Design and implement a

Machine Learning Model to predict the spending of ZaloPay customers in a period of time

Nguyen Thi Thuy Tien
Nguyen Quoc Cuong
Bui Thanh Long
Nguyen Hoang Khang

With the mass adoption of machine learning and big data tools, companies now have the ability to predict crucial information about their customer base, especially their purchasing power. The capabilities of predicting customer spending over a period of time is a crucial task for marketers in making strategic decisions about advertising. 

That is why in this project, with the collaboration with ZaloPay, the team created a time-series machine learning system that can forecast the expenditure of the customer at a specific time. The model’s goal is to tell each customer how much they are about to spend on the ZaloPay service during a particular time. Using ZaloPay data, the team analyzes 40 million transactions, extracting them to lag-feature for time-series prediction. The final model was 859,330 VND in RMSE and 274,024 VND in MAE. For ease of use, the model is served through a dashboard for non-technical users and more advanced filtering.

By forecasting the customer spending, the marketers can easily target a selected group of customers based on the amount that they are willing to spend in a certain time by giving vouchers or increasing the products at that moment. Thus, the investment is put in the right place, which saves the costs for the business.