Optimizing a digital marketing campaign is tremendous complex since lots of variables are involved, like channels, strategy, content quality, investment, makets, social networks etc.
My question is: can we really optimize this process using machine learning approaches to the gigabytes of data we can now collect from analytics? Any reference on this topics? I look at some major journals without much sucess.
The problem is very challenging, from different perspectives. First, do the data arrive in form of (multiple) streams, rather than being available in advance for training? Then, is time frequency updates very different among the variables? Furthermore, are data noisy and/or incomplete? Finally, is the decision process distributed, rather than centralized, and is it based on some other decisions and/or incomplete knowledge?
A suitable solution to this problem would require the analysis and planning of the data to be analyzed, of distributed software architecture for data management, and then of the machine learning and data mining algorithms.
There is a nice paper http://www.db-net.aueb.gr/matina/docs/dedm2013_thomaidou.pdf that deals with Adwords optimization using Genetic Algorithms.
But I took a different approach using an iterative algorithm by re-estimating the CTR and CR ratios over several channels (Adwords, Facebook, Email and SEO). To deal with high dimensionality I used a semi-supervised Isomap to cluster data. I'm testing on a small data set, but will be glad to tested over other data
- Badges
- Science topic