Guavafusionformer: A Dual-Branch Soil-Climate Fusion Network for Intelligent Guava Health Classification

Abstract Guava is a fruit economically relevant in both tropical and subtropical climates and accurate assessment of health is critical to ensure sustainable production and yield. However, traditional methods might be inadequate in capturing complex relationships existing in diverse soil-climatic variables. This paper presents GuavaFusionFormer, a two-branch soil-climate fusion model for intelligent classification of guava health. The model consists of nitrogen, phosphorus, potassium, pH, soil moisture, temperature, humidity, and rainfall fed into the Dual-Branch Contextual Attention Encoder (DCAE), Adaptive Feature Enhancement (AFE), Multi-scale Dependency Learning (MDL), Adaptive Soil-Climate Fusion (ASCF), and Gated Residual Network (GRN). Employing the Guava Soil-Climate Disease Dataset collected from Maharashtra, India, the framework makes classification of eight guava health and disease states. The GuavaFusionFormer model obtains 0.9754% accuracy, 0.9757% macro precision, 0.9754% macro recall, and 0.9755% macro F1-score, and its specificity and FPR are 0.9965% and 0.35%, respectively. In addition to evaluating its classification capability via comparative experiments on traditional and Transformer-based deep learning approaches, SHAP and t-SNE analysis provides an interpretation of feature contribution and representation separability for guava health.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-11317368/v1
Primary Topic
Smart Agriculture and AI
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Guavafusionformer: A Dual-Branch Soil-Climate Fusion Network for Intelligent Guava Health Classification

Archana Dirgule, Archana Pakhare, Amarsinh Vidhate, Tukaram Chavan
3 links
Research Square
Smart Agriculture and AI
preprint

Guavafusionformer: A Dual-Branch Soil-Climate Fusion Network for Intelligent Guava Health Classification

Archana Dirgule, Archana Pakhare, Amarsinh Vidhate, Tukaram Chavan
preprint en

Abstract

Abstract Guava is a fruit economically relevant in both tropical and subtropical climates and accurate assessment of health is critical to ensure sustainable production and yield. However, traditional methods might be inadequate in capturing complex relationships existing in diverse soil-climatic variables. This paper presents GuavaFusionFormer, a two-branch soil-climate fusion model for intelligent classification of guava health. The model consists of nitrogen, phosphorus, potassium, pH, soil moisture, temperature, humidity, and rainfall fed into the Dual-Branch Contextual Attention Encoder (DCAE), Adaptive Feature Enhancement (AFE), Multi-scale Dependency Learning (MDL), Adaptive Soil-Climate Fusion (ASCF), and Gated Residual Network (GRN). Employing the Guava Soil-Climate Disease Dataset collected from Maharashtra, India, the framework makes classification of eight guava health and disease states. The GuavaFusionFormer model obtains 0.9754% accuracy, 0.9757% macro precision, 0.9754% macro recall, and 0.9755% macro F1-score, and its specificity and FPR are 0.9965% and 0.35%, respectively. In addition to evaluating its classification capability via comparative experiments on traditional and Transformer-based deep learning approaches, SHAP and t-SNE analysis provides an interpretation of feature contribution and representation separability for guava health.

Research Square
DY Patil Deemed to be University (IN), MIT Art, Design and Technology University (IN)
Zero hunger
Smart Agriculture and AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon