Silei Huo
Data-driven problem solver with extensive experience in Banking and Technology (Payments), focusing on Strategic Planning, Product Management, and Data Analytics. MBA from London Business School.
Data Science | FinTech | Payment Processing | Banking
View My LinkedIn Profile
Children’s Book Recommendation - Capstone Project
Project Background
How to support non-native speaker parents to find the most suitable books for their children? How to help them search for books in a much easier and simpler way?
Recommendation Mini-APP
Children’s Book Recommendation - Deployed on Streamlit
Dataset
Dataset source: USCD Book Graph
Original Children’s book data szie: 10M ratings, 128K books, 830K authors
Processed data size (dataset is reduced due to limited resources): 1M reviews, ~40K unique books, ~38K unique users
Key features: Book Title, Book Description, Book Ratings
Text Preprocessing
NLP(Book Description): Conducted word vectorization using TFIDF & NLTK WordNetLemmatizer on each book’s description text data
Recommendation Systems
- Content Based Recommendation: Analyse similarity between each book pair based on book description text, recommend top ranked books for unknown users;
- User-Item Collaborative Recommendation: Analyse similarity between each book pair based on user rating record, recommend top ranked books for unknown users;
- FunkSVD Recommendation: Predict rating for each book that user haven’t read before, and recommend top ranked books to specific known user;