Combination of individual and group patterns for time-sensitive purchase recommendation
- PDF / 748,211 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 59 Downloads / 163 Views
REGULAR PAPER
Combination of individual and group patterns for time-sensitive purchase recommendation Anton Lysenko1
· Egor Shikov1 · Klavdiya Bochenina1
Received: 20 January 2020 / Accepted: 22 September 2020 © Springer Nature Switzerland AG 2020
Abstract Due to the availability of large amounts of data, recommender systems have quickly gained popularity in the banking sphere. However, time-sensitive recommender systems, which take into account the temporal behavior and the recurrent activities of users to predict the expected time and category of next purchase, are still an active field of research. Many researchers tend to use population-level features or their low-rank approximations because the client’s purchase history is very sparse with few observations for some time intervals and product categories. But such approaches inevitably lead to a loss of accuracy. In this paper, we present a generative model of client spending based on the temporal point processes framework. The model is built in the way, to bring more individuality for the clients’ purchase behavior which takes into account individual purchase histories of clients. We also tackle the problem of poor statistics for people with a low transactional activity using effective intensity function parameterizations, and several other techniques such as smoothing daily intensity levels and taking into account population-level purchase rates for clients with a small number of transactions. The model is highly interpretable, and its training time scales linearly to millions of transactions and cubically to hundreds of thousands of users. Different temporal-process models were tested, and our model with all the incorporated modifications has shown the best results in terms of both error of time prediction and the accuracy of category prediction. Keywords Point processes · Transactional data · Mixture models · Recommendation · Machine learning
1 Introduction Banks have been using corporate databases for a long time, which led to the accumulation of a large amount of different data on the purchasing behavior of customers. Thanks to this, as well as the development of machine learning algorithms, banks have moved from using simple models, such as LRFM (length, recency, frequency, and monetary) model to more complex recommendation models. Typically, these models A preliminary version of this work was presented at the Most-Rec Workshop at CIKM 2019, but it has not been published in any proceedings or journal before.
B
Anton Lysenko [email protected] Egor Shikov [email protected] Klavdiya Bochenina [email protected]
1
ITMO University, 49 Kronverkskiy prospect, Saint Petersburg, Russian Federation
were used to back up bonus programs developed together with trade and service enterprises for a long fixed period, such as a month. However, the use of time-limited offers can be much more profitable. They may sound as follows: “Hurry up and spend 100 dollars at our partner’s restaurant and get double cash-back. The offer is valid until 10 pm. April the
Data Loading...
