Abstract:Abstract The objectives of this study were to apply alternative machine learning (ML) algorithms to predict consumers’ garment fit satisfactions (real fit satisfaction [RFS]) and compare the efficiencies of these algorithms to predict RFS. Skirts made from different fabrics were used as test garments. Mechanical properties of the skirts’ fabrics were assigned as predictor variables to estimate RFS. Study participants’ virtual body models were created by using 3D body scanner and used for virtual fitting. Each participant physically tried on the skirts and evaluated the fit. Participants also viewed the skirt simulations on their avatars and evaluated the virtual fit, which represented participants’ virtual fit satisfactions (VFS). Random Forest (RF), support vector machine (SVM), and conditional tree (CT) algorithms were used to learn from the data to predict participants’ RFSs. The mean correlations between the predicted and observed RFS values in the validation sets were 0.74 (RF), 0.70 (SVM-linear kernel), 0.72 (SVM-radial kernel), and 0.55 (CT). According to the variable importance analysis, VFS had the highest importance among 35 predictor variables. ML is used mostly for sales forecasting and manufacturing purposes in the fashion industry. However, garment fit, which restrains consumers from shopping online, did not get enough attention in ML studies. Besides, the ML algorithms used in fashion and apparel studies are often genetic algorithms and neural networks; therefore, there is a need to test other algorithm types. In this study, we offered alternative ML algorithms (i.e., RF, SVM, and CT) to predict consumers’ garment fit satisfactions.

What problem does this paper attempt to address?

The paper attempts to address the issue of reducing return rates due to poor fit by predicting consumer satisfaction with clothing fit (Real Fit Satisfaction [RFS]) using machine learning (ML) algorithms. Specifically, the study aims to: 1. **Apply different machine learning algorithms**: The study employs algorithms such as Random Forest (RF), Support Vector Machine (SVM), and Conditional Tree (CT) to predict consumer satisfaction with clothing fit. 2. **Compare the efficiency of different algorithms**: The study compares the efficiency of these algorithms in predicting RFS to determine which algorithm is best suited for predicting consumer fit satisfaction. 3. **Reduce return rates in online shopping**: By improving consumer satisfaction with virtual fitting (Virtual Fit Satisfaction [VFS]), the study aims to reduce returns due to poor fit, especially in online shopping. ### Background and Motivation - **High return rates**: According to consumer reports, approximately 23% of clothing purchases in the United States are returned due to poor fit, particularly in online shopping. High return rates have negative impacts on both the environment and the economy. - **Technology application**: Many clothing companies are adopting virtual fitting and augmented reality (AR) technologies to reduce return rates and enhance the shopping experience. - **Application of artificial intelligence**: In recent years, the application of artificial intelligence (AI) in the fashion industry has accelerated, particularly in design, manufacturing, retail, and supply chain management. However, the application of AI in clothing fit research is still insufficient. ### Research Methods - **Experimental design**: The study selected skirts of different materials as test garments and used 3D body scanners to create virtual body models of the participants. - **Data collection**: Participants physically tried on the garments and evaluated the fit, while also viewing virtual fitting simulations and evaluating the Virtual Fit Satisfaction (VFS). - **Machine learning models**: RF, SVM, and CT algorithms were used to learn from the data to predict participants' RFS. ### Main Findings - **Algorithm performance**: In the validation set, the RF algorithm had the highest average correlation (0.74), followed by the SVM linear kernel (0.70) and SVM radial kernel (0.72), with the CT algorithm having the lowest correlation (0.55). - **Variable importance analysis**: VFS had the highest importance among the 35 predictive variables. ### Conclusion - **Reducing return rates**: Using ML algorithms to predict consumer fit satisfaction can effectively reduce returns due to poor fit, especially in online shopping. - **Technology integration**: The combination of virtual fitting technology and AI shows great potential in addressing clothing fit issues. This study not only provides new tools and technologies for the fashion industry but also offers scientific evidence for improving consumer shopping experiences and reducing environmental impact.

Predicting consumers’ garment fit satisfactions by using machine learning

Automatic Measurement of Garment Sizes Using Image Recognition

A Machine Learning-Enhanced 3D Reverse Design Approach to Personalized Garments in Pursuit of Sustainability

An evaluation of garment fit to improve customer body fit of fashion design clothing

A deep learning system for predicting size and fit in fashion e-commerce

Clothes Size Prediction from Dressed-Human Silhouettes.

Design of Customized Garments Towards Sustainable Fashion Using 3D Digital Simulation and Machine Learning-Supported Human–Product Interactions

Automated machine learning for fabric quality prediction: a comparative analysis

Applications of Machine Learning in Textile Industry

Cluster Size Intelligence Prediction System for Young Women's Clothing Using 3D Body Scan Data

Machine Learning Approach for Prediction of the Online User Intention for a Product Purchase

Machine learning for digital try-on: Challenges and progress

Efficient Clothing Fashion Prediction using Machine Learning

Evaluating the sales potential of new products using machine learning techniques and data collected from mobile applications

Predicting ready-made garment dressing fit for individuals based on highly reliable examples.

Using Machine Learning Methods for Automation of Size Grid Building and Management

Personalized Clothing Prediction Algorithm Based on Multi-modal Feature Fusion

SmartFit: Smartphone Application for Garment Fit Detection

Deep Learning in Virtual Try-On: A Comprehensive Survey

Developing a Generalized Regression Forecasting Network for the Prediction of Human Body Dimensions

ClothFit: Cloth-Human-Attribute Guided Virtual Try-On Network Using 3D Simulated Dataset