近年來，隨著電腦需求市場的快速變化，產品測試和驗證階段至關重要，以確保產品功能的正常運作。然而，現有的問題追蹤系統在查詢異常問題時僅能使用關鍵字，且搜尋出來的結果需人工逐筆過濾與判斷，導致效率低下且耗費大量時間。在快速發展的人工智慧技術中，生成式AI因其在文字處理上展現了驚人的創造力與對自然語言的理解能力，在醫療、金融等領域被快速採用。然而在過去研究中較少有研究將大型語言模型(Large Language Model, LLM)應用於電腦製造業中，因此本研究將針對電腦產品之測試案例所產生的異常事件，應用LLM建置一個要因預測分析系統，並評估零樣本與少樣本提示工程在回答上的表現，使LLM在電腦產品測試領域中能夠更有效地理解專業知識。該系統將允許測試人員以自然語言提問，系統將生成可能的要因與解決方案，提供測試人員未知或未注意到的觀點和思路，特別是針對較無經驗之新進人員，能夠協助其更快速地識別出問題發生之要因，同時達到組織內部的知識共享。;Product testing and verification have become essential to ensuring the proper functioning of products. However, current issue tracking systems only allow users to employ keywords to query abnormal issues, and results still need to be manually filtered and judged by an expert, leading to low efficiency and taking considerable time. Generative artificial intelligence has been adopted in several industries, including health care and finance, due to its creativity and ability to understand natural language; however, large language models (LLMs), a type of generative artificial intelligence, have not yet been extensively applied in the computer manufacturing industry. The present study implemented Llama 2 to build a cause prediction analysis system that can predict abnormalities in computer products. At the same time, this study evaluates the performance of zero-shot and few-shot prompt engineering to enhance its understanding of domain-specific knowledge. With the LLM, a tester can ask questions using natural language, and the LLM generates possible causes and solutions, providing perspectives and ideas that the tester may be unaware of or have overlooked. The system is especially beneficial for newcomers—it helps them rapidly identify the causes of issues and facilitates knowledge sharing within an organization.
