autofuse

This repository is the sample code for the paper

Automatic Fusion Network for Cold-start CVR Prediction with Explicit Multi-Level Representation

Environment

The code has been tested running under Python 3.6.10 and Centos7, with the following packages installed (along with their dependencies):

• tensorflow == 2.1.0
• numpy == 1.19.5
• pandas == 1.1.5
• keras == 2.9.0
• scikit-learn == 0.24.2
• scipy == 1.4.1

Dataset

Dataset Link: https://nijianmo.github.io/amazon/index.html (Or https://jmcauley.uscd.edu/dataset/amazon)

Data Preprocess:

merge the reviews information and meta information similar as G. Zhou, C. Song, X. Zhu, X. Ma, Y. Yan, X. Dai, H. Zhu, J. Jin,H. Li, and K. Gai, “Deep interest network for click-through rate prediction.” https://github.com/zhougr1993/DeepInterestNetwork

Feature Meaning:

• asin: ID of the product
• title: name of the product
• description: description of the product
• price: price in US dollars
• reviewerID: ID of the reviewer
• reviewTime: time of the review
• reviewerName: name of the reviewer
• vote: helpful votes of the review
• style: a disctionary of the product metadata
• reviewText: text of the review
• unixReviewTime: time of the review
• imageURL: url of the high resolution product image
• brand: brand name
• overall: rating of the product

Two special feaures

• popularity: historically cumulative conversions of each product in the dataset (statistical results)
• label: we set “overall” over 3 as conversion behavior, labeled 1, otherwise 0.

Feature Classification:

• user-side feautres: reviewerID, reviewerName
• item-side features: asin, title, description, price, reviewTime, vote, style, reviewText, unixReviewTime, imageURL, brand
• dense features: price
• sparse features: reviewerID, reviewerName, asin, title, description,reviewTime, vote, style, reviewText, unixReviewTime, imageURL, brand
• coarse features(only item-side): price, vote, unixReviewTime, brand, style
• fine featuress(only item-side): asin, reviewText, title, summary, imageURL, description

Running the code

The sample code is built based on Weichen Shen. (2017). DeepCTR: Easy-to-use, Modular and Extendible package of deep-learning based CTR models. https://github.com/shenweichen/deepctr.

Step1:

$ git clone https://github.com/shenweichen/DeepCTR.git 

Step1:

$ mv autofuse.py DeepCTR/deepctr/models

Step3:

$ python3 main.py