#South Korea #Pusan National University #Busan #ChatGPT #Fashion Trends

Fashion Trend Forecasting with AI: A New Approach from Pusan National University

Introduction

Fashion trend forecasting is an essential component in the clothing industry, enabling companies to anticipate which styles will dominate in future seasons. Traditionally, this process has been reliant on the creative insights of industry experts. However, researchers from Pusan National University in South Korea are pioneering a new method using artificial intelligence (AI), specifically the ChatGPT model, to revolutionize fashion trend predictions. This innovative approach promises to make forecasting more accessible, particularly for fashion students and smaller brands that may not have the resources to conduct extensive data analysis.

The Traditional Model and Its Limitations

Historically, fashion trend predictions have depended heavily on the intuition and experience of top designers and trend forecasters. These experts analyze cultural shifts, societal trends, and broader consumer behavior to make educated guesses about future fashion. While their insights can offer valuable direction, the approach often lacks systematic analysis, limiting its effectiveness, especially in a rapidly changing market where data-driven decisions are key.

The Role of AI in Modern Forecasting

The introduction of AI, particularly large language models (LLMs) like ChatGPT, has the potential to democratize this analytical capability. ChatGPT, which processes vast amounts of societal and cultural data, can assist in identifying patterns and predicting future trends. However, it has been critiqued for issues like generating inaccurate data or hallucinations, which raise questions about its reliability in trend forecasting.

Research Breakthrough

In light of these challenges, Assistant Professor Yoon Kyung Lee and Master's student Chaehi Ryu from the Department of Clothing and Textiles recently published a study on a new strategy that harnesses ChatGPT for fashion trend prediction. Their findings were published in the Clothing and Textiles Research Journal on September 26, 2025. Dr. Lee emphasizes that they devised an innovative systematic prompting strategy, moving beyond generic questions to more targeted inquiries that yield specific and reliable answers.

The Top-Down Prompting (TDP) Technique

This novel approach, termed Top-Down Prompting (TDP), draws from the Lotus Blossom brainstorming method. TDP initiates with a broad question on fashion trends and cascades into more detailed queries regarding various aspects of fashion, including silhouettes, fabrics, key items, and color palettes. For their research, the team used this structured technique to predict men's fashion trends for the fall/winter 2024 season, employing both ChatGPT-3.5 and ChatGPT-4 Classic.

The efficacy of the predictions was then compared against those of the Official Fashion Trend Information Company (OFTIC) and analyzed by fashion experts. Although the AI's forecast reflected some established trends, it was found to accurately identify only 9 out of the 39 trends outlined in OFTIC's report. Yet, intriguingly, the AI models also pinpointed emergent themes such as gender fluidity and bold outerwear, which were not part of existing datasets.

Implications for the Fashion Industry

While the research indicates that ChatGPT does not yet replace expert analysis, its ability to recognize and articulate new and unexpected trends presents a valuable tool for fashion students and small brands that may lack the resources to conduct comprehensive analyses. This collaborative potential between AI and human insight could foster a new dynamic in trend forecasting, providing a more nuanced understanding of what consumers may desire in the future.

Dr. Lee notes, "What’s particularly remarkable is that despite their lower accuracy, the AI tools captured cultural shifts, hinting at untapped creative possibilities." Thus, while traditional forecasting may still hold significant weight, the integration of AI represents a necessary evolution in the fashion industry, facilitating more systemic and inclusive approaches to trend analysis.

Future Directions

This research not only paves the way for enhanced forecasting models but also encourages the integration of technology within fashion education. The development of a TDP-based framework tailored for training future fashion professionals underscores the potential for AI to enrich educational methodologies, expanding students' understanding of how to pinpoint trends within a complex marketplace. With continued advancements, AI could well become an indispensable ally in the quest for trend accuracy, creativity, and innovation within the fashion realm.

This study exemplifies how AI can streamline fashion trend forecasting, making it not just a tool for industry giants but an accessible resource for all aspiring fashion creators.