Start with linear regression for your vectors and then also try random forests and other algorithms and compare what works better in your case (the trick to evaluate the algorithm's quality is simple: you take all your the data you have, use 70-80% of it to train, and then use the remainder to get estimations from the trained engine – and then use some function to calculate the deviation error, usually people use mean square error approach).

- Stanford online course, with online book and lectures published on Youtube http://mmds.org/.

Includes various modern approaches to building **recommendation engines** are described very well in it.

Popular python machine learning algorithms can be used with pl/python to create fast inbuilt machine learning algorithms e.g.

- http://www.madlibs.com/ - Apache MADlib: Big Data Machine Learning in SQL

so in PL/Python we can

$$ from sklearn import linear_model

- https://www.youtube.com/watch?v=_f3URz9RlCY - Scalable in-database machine learning with PL/Python