Time
2 weeks
Location
USA
Sector
Marketing
Description
The customer’s objective was to target various Facebook users with ads and achieve maximum conversion rates, taking into account factors such as the day of the week, time, and other target features. The goal was to build and train a machine learning model capable of predicting conversion rates for a given click, represented by a set of features. This would enable the customer to select the optimal audience and ad prices in advance.
Solution
The project involved improving the accuracy of the base model through model selection among different types of models and model ensembles, including random forest and logistic regression. Various feature engineering and selection strategies were applied to fine-tune accuracy scoring.
Results
The developed model allowed for the optimization of advertising campaigns in 97% of cases. This system significantly increased advertising revenue without the need to increase traffic and placement costs.