SkinSolution | Skincare Recommender System

1. Introduction

Finding the right skincare product can be challenging due to individual skin differences. SkinSolution is a Flask-based web application designed to recommend skincare products based on user preferences. It uses machine learning models trained on FemaleDaily Reviews to analyze reviews, predict ratings, and determine product recommendations.

2. Data Collection & Preprocessing

The dataset comes from FemaleDaily Reviews, where user-generated reviews contain product feedback, skin type, age range, and ratings. Reviews were preprocessed using lexicon-based sentiment analysis to derive an implicit rating, mapping textual sentiment to a numerical score.

3. Model Development

The recommendation system follows a three-step modeling process:

1. Decision Tree Regressors (DTR) predict:
   • (product_id, skin_type, age_range) → explicit rating
   • (product_id, skin_type, age_range) → implicit rating (from sentiment analysis)
2. Support Vector Machine (SVM) then classifies:
   • (explicit rating, implicit rating) → is_recommended

4. System Implementation

The system is built using Flask for the backend, processing user input (age, skin type, skincare category) and returning product recommendations. The frontend is styled with a soft pink theme, designed for a smooth user experience.

5. Results & Performance

• The sentiment-based rating model shows a strong correlation with user-given ratings.
• The DTR model achieves high accuracy in predicting ratings.
• The SVM model effectively classifies recommended products, improving user satisfaction.

6. Conclusion & Future Improvements

SkinSolution successfully provides personalized skincare recommendations using AI models trained on real-world reviews. Future improvements could include deep learning-based NLP for sentiment analysis and a more extensive product database for better recommendations.