Predictive Machine Learning Framework for Urban Ornamental Plant Recommendation via Multi-Parameter Environmental and Air Quality Assessment

Abstract

Purpose

Urban environments face increasing challenges related to air pollution, limited green spaces, and environmental sustainability. Ornamental plants can improve urban environmental quality; however, selecting the most suitable plant species for specific urban conditions remains difficult. This study aims to develop an intelligent recommender system that assists users in selecting suitable ornamental plants based on environmental conditions and plant performance indicators.

Objectives

The research focuses on five key objectives: (1) Identifying important environmental factors affecting ornamental plant suitability. (2) Developing a linear regression model for predicting Hybrid Recommendation Scores (HRS). (3)  Designing a recommender system for ornamental plant selection. (4) Ranking plant species based on predicted suitability scores. (5)  Evaluating the effectiveness of the proposed recommendation model in urban environments.

Methodology

 Environmental and plant-related parameters, including sunlight, season, plant density, and growth rate were collected and analysed. A multiple linear regression model was developed to estimate the SHybrid Recommendation Score (HRS). The predicted scores were integrated into a recommendation engine that ranks ornamental plants according to their suitability under different urban conditions. Statistical validation was performed using paired sample t-tests to evaluate the reliability of the proposed system.

"While advanced computational architectures such as Support Vector Machines (SVM), Deep Learning, and Transformers represent potential avenues for scaling high-complexity environmental datasets, they fall outside the immediate implementation scope of this study. The current architecture deliberately focuses on Multiple Linear Regression and Ensemble Learning models (Random Forest and Gradient Boosting) to maintain computational efficiency, explainability, and direct interpretability for urban horticulture applications."

Try-outs

The proposed recommender system was evaluated using environmental and plant-related parameters including sunlight, season, plant suitability score, and growth rate. These variables were processed through the developed linear regression model to calculate the Hybrid Recommendation Score (HRS).

The calculated HRS values were compared with benchmark HRS values obtained from previous studies to assess prediction accuracy. Furthermore, statistical validation was performed using a paired sample t-test to determine whether significant differences existed between the standard and predicted scores.

The calculated Hybrid Recommendation Scores (HRS) showed strong agreement with benchmark values obtained from published studies. Statistical validation confirmed that no significant difference existed between the standard and predicted values, demonstrating the reliability of the proposed regression model. The developed recommender system successfully ranked ornamental plants according to environmental suitability and can support sustainable urban greening initiatives by providing intelligent plant recommendations.