2025, 3(1): 94. doi:10.1038/s44271-025-00281-y
ZhuQJ, XiongJC, PengLL. College students' mental health evaluation model based on tensor fusion network with multimodal data during the COVID-19 pandemic[J]. Biotechnol Genet Eng Rev, 2024, 40(3): 1821-35. doi:10.1080/02648725.2023.2196846
[14]
FuZW, LiuF, XuQ, et al. LMR-CBT: learning modality-fused representations with CB-Transformer for multimodal emotion recognition from unaligned multimodal sequences[J]. Front Comput Sci, 2023, 18(4): 184314. doi:10.1007/s11704-023-2444-y
[15]
KalınkaraY, TalanT. Psychological balances in the digital world: dynamic relationships among social media addiction, depression, anxiety, academic self-efficacy, general belongingness, and life satisfaction[J]. J Genet Psychol, 2025, 186(2): 85-113. doi:10.1080/00221325.2024.2400342
[16]
Chandra GuntukuS, BuffoneA, JaidkaK, et al. Understanding and measuring psychological stress using social media[J]. Proc Int AAAI Conf Web Soc Medium, 2019, 13: 214-25. doi:10.1609/icwsm.v13i01.3223
[17]
WolfersLN, UtzS. Social media use, stress, and coping[J]. Curr Opin Psychol, 2022, 45: 101305. doi:10.1016/j.copsyc.2022.101305
[18]
KhooLS, LimMK, ChongCY, et al. Machine learning for multimodal mental health detection: a systematic review of passive sensing approaches[J]. Sensors (Basel), 2024, 24(2): 348. doi:10.3390/s24020348
[19]
DuC, LiuC, BalamuruganP, et al. Deep learning-based mental health monitoring scheme for college students using convolutional neural network[J]. Int J Artif Intell Tools, 2021, 30(6n08): 2140014.
[20]
AsadMM, ErumD, ChuriP, et al. Effect of technostress on psychological well-being of post-graduate students: a perspective and correlational study of higher education management[J]. Int J Inf Manag Data Insights, 2023, 3(1): 100149. doi:10.1016/j.jjimei.2022.100149
[21]
TaylorJM. Psychometric analysis of the ten-item perceived stress scale[J]. Psychol Assess, 2015, 27(1): 90-101. doi:10.1037/a0038100
ZhaoF, ZhangCC, GengBC. Deep multimodal data fusion[J]. ACM Comput Surv, 2024, 56(9): 1-36. doi:10.1145/3649447
[24]
MuktaMSH, AhmadJ, ZamanA, et al. Attention and meta-heuristic based general self-efficacy prediction model from multimodal social media dataset[J]. IEEE Access, 2024, 12: 36853-73. doi:10.1109/access.2024.3373558
[25]
DengH, YangZG, HaoTY, et al. Multimodal affective computing with dense fusion transformer for inter- and intra-modality interactions[J]. IEEE Trans Multimed, 2022, 25: 6575-87. doi:10.1109/tmm.2022.3211197
[26]
ZhaoYX, CaoXY, LinJL, et al. Multimodal affective states recognition based on multiscale CNNs and biologically inspired decision fusion model[J]. IEEE Trans Affect Comput, 2023, 14(2): 1391-403. doi:10.1109/taffc.2021.3093923
[27]
LiWB, TanRC, XingY, et al. A multimodal psychological, physiological and behavioural dataset for human emotions in driving tasks[J]. Sci Data, 2022, 9(1): 481. doi:10.1038/s41597-022-01557-2
[28]
ZhuLN, ZhuZC, ZhangCW, et al. Multimodal sentiment analysis based on fusion methods: a survey[J]. Inf Fusion, 2023, 95: 306-25. doi:10.1016/j.inffus.2023.02.028
[29]
UniversityDP, Saleem AbdullahSM, AbdulazeezAM, et al. Facial expression recognition based on deep learning convolution neural network: a review[J]. J Soft Comput Data Min, 2021, 2(1): 53-65.
FuLY, LiSW. A new semantic segmentation framework based on UNet[J]. Sensors (Basel), 2023, 23(19): 8123. doi:10.3390/s23198123
[32]
WangX, JingSH, DaiHF, et al. High-resolution remote sensing images semantic segmentation using improved UNet and SegNet[J]. Comput Electr Eng, 2023, 108: 108734. doi:10.1016/j.compeleceng.2023.108734
[33]
AbdollahiA, PradhanB, AlamriAM. An ensemble architecture of deep convolutional Segnet and Unet networks for building semantic segmentation from high-resolution aerial images[J]. Geocarto Int, 2022, 37(12): 3355-70. doi:10.1080/10106049.2020.1856199
[34]
BeheraRK, JenaM, RathSK, et al. Co-LSTM: Convolutional LSTM model for sentiment analysis in social big data[J]. Inf Process Manag, 2021, 58(1): 102435. doi:10.1016/j.ipm.2020.102435
[35]
LindemannB, MüllerT, VietzH, et al. A survey on long short-term memory networks for time series prediction[J]. Procedia CIRP, 2021, 99: 650-5. doi:10.1016/j.procir.2021.03.088
[36]
ShanmuganathanV, SureshA. LSTM-Markov based efficient anomaly detection algorithm for IoT environment[J]. Appl Soft Comput, 2023, 136: 110054. doi:10.1016/j.asoc.2023.110054
[37]
WuJ, ZhuTL, ZhuJH, et al. A optimized BERT for multimodal sentiment analysis[J]. ACM Trans Multimedia Comput Commun Appl, 2023, 19(2s): 1-12. doi:10.1145/3566126