Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning

Evaluating the environmental perception of urban parks is highly significant for optimizing urban planning. To address the limitations of traditional evaluation methods, a multimodal deep learning framework that integrates pre-training and reinforcement learning strategies for the comprehensive asse...

Full description

Saved in:
Bibliographic Details
Main Authors: Kangen Chen, Tao Xia, Zhoutong Cao, Yiwen Li, Xiuhong Lin, Rushan Bai
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Buildings
Subjects:
Online Access:https://www.mdpi.com/2075-5309/15/13/2364
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850117908608319488
author Kangen Chen
Tao Xia
Zhoutong Cao
Yiwen Li
Xiuhong Lin
Rushan Bai
author_facet Kangen Chen
Tao Xia
Zhoutong Cao
Yiwen Li
Xiuhong Lin
Rushan Bai
author_sort Kangen Chen
collection DOAJ
description Evaluating the environmental perception of urban parks is highly significant for optimizing urban planning. To address the limitations of traditional evaluation methods, a multimodal deep learning framework that integrates pre-training and reinforcement learning strategies for the comprehensive assessment of various park types (seaside, urban, mountain, and wetland) across three dimensions—accessibility, usability, and aesthetics—is proposed herein. By combining image data and user review texts, a unified architecture is constructed, including a text encoder, image visual encoder, and multimodal fusion module. During the pre-training phase, the model captured latent features in images and texts through a self-supervised learning strategy. In the subsequent training phase, a reinforcement learning strategy was introduced to optimize the sample selection and modal fusion paths to enhance the model’s generalization capability. To validate the cross-type prediction ability of the model, the experimental design uses data from three types of parks for training, with the remaining type as a test set. Results demonstrate that the proposed method outperforms LSTM and CNN architectures across accuracy, precision, recall, and F1 Score metrics. Compared with CNN, the proposed method improves accuracy by 5.1% and F1 Score by 6.6%. Further analysis shows that pre-training enhances the robust fusion of visual and textual features, while reinforcement learning optimizes the sample selection and feature fusion strategies during training.
format Article
id doaj-art-7e31472a3e044f059f0bda3a24f1ec00
institution OA Journals
issn 2075-5309
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Buildings
spelling doaj-art-7e31472a3e044f059f0bda3a24f1ec002025-08-20T02:35:59ZengMDPI AGBuildings2075-53092025-07-011513236410.3390/buildings15132364Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement LearningKangen Chen0Tao Xia1Zhoutong Cao2Yiwen Li3Xiuhong Lin4Rushan Bai5Faculty of Innovation and Design, City University of Macau, Macau 999078, ChinaZhuhai Dechuang Construction Engineering Consulting Co., Ltd., Zhuhai 519000, ChinaLeeds University Business School, University of Leeds, Leeds LS2 9JT, UKLand Development and Reclamation Center of Guangdong Province, Guangzhou 510635, ChinaFaculty of Innovation and Design, City University of Macau, Macau 999078, ChinaFaculty of Innovation and Design, City University of Macau, Macau 999078, ChinaEvaluating the environmental perception of urban parks is highly significant for optimizing urban planning. To address the limitations of traditional evaluation methods, a multimodal deep learning framework that integrates pre-training and reinforcement learning strategies for the comprehensive assessment of various park types (seaside, urban, mountain, and wetland) across three dimensions—accessibility, usability, and aesthetics—is proposed herein. By combining image data and user review texts, a unified architecture is constructed, including a text encoder, image visual encoder, and multimodal fusion module. During the pre-training phase, the model captured latent features in images and texts through a self-supervised learning strategy. In the subsequent training phase, a reinforcement learning strategy was introduced to optimize the sample selection and modal fusion paths to enhance the model’s generalization capability. To validate the cross-type prediction ability of the model, the experimental design uses data from three types of parks for training, with the remaining type as a test set. Results demonstrate that the proposed method outperforms LSTM and CNN architectures across accuracy, precision, recall, and F1 Score metrics. Compared with CNN, the proposed method improves accuracy by 5.1% and F1 Score by 6.6%. Further analysis shows that pre-training enhances the robust fusion of visual and textual features, while reinforcement learning optimizes the sample selection and feature fusion strategies during training.https://www.mdpi.com/2075-5309/15/13/2364park environment perceptioncross-type generalizationpre-trainingzero-shot learningreinforcement learning
spellingShingle Kangen Chen
Tao Xia
Zhoutong Cao
Yiwen Li
Xiuhong Lin
Rushan Bai
Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning
Buildings
park environment perception
cross-type generalization
pre-training
zero-shot learning
reinforcement learning
title Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning
title_full Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning
title_fullStr Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning
title_full_unstemmed Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning
title_short Predictive Models for Environmental Perception in Multi-Type Parks and Their Generalization Ability: Integrating Pre-Training and Reinforcement Learning
title_sort predictive models for environmental perception in multi type parks and their generalization ability integrating pre training and reinforcement learning
topic park environment perception
cross-type generalization
pre-training
zero-shot learning
reinforcement learning
url https://www.mdpi.com/2075-5309/15/13/2364
work_keys_str_mv AT kangenchen predictivemodelsforenvironmentalperceptioninmultitypeparksandtheirgeneralizationabilityintegratingpretrainingandreinforcementlearning
AT taoxia predictivemodelsforenvironmentalperceptioninmultitypeparksandtheirgeneralizationabilityintegratingpretrainingandreinforcementlearning
AT zhoutongcao predictivemodelsforenvironmentalperceptioninmultitypeparksandtheirgeneralizationabilityintegratingpretrainingandreinforcementlearning
AT yiwenli predictivemodelsforenvironmentalperceptioninmultitypeparksandtheirgeneralizationabilityintegratingpretrainingandreinforcementlearning
AT xiuhonglin predictivemodelsforenvironmentalperceptioninmultitypeparksandtheirgeneralizationabilityintegratingpretrainingandreinforcementlearning
AT rushanbai predictivemodelsforenvironmentalperceptioninmultitypeparksandtheirgeneralizationabilityintegratingpretrainingandreinforcementlearning