Code Migration from AngularJS to Angular Using Artificial Intelligence
Burman, Ashutosh (2025)
2025
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025051512024
https://urn.fi/URN:NBN:fi:amk-2025051512024
Tiivistelmä
The feasibility and effectiveness of using artificial intelligence (AI) models for prompt-based code migration from legacy AngularJS applications to the modern Angular framework were analysed. Migration was carried out on a real-world application commissioned by Music.Info Finland Oy. to evaluate how far AI could automate the process with minimal human intervention. Multiple AI models OpenAI o3-mini (high), DeepSeek R1, Gemini 2.0 Pro, and Claude 3.5 Sonnet were as-sessed using the LiveCodeBench benchmark, with OpenAI o3-mini (high) selected for based on the coding performance. A qualitative research method was adopted, to analyse the AI generated code.
The results showed that AI was successfully migrated components, services, and styles into Angu-lar modern standards, handling routine transformations efficiently. However, several limitations were observed, such as syntax errors, incomplete dependency injection, and a lack of understand-ing of full project architecture, requiring human validation and corrections. AI indicated strong potential for reducing developer workload in basic tasks but lacks in successful migration without manual supervision. Future research was suggested to fine-tune AI models for specific migration tasks and integrate AI systems into development environments to enhance performance, reliabil-ity, and sustainability in large-scale code migration projects.
The results showed that AI was successfully migrated components, services, and styles into Angu-lar modern standards, handling routine transformations efficiently. However, several limitations were observed, such as syntax errors, incomplete dependency injection, and a lack of understand-ing of full project architecture, requiring human validation and corrections. AI indicated strong potential for reducing developer workload in basic tasks but lacks in successful migration without manual supervision. Future research was suggested to fine-tune AI models for specific migration tasks and integrate AI systems into development environments to enhance performance, reliabil-ity, and sustainability in large-scale code migration projects.