Khush Libaas – Fashion Recommendation Dataset

Project Background
This dataset was developed as part of a university-level coursework project focused on applying machine learning to fashion recommendation systems. The objective is to predict suitable clothing items based on features such as color, brand, size, fabric type, gender, event type, and regional trends.

Dataset Purpose
The dataset facilitates the training, validation, and testing of classification models to recommend appropriate fashion items for various events, such as weddings, office wear, religious gatherings, and more. It can also serve as a benchmarking resource for fashion-related machine learning tasks.

Dataset Structure

• khush_libaas_train.csv: Main training dataset

• khush_libaas_valid.csv: Validation set used during training

• khush_libaas_test.csv: Test dataset for final model evaluation

• khush_libaas_model.pkl: Trained RandomForest model (Scikit-learn)

• feature_importance_chart.png: Feature importance visualization

• confusion_matrix.png: Confusion matrix for event classification

• evaluation_metrics.json: Evaluation metrics (accuracy, precision, recall)

• recommendations.csv: Sample model output recommendations

• README.md: Dataset usage guide and structure overview

Technical Information

• Algorithm: RandomForestClassifier (Scikit-learn)

• Metrics: Accuracy, Precision, Recall, Confusion Matrix

• Language: Python

• Libraries: pandas, scikit-learn, matplotlib, seaborn

Reuse Guidance
This dataset is ideal for students, educators, and researchers working on fashion analytics, recommendation engines, or feature engineering, particularly with imbalanced data.

Acknowledgements
Created and submitted as part of an academic project at TU Wien.