Books to Scrape Price Predictor

Model Characteristics

• Type: Regression (RandomForestRegressor, 300 trees)
• Inputs: star rating (1-5), title length, title word count, presence of ':', + 8 title TF-IDF topics via TruncatedSVD
• Output: predicted price in GBP
• CV MAE: 13.3 GBP (+/- 0.33) on 5-fold cross-validation

Data Overview

• Dataset: books-to-scrape-catalog-dataset (1000 rows)
• Raw HTML-wrapped prices cleaned via regex; word ratings (One-Five) mapped to integers

Evaluation Results

• 5-fold CV Mean Absolute Error: 13.3 GBP
• Baseline (predict median): 12.5 GBP
• Title-only features give weak signal; adding category/genre would improve accuracy

Files

• model.joblib — trained RandomForestRegressor
• tfidf.joblib — fitted TfidfVectorizer
• svd.joblib — fitted TruncatedSVD
• numeric_feats.joblib — list of numeric feature names
• feature_names.npy — full feature names

Usage

Provenance