Typical Machine Learning Applications
Getting to grips with big data means letting the machine take the strain using the following machine learning applications.
- Decision Trees
- Cluster Analysis
- Support Vector Machines
- Facebook’s ability to identify target populations
What is a decision tree?
You can use this technique for classification but also for regression analysis. It effectively takes your data and splits it according to criteria that you choose. Once the algorithm is trained it will assign a new example to a category based on the previous sample. So for example if we know that 240 out of 245 previous sign ups returned to the site, then we can predict that the next signee will also return to the site.
Decision trees are used for reporting in a couple of places in Google Analytics.
What is cluster analysis
It’s a means of classifying data. K-means clustering means that you assign a certain number of preselected mean positions in the data environment and the machine builds and existing map of these in phase space and then allocates the incoming population to the appropriate cluster locus. We might for instance be measuring number of purchases against advertisement exposure.
What are support Vector Machines?
Again used for classification – the Support Vector Machine is the boundary between two classes e.g subscribers vs non subscribers.
If we analyse purchasers vs non purchasers based on number of ads seen and time spent on site, we calculate the line that separates the two giving the maximum differences between the line and the point. Then when a new client arrives we can see from their behaviour which side of the line they are going to fall.
How does Facebook help you use big data?
Facebook lets you advertise to either people who have liked your page or post or to a specified group of people or characteristics. If you can supply a list of your existing customers for example or the defining characteristics in terms of interest or facebook behaviour you can advertise to them.
More importantly you can ask facebook to identify a larger population in a given geographical area to match the characteristics of your existing population. Then you can advertise to them also.
Consider downloading our Digital Marketing Papers