What are some of the applications of machine learning in online advertising analytics?

There are many challenging data science problems to be found in the online advertising world. It goes much further then purely analytics. Over the past years, online advertising has become more and more about learning and making decisions based on data. Especially since we are often dealing with vast amounts of data (‘big data’),machine learning can be a great help in automated decision making and advanced analytics.

Simple analytics applications:
– Forecasting campaign performance in terms of media spend and impressions
– Optimization of campaigns
– Automated bidding strategies for ads
– Determining impact of creatives
– Predicting email responsiveness

Also there are still quite a bit of unsolved challenges in the online advertising domain. For instance, many of which will most likely be solved using machine learning approaches.
– Segmentation of audiences; audiences often consist of multiple segmented groups, in addition ,finding the segmentations could help in gaining a better perspective on an audience
– Look-alike modelling; targeting specific groups or prospecting on domains may prove favorable on conversions. Most important the key characteristics leading to these conversions could be a far more effective to target.
– Detecting fraudulent behavior; a certain amount of budget in online advertisement the budgets are spend on fraudulant clicking or conversions. Detecting fraudulant behavior would reduce wastes of budget.

All of these open problems emphasize reduced costs (e.g. of buying inventory). better return on ad spend, increased reach and/or finding behavioral patterns.

If you’d like to try out some of the data challenges. have a look at earlier and current Kaggle competitions related to the online advertising domain. Often the winning solutions are open source which could spark your ideas.