AdKDD - Bayesian Time-Varying Coefficients Model
What about the model?
There is a global trend in raising the bar of protecting users data privacy. Matching the high privacy standards while preserving the quality of marketing effectiveness and measurement becomes one of the hot topics in marketing science. My team proposed an adaptive marketing mix models and are presenting in AdKDD 2021. The paper borrows technique and concepts such as kernel regression, spline, state-space models and Bayesian framework to propose called Bayesian Time-Varying Coefficients (BTVC). Part of the reasons we like BTVC is that it can flexibly calibrate with additional insights such as A/B tests and perform pretty well as a time-series model by testing its forecast accuracy and capability in reading time-varying regression coefficients. We also plan to fully implement it under Orbit.
AdKDD
From their own words: “The AdKDD workshops held in conjunction with KDD conference over the past 15 years continue to generate interest from academia and industry–as one of the top venues specifically for advertising research. We believe this is a unique forum for folks interested in aspects of digital advertising to get together, exchange notes and get a pulse for the state of the art, especially in the industry.”
This year AdKDD is hosted in Singapore (Virtual) and is part of the The 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2021. Unfortunately, due to the pandemic, I couldn’t visit Singapore :(
Details of our paper and presentation:
Enjoy Reading This Article?
Here are some more articles you might like to read next: