Improving fashion trend prediction using an ontology-based approach



Fashion is a huge revenue-generating industry affecting millions of people worldwide.  Fashion trends play a big role in influencing various stakeholders’ decisions.  Fashion Trend Prediction (FTP) is a complex task that traditionally involved trained professionals analysing large amount of information.  Artificial intelligence (AI) techniques can help in automating and simplifying this task for humans.  In this project, we used a human-centred approach to explore the design of an ideal FTP system and implemented an early prototype that focussed on predicting colour.  A basic ontology was built based on a core vocabulary used for describing colour in fashion.  This was utilised in both visual and textual analyses to derive an overall output.  The ontology provided a bridge for the integration of visual and non-visual elements in fashion and forms a basis for future work where more complex reasoning tasks can be performed.  Possible avenues of exploration include extending the ontology to other identified relevant trend features such as texture, style, and fabric and working on developing appropriate algorithms for prediction.  Carrying out user evaluation on the system is also of interest. 


The goal of this project is to build on previous work to further explore and improve a fashion trend predictions system.  This can involve doing user studies to better understand user needs and building a fashion ontology that uses relevant vocabulary and captures important relations between various concepts.  We expect to be able to use a neuro-symbolic approach to improve fashion trend prediction results.


The project student is expected to have the following:

  • a strong programming background (e.g. has completed COMP2100 or COMP6442)
  • has done logic-based courses or machine learning or both
  • has preferably done the course COMP3900
  • has an interest to learn about neuro-symbolic approaches and ontology
  • is planning to do a research project in at least two semesters



fashion, trend prediction, Artificial Intelligence, ontology, human-centred computing, design, neuro-symbolic

Updated:  10 August 2021/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing