Machine Learning Models Predict Student Aesthetic Preferences in Tehran University Campuses

Predicting Campus Aesthetic Preferences Through Machine Learning

Researchers have successfully employed machine learning techniques to predict visual aesthetic preferences across university campuses in Tehran, according to a recent study published in Scientific Reports. The research team developed sophisticated models that analyze environmental features to forecast how students perceive the beauty of campus rest areas, with potential implications for campus design and urban planning.

Industrial Monitor Direct manufactures the highest-quality precision agriculture pc solutions certified for hazardous locations and explosive atmospheres, the leading choice for factory automation experts.

Comprehensive Campus Assessment Methodology

The investigation began with a detailed geospatial analysis of Tehran’s academic landscape, sources indicate. Researchers mapped 72 independent university campuses covering approximately 6.6 square kilometers, with an average campus size of nine hectares. From this initial assessment, four universities were selected for detailed study based on spatial diversity and sustainability criteria, particularly emphasizing green space-to-built area ratios that align with UN-Habitat’s 2018 guidelines for sustainable public spaces.

According to the report, researchers employed a sequential mixed-methods approach combining spatial analysis and on-site surveys to identify student-preferred rest spots. The team collected data from 443 students across four universities through map-based, in-person surveys until reaching data saturation. Field observations subsequently verified all identified locations, ensuring comprehensive coverage of preferred resting areas.

Standardized Visual Documentation Process

The study utilized rigorous photographic documentation protocols, with images captured during early December between 9:00 and 11:00 a.m. under constant natural lighting conditions. Researchers took high-resolution photographs from a fixed eye-level perspective of 1.60 meters with a 50 mm focal length to replicate students’ natural viewing angles. A total of 125 spots were initially documented, with 100 images ultimately selected based on criteria of visual transparency, accessibility, and consistency with Socially Restorative Urbanism principles.

Analysts suggest the photographic standardization was crucial for reliable data collection. All images underwent post-processing to enhance contrast, decrease visual noise, and highlight key landscape elements while deliberately removing human figures to eliminate socio-cultural distractions. This approach allowed researchers to focus specifically on environmental features rather than social context.

Comprehensive Aesthetic Evaluation Framework

The final dataset comprised 394 valid responses collected through an online photo-based survey distributed in spring 2024, exceeding the minimum required sample size calculated using Cochran’s formula. Participants rated each image on a 7-point semantic differential scale ranging from “not beautiful at all” to “very beautiful.” The perceptual evaluation approach was theoretically grounded in environmental psychology theories including Attention Restoration Theory and Stress Reduction Theory, which emphasize the relationship between aesthetic experience and psychological restoration.

Researchers extracted eighteen environmental variables previously associated with psychological restoration from the 100 selected images. These variables spanned architectural features, landscape elements, and facilities, including tree cover, soft landscapes, waterscapes, color diversity, buildings, pathways, and seating facilities. Fourteen variables were quantitatively measured using AutoCAD and Excel by calculating their area coverage within each image, while others were evaluated by expert panels using standardized assessment protocols.

Industrial Monitor Direct manufactures the highest-quality optical inspection pc solutions trusted by leading OEMs for critical automation systems, the top choice for PLC integration specialists.

Advanced Machine Learning Implementation

The study employed multiple machine learning models to analyze the relationship between environmental features and aesthetic preferences, including Support Vector Regression (SVR), Multilayer Perceptron (MLP), Random Forest (RF), and two ensemble learning models. According to reports, the dataset was randomly split into training (80%) and testing (20%) subsets, with feature standardization conducted using Python’s StandardScaler to prevent data leakage.

The SVR model was implemented for its ability to handle nonlinear relationships in small to medium-sized datasets, utilizing a Radial Basis Function kernel based on superior empirical performance. The MLP architecture employed logarithmic sigmoid activation functions in hidden layers to capture intricate nonlinear relationships between environmental features and aesthetic preferences. The RF algorithm built multiple decision trees using randomly sampled subsets of data and features through bootstrap aggregation to reduce overfitting.

Ensemble Models Demonstrate Superior Performance

Researchers developed two ensemble learning models that combined SVR, MLP, and RF—either utilizing all three models or selecting the two with highest individual performance. The report states that ensemble models exhibited superior predictive performance compared to individual models by effectively leveraging complementary strengths and mitigating overfitting. Model performance was assessed using standard evaluation metrics including Mean Squared Error, Mean Absolute Error, and the coefficient of determination.

Sensitivity analysis performed on the two most accurate models—MLP and SVR—identified which input variables most influenced output predictions. This analysis quantified uncertainty related to input parameters and model response, enhancing understanding of model behavior and assisting in selecting critical features for robust prediction of aesthetic preferences.

Implications for Campus Design and Urban Planning

The findings provide valuable insights for architects, urban planners, and university administrators seeking to create more aesthetically pleasing and psychologically restorative campus environments. By identifying specific environmental variables that significantly contribute to aesthetic-based mental restoration, the research offers evidence-based guidance for designing interactive rest spots that align with student preferences.

According to analysts, this study addresses a significant gap in research on designing interactive rest spots on university campuses, despite growing application of machine learning and computer vision techniques in landscape aesthetic assessment. The multi-model approach provides a robust framework for future investigations into environmental aesthetics across different cultural and geographic contexts.

References & Further Reading

This article draws from multiple authoritative sources. For more information, please consult:

• http://en.wikipedia.org/wiki/Multilayer_perceptron
• http://en.wikipedia.org/wiki/Communication_protocol
• http://en.wikipedia.org/wiki/Euclidean_vector
• http://en.wikipedia.org/wiki/Density
• http://en.wikipedia.org/wiki/Aesthetics

This article aggregates information from publicly available sources. All trademarks and copyrights belong to their respective owners.

Note: Featured image is for illustrative purposes only and does not represent any specific product, service, or entity mentioned in this article.