As a practitioner, in my lens any of the analytical problem will be solved by using the either of these approaches
- Prediction – Predicting the future
- Grouping
- Association
- Exploring the data in either higher dimensional Space (kernels) or lower dimensional Space
And we do have different algorithms because of the variable type of what we are predicting, in case of continuous prediction we have regression algorithms, in case of discrete or nominal variables we have classification algorithms and if you want to predict the future by using past variables we have time series algorithms.
Whereas when it comes to grouping, in case you are interested to group the variables we have data reduction techniques and if you want to group the observations we have cluster algorithms.
Similarly for the Associations we have Market Basket Analysis or recommendation systems which are based on either content as content based recommendation systems or if you want to provide recommendations based on finding similar people we have collaborative filtering.
All these basically learning based approaches and we do have different ways of finding these relations using different approaches the below are different approaches and models that exists to solve the analytical problem which might fall under prediction, grouping and association categories.
As there is “No Free Lunch” theorem states that there is no one model that works best for every problem. The assumptions of a great model for one problem may not hold for another problem, so it is common in machine learning to try multiple models and find one that works best for a particular problem. That’s the reason we practitioner always build the pipeline of algorithms and select the best model on the basis of cross validation in both external(domain) and internal(data) prospective
