README.md

metadata

license: mit
task_categories:
  - text-classification
  - text-retrieval
language:
  - en
tags:
  - recommendation-system
  - collaborative-filtering
  - amazon-reviews
  - musical-instruments
size_categories:
  - 100K<n<1M
configs:
  - config_name: interactions
    data_files:
      - split: train
        path: interactions/train-*
  - config_name: items
    data_files:
      - split: train
        path: items/train-*
  - config_name: cold_start
    data_files:
      - split: train
        path: cold_start/train-*

Amazon Musical Instruments — 5-Core Sampled Dataset (2018-2023)

Sampled subset of Amazon Reviews 2023 Musical Instruments category, created for a Data Mining course recommendation system project.

Sampling Strategy

1. Time filter: 2018-01-01 to 2023-09-12
2. Iterative 5-core filtering: Both users and items have at least 5 interactions (iteratively until stable)
3. Cold-start test set: 5000 users with 1-2 reviews held out for cold-start evaluation
4. Random seed: 42

Configs

This dataset has three configs. Load each one separately:

from datasets import load_dataset

interactions = load_dataset("oyku-tugana/amazon-musical-instruments-2018-2023-5core", "interactions", split="train")
items        = load_dataset("oyku-tugana/amazon-musical-instruments-2018-2023-5core", "items",        split="train")
cold_start   = load_dataset("oyku-tugana/amazon-musical-instruments-2018-2023-5core", "cold_start",   split="train")

Statistics

Field	Value
Users (main)	25,746
Items (main)	13,575
Interactions (main)	222,276
Sparsity	0.999364
Mean rating	4.451
Items w/ image	100.0%
Items w/ description	61.6%
Items w/ price	77.8%
Cold-start users	2,991

Schema

interactions / cold_start

• user_id, parent_asin, rating, timestamp (ms), date
• helpful_vote, verified_purchase
• review_title, review_text

items

• parent_asin, title, description, features
• main_category, categories, store
• price (nullable), average_rating, rating_number
• image_url (nullable), has_image, has_description, has_price