Code and data for "Detecting Fake Review Buyers Using Network Structure: Direct Evidence from Amazon"

Datasets description

There are five datasets that are needed for replication:

1. product_level_data_text_feats.csv.gz: Product-level data that contains top-1000 TF-IDF features along with product ID and fake review product label (fake = 1 if the product buys fake reviews, 0 otherwise.)
2. product_level_data_with_img_feats.csv.gz: Product-level data that contains metadata, network, and image features along with product ID and fake review product label (fake = 1 if the product buys fake reviews, 0 otherwise.)
3. product_level_data_without_img_feats.csv.gz: Product-level data that contains metadata and network features along with product ID and fake review product label (fake = 1 if the product buys fake reviews, 0 otherwise.)
4. UCSD_product_level_data.csv.gz: Product-level data that contain metadata and network features of products in the home and kitchen category part of the UCSD Amazon Review Data.
5. UCSD_home_and_kitchen_reviews.csv.gz (due to its large size (~ 1GB), this dataset needs to be downloaded using this link): Review-level data that contains home and kitchen reviews of the products used to obtain product-level data in dataset 4 above.

See Table 1 in the paper for a description of the features.

Code

The code is written with Python 3.9.12.

Under the folder "code", there are two scripts:

1. classification.py which will reproduce the supervised results discussed in the paper. It equires the product-level data (datasets 1 and 2) that contain the metadata, image, network, and text features included in the "data" folder. To run the script simply type:

 python classification.py

2. clustering.py which will reproduce the unsupervised results discussed in the paper. It requires the product-level data (datasets 3 and 4) and review-level data (dataset 5) that can be downloaded from here. Tu run the script, simply type:

 python clustering.py

Reference: He, S., Hollenbeck, B., Overgoor, G., Proserpio, D., & Tosyali, A. (2022). Detecting Fake Review Buyers Using Network Structure: Direct Evidence from Amazon. The paper can be downloaded here.