Integrasi Model AI pada Platform Web Modular Prediksi Penyakit Tanaman dengan Metode Strategy Pattern

Abstract

Artificial Intelligence, Strategy Design Pattern, Software Architecture, Modular Web Platform, Plant Disease Prediction, FastAPI, CNN, YOLOv5, Vision Transformer, AI Integration, Modular AI System

Abstract This research focuses on the development of a modular web-based platform for plant disease prediction through the integration of multiple Artificial Intelligence (AI) models using the Strategy Design Pattern. The study addresses the architectural challenges commonly found in AI systems, particularly tight coupling, low extensibility, and difficulties in integrating or replacing AI models without modifying the core system structure.

The proposed system applies the Strategy Design Pattern to separate AI model implementations from the main application logic, enabling dynamic model selection and flexible integration. The platform was developed using a modular architecture with FastAPI as the backend framework and React-based frontend integration. Several AI models, including Convolutional Neural Network (CNN), YOLOv5, and Vision Transformer (ViT), were integrated as independent strategies for plant disease prediction.

System evaluation was conducted using formal design pattern metrics, namely Property Satisfaction Rate (PSR), Critical Property Coverage (CPC), and Pattern Implementation Quality Score (PIQS). Additional evaluation was performed at the code-design level through cohesion, coupling, and extensibility analysis. The results indicate that the implementation successfully satisfies the structural and behavioral characteristics of the Strategy Design Pattern, achieving high modularity, maintainability, and extensibility. Furthermore, the platform supports the addition of new AI models without requiring significant changes to the core architecture.

This research demonstrates that the Strategy Design Pattern is an effective architectural approach for developing scalable and maintainable AI-based web platforms, particularly for systems requiring multi-model AI integration in digital agriculture applications.