Comparing state of the art machine learning frameworks for sequence based recommendations
A comparative thesis project using click-streamdata in a children’s fashion setting
About Babyshop Group
E-commerce is at the heart of Babyshop Group where premium childrens’ fashion is retailed on a multitude of websites (eg. lekmer.se, babyshop.se, alexandalexa.com, melijoe.com). Babyshop Group has grown rapidly and we sell and ship all over the world with a revenue of more than 1 billion SEK. In order to sustain such growth, Babyshop Group will continue to work in a data driven environment, which you hopefully can be a part of! We work with machine learning (ML) on a daily basis, whether it be exploratory data analysis and modelling in a Jupyter Notebook or building ML-pipelines integrating cloud solutions for personalization, demand prediction, recommender systems, learning-to-rank and more. Our head office is located by Karlaplan where we regularly meet for fika and fun after-works.
Thesis proposal
On a daily basis at Babyshop Group, there are hundreds of thousands of users browsing the tens of thousands of articles in our assortment. With such an extensive assortment, finding relevant and inspiring articles becomes far from easy, and the need for recommender systems is high. In ecommerce, providing relevant recommendations is critical for an engaging and satisfactory user experience. Many different methods for creating personalized recommendations exist, e.g., collaborative filtering, content-based filtering and session-based filtering to name but a few. Within the latter, a subset focuses on what is referred to as sequence based recommendations, where the users’ historical behaviour on the site is used to generate recommendations that they would like to interact with in the future.
Previous work include K-Nearest Neighbours, Classic Recurrent Neural Networks, Long-Short Term Memory, Convolutional Sequence Embedding Recommendation Model (Caser), Graph Neural Network and Bidirectional Encoder Representations from Transformers for sequential Recommendation (BERT4Rec). Within the fashion domain, previous work have primarily focused on adult’s fashion, for instance Zalando’s session-based complementary item recommendation algorithm, while lacking coverage of the sub-domain of children’s fashion. Because of this discrepancy, and vast amounts of click-stream data collected by Babyshop Group, it becomes interesting to evaluate how these different approaches perform in a different, but comparable, domain. This thesis project entails a comparative review of these approaches, where the exact choice of methodologies to implement and compare, as well as the overall scope of the review would be in the hands of the student. For inspiration, most papers from the yearly RecSys conference can be found here.
For this thesis, we believe it to be preferable if you have the following qualifications:
● Student of Computer Science, Machine Learning, Computer Engineering, Mathematics, Physics, or related field (e.g. Applied Mathematics/Statistics).
● A good understanding of the use of some of the current state of the art machine learning frameworks such as Keras, TensorFlow, Scikit-Learn, Spark etc. in developing ML applications.
● Proficiency in Python and Git.
● Knowledge about data wrangling and data munging, using SQL, Pandas and Numpy.
● Good communication skills in written and spoken English.
What we offer
● Supervision from experienced data scientists and ML practitioners.
● The compute-resources needed for such a modelling-heavy task.
● 30 000 SEK on completion of the thesis.
● Inspiring culture and fun after-works!
What we unfortunately do not offer
● Relocation package
● Swedish Visa assistance
Starting, early January
Contact: Marcus Svensson, Data Scientist at Babyshop Group marcus.svensson@babyshop.se
Last application date is 17/12, but continuous selection is carried out.
