ارزیابی ادراکات دانشجویان از اثرات زیان‌بار میکروپلاستیک‌ها به منظور تقویت آگاهی زیست‌محیطی: رویکرد مبتنی بر شبکه‌ عصبی مصنوعی

Abstract
Introduction
Microplastics (MPs)—plastic particles smaller than 5 mm—are pervasive pollutants that threaten ecosystems and human health across terrestrial, aquatic, and atmospheric environments. By 2015, global plastic production had reached 6,300 million tons, of which only 9% was recycled, and projections indicate that plastics will account for 20% of global oil consumption by 2050. In the Persian Gulf—a semi-enclosed sea—anthropogenic activities such as oil extraction, industrialization, and coastal development contribute between 155,000 and 413,000 tons of plastic waste annually. MPs infiltrate food chains, induce toxicity, and pose considerable health risks; humans are estimated to ingest between 39,000 and 74,000 microplastic particles each year. The widespread use of single-use plastics, coupled with inadequate waste management systems, has further intensified this pollution crisis. Environmental education, informed by theoretical frameworks such as social identity theory and experiential learning, plays a pivotal role in fostering sustainable behaviors. In Bushehr, urban runoff and maritime activities further exacerbate MP contamination. Although global research increasingly explores strategies for reducing plastic pollution, port-specific sustainability initiatives and educational interventions remain insufficiently addressed in the Middle East.
The novelty of this study lies in applying an artificial neural network (ANN) to assess and rank microplastic awareness among students at Persian Gulf University, thereby integrating machine learning with educational strategies to propose actionable, context-specific solutions for MP mitigation. The research contributes by identifying education-driven dimensions of awareness and offering policy recommendations tailored to the Persian Gulf region.
 
Methodology
This quantitative, developmental-applied study employed a descriptive-survey design to evaluate MP awareness among Persian Gulf University students during 2024–2025. The target population included 4,000 students across bachelors, masters, and PhD levels. A simple random sample of 450 participants was selected using Cochran’s formula (95% confidence level; 5% margin of error), yielding 408 valid responses (195 female, 213 male; 228 undergraduate, 153 master’s, 27 PhD), and representing a 90.67% response rate. Inclusion criteria required active enrollment and accessibility.
Data collection utilized a researcher-developed questionnaire containing 23 items: 22 closed-ended questions on a 5-point Likert scale measuring awareness, attitudes, and behaviors related to MP reduction, and one open-ended question identifying perceived sources of pollution (e.g., seafood, water bottles). Content validity was confirmed by five experts in environmental science and sustainability, and internal consistency reliability was high (Cronbach’s α = 0.87). Data were analyzed using a three-layer ANN (input, hidden, output) with softmax activation and a cross-entropy loss function. Variables were normalized using (x − min)/(x − max) and divided into training (70.1%, n = 282), testing (29.4%, n = 120), and holdout (1.5%, n = 6) sets. Independent variables included demographic characteristics (gender, age, education level, prior experience) and six thematic dimensions: toxicity awareness, environmental damage, degradation deterrents, regulatory strengthening, aquatic controls, and knowledge promotion. The ANN model ranked variable importance and predicted perceptions of MP-related hazards.
Results and Discussion
The ANN demonstrated strong performance, achieving cross-entropy errors of 0.1% (training), 0.05% (testing), and 0.05% (holdout) after five-fold cross-validation, with overfitting mitigated through L2 regularization and early stopping. Classification accuracy reached 95.2% for the training dataset and 94.8% for the holdout dataset, with sensitivities of 96.5% for non-hazardous and 95.6% for hazardous MP perceptions. Cumulative gains analysis showed that 40% of the dataset accounted for 62% of high-risk cases, with 100% hazard detection obtained beyond the 68th percentile.
Key predictive variables included awareness of environmental damage, livestock ingestion, soil toxicity, cancer risks, non-recyclability, and toxic production processes—each significantly moderated by education level. Open-ended responses revealed gender-based differences: female respondents emphasized seafood contamination, whereas male participants highlighted water bottles and synthetic clothing as major sources. Six core dimensions emerged from the analysis: (1) toxicity awareness (99% education-driven deterrence), (2) environmental damage, (3) degradation deterrents, (4) regulatory strengthening (57.4% education-linked), (5) aquatic controls (77.2% education-linked), and (6) awareness gaps (44.9% education-linked).
These findings align with previous research on MP toxicity, ecosystem disruption, aquaculture threats, and deficits in public awareness. The pronounced role of education in behavioral deterrence underscores the importance of integrated policy-education strategies. The observed gender differences further suggest the relevance of tailored educational messaging to enhance the effectiveness of MP reduction interventions.
Conclusions
This study confirms that university-level education significantly enhances microplastic awareness, identifying six key dimensions through ANN-based analysis. It directly answers the research question regarding education’s influence on MP awareness, revealing its substantial impact—particularly a 99% association with deterrence behaviors and a 44.9% contribution to reducing awareness gaps. Gender-specific concerns highlight the need for differentiated communication strategies. Practical recommendations include implementing targeted educational programs, enforcing stricter regulations, and providing incentives to reduce plastic consumption—interventions that offer co-benefits for aquaculture, tourism, and public health. The study’s significance lies in its innovative use of machine learning and its region-specific focus, offering a transferable model for sustainable microplastic management in the Persian Gulf. Future research should extend to broader demographic groups and adopt longitudinal approaches to evaluate the durability of educational impacts.