Hierarchical User Profiling for E-commerce Recommender Systems (WSDM 2020)

HUP:

HUP_code is the code for the paper "Hierarchical User Profiling for E-commerce Recommender Systems", which is published in WSDM 2020. The code is written in Keras 2.2 and theano 1.0.

HUP_Data is the folder that contains the dataset.

Requirements

python==2.7

keras==2.2.0

theano=1.0.1

numpy=1.13.1

pandas=0.20.3

Usage

sh run.sh

Datasets:

JD Micro Behaviors Datasets.

Description:

The datasets are used for research on modeling users' sequential micro-behaviors in "Hierarchical User Profiling for E-commerce Recommender Systems", which is published in WSDM 2020.

The datasets contain users’ micro-behaviors in two product categories “Appliances” and “Computers”, where each line is a sequence of a user’s micro behaviors in a session.

Statistics:

Dataset	JD-Applicances	JD-Computers	Desc
Users	6,166,916	3,191,573	users
Products	169,856	419,388	SKUs
Categories	103	93	Leaf categories
Number of Micro behaviors	176,483,033	88,766,833	micro behaviors

Research Topics:

This dataset can be used for research on User Profiling, Recommender Systems, Session-based Recommender Systems, Micro-behaviors Modeling, and so on. To protect the privacy of users, we remove the users' information. This dataset should only be used for research purpose!

Citation:

Please cite the following paper if you use the data in any way.

@inproceedings{gu2020hierarchical,
  title={Hierarchical User Profiling for E-commerce Recommender Systems},
  author={Gu, Yulong and Ding, Zhuoye and Wang, Shuaiqiang and Yin, Dawei},
  booktitle={Proceedings of the 13th International Conference on Web Search and Data Mining},
  pages={223--231},
  year={2020}
}

Download link:

The full dataset can be downloaded from https://drive.google.com/open?id=1kz7Dkq5jQ8WR82xBITyQ7S4vUgBFNsDn.

File Description:

Each line is a sequence of a user’s micro behaviors in a session.

Each micro-behavior is in the format of "sku + behavior_type + category + time_interval + dwell_time", where "SKU" is id of the product, "behavior_type" is the type of micro-behavior, "category“ is the leaf category of the product, time_interval is the time interval between two consecutive micro behiaviors, dwell_time is the dwell time in each product.

For example, "1993092+7+870+22+27" is a micro-behavior, which means that a user read the specification of product "1993092" (in the leaf category "870"). The time interval between this micro behavior and next micro behavior is 22 seconds. The user spends 27 seconds in this product.

SKU:

"SKU", which is Short for "Stock Keeping Unit", is the alias of product (i.e. item) in E-commerce.

For a SKU "1993092" in the "JD-Applicances" Dataset, the information of the item is in the webpage "https://item.jd.com/1993092.html".

behavior_type:

behavior_type	Micro behaviors	Description
1	Home2Product	Browse the product from the homepage
2	ShopList2Product	Browse the product from the category page
3	Sale2Product	Browse the product from the sale page
4	Cart2Product	Browse the product from the carted page
5	SearchList2Product	Browse the product from the searched results
6	Detail_comments	Read the comments of the product
7	Detail_specification	Read the specification of the product
8	Detail_bottom	Read the bottom of page of the product
9	Cart	Add the product to the shopping cart
10	Order	Make an order

category:

It is the leaf category of a product.

time_interval:

Time intervals is the time between two micro-behaviors.

In the experiments, we discretize them into 5 buckets as following.

time interval bucket	time interval range
1	0~1 seconds
2	1~15 seconds
3	15~40 seconds
4	40~90 seconds
5	90+ seconds

dwell_time:

Dwell time is the length of time a user spends in an item.

It should be noted that multiple micro behaviors under a product share the same dwell time, during which the user may perform multiple micro-behaivors on the product.