Masters Theses

Abstract

Accurate and timely prediction of pest outbreaks is a cornerstone of Agriculture 5.0, which emphasizes intelligent, data-driven, and sustainable decision-making in crop production. This research presents a multimodal deep learning framework that integrates heterogeneous data sources, including weather parameters, satellite-derived vegetation indices, and static and dynamic soil attributes, to forecast pest population dynamics under varying management and ecological conditions. The proposed framework employs modality-specific deep encoders to capture distinct temporal and spatial representations from each data stream and merges them through a late-fusion architecture that learns cross-modal dependencies critical to pest emergence. The design further incorporates treatment-aware and multiclass extensions, allowing the model to differentiate pest dynamics across management regimes such as Conventional Management and Integrated Pest Management (IPM), while simultaneously forecasting multiple pest categories within a unified framework. This multimodal approach enhances interpretability, scalability, and ecological relevance by linking environmental, spectral, and agronomic factors to observed pest patterns. The results demonstrate the framework’s potential to support precision pest forecasting and sustainable decision-making across diverse cropping systems, establishing a foundation for real-time, adaptive pest management aligned with the principles of Agriculture 5.0.

Advisor(s)

Das, Sajal K.

Committee Member(s)

Arifuzzaman, Md
Morales, Ricardo

Department(s)

Computer Science

Degree Name

M.S. in Computer Science

Publisher

Missouri University of Science and Technology

Publication Date

Fall 2025

Pagination

x, 76 pages

Note about bibliography

Includes_bibliographical_references_(pages 70-75)

Rights

© 2026 V N S Kameswari Sri Sindhu Manchikanti , All Rights Reserved

Document Type

Thesis - Open Access

File Type

text

Language

English

Thesis Number

T 12571

Share

 
COinS