Product size recommendation and fit prediction are crucial for improving customer shopping experiences and reducing product returns. Modeling customers' fit feedback is challenging to its nuanced semantics, which result from subjective product evaluation and imbalanced label distribution. We explored various prediction frameworks to address the product fit problem in this project, which captures the semantics behind consumers' fit feedback and applies metrics to handle label imbalance concerns.
Please refer to project_report.pdf for details.
1 DATASET We are using the RentTheRunway clothing fit dataset (https://cseweb.ucsd.edu//~jmcauley/datasets.html#clothing_fit) for this assignment.
2 Models
Multiclass Logistic Regression, K Nearest Neighbors, Naive Bayes, Random Forest, XGBoost, Deep Learning With Pre-Trained Word Embeddings (Recommended Model).