A ROBUST MULTI-SENSOR DEEP REGRESSION FRAMEWORK FOR PREDICTIVE PLANT GROWTH MODELLING IN CONTROLLED ENVIRONMENTS

Abstract

The analysis of plant growth and development in controlled agricultural environments has become increasingly important to ensure sustainable and efficient food production. Traditionally, plant monitoring relied on manual observations and basic statistical approaches, which provided limited insights into the complex interactions between environmental factors and plant physiology. To address these challenges, this study proposes a machine learning–driven analytical framework designed for comprehensive plant development analysis. The system integrates data preprocessing, exploratory data analysis, classification, regression, and hybrid deep learning approaches within a unified pipeline. Classification algorithms such as Support Vector Classifier (SVC), Bernoulli Naive Bayes (BNC), and Multinomial Naive Bayes (MNC) are employed to identify plant growth stages. Regression models, including Decision Tree Regressor (DTR), Support Vector Regressor (SVR), and Ridge Regressor (RR), are utilized to estimate growth-related parameters. Furthermore, two hybrid models are introduced to enhance predictive performance. The Deep Feature Probabilistic Classifier (DFPC) combines Feed Forward Neural Networks (FFNN) with Gaussian Naive Bayes (GNB) for improved classification accuracy, while the Hybrid Deep Ridge Predictor (HDRP) integrates FFNN with Ridge Regressor to achieve precise regression outcomes. Experimental results demonstrate that the DFPC model attains an accuracy of 94.42%, whereas the HDRP model achieves an R² score of 0.999. These findings highlight the effectiveness of combining deep learning and machine learning techniques for accurate plant growth analysis and informed decision-making in controlled agricultural systems.

Plant phenotyping relevance

植物の成長段階と成長関連パラメータを推定する機械学習・深層学習パイプラインが研究の中心であり、表現型推定手法の開発に該当する。

Code and data availability

The article describes a CSV-based multi-sensor plant growth dataset and a hybrid deep learning + Ridge regression pipeline, but provides no public dataset link, repository, code deposit, or availability statement. All URLs in the text are citations to prior published works (references [2]-[19]), which are not paper-own

No evidence-backed public reproduction asset is currently recorded.