...
| Excerpt |
|---|
The theory behind how recommendations work in Squirro and the 3 different methods we provide based on composition, conditional probability non-correlated facets, correlated facets or machine _ learning |
Recommendation The recommendation problem could can be formulated in this the following way:
Given input features f1, f2, .., fn (features could be facets of an item or properties of entity), we recommend the top classes C (Class could be a Client or Investor in the Investment Banking App) based on a score(C) = score(f1, f2, ...fn).
We provide 3 methods: composition, conditional probability non-correlated facets, correlated facets and machine _ learning for computing the score(f1, f2, ...fn). With composition and conditional probability methodsthe non-correlated facets and correlated facets methods, the data could be recommended immediately after loading to the storage without any training process (we also call it adhoc query method). For the machine learning method methods, you need to know ml_workflow_id after model is trained.
...
The score of a class C is computed based on average score of each individual feature belong that belongs to it. Score Scores of individual feature is the probability of the that feature f co-occurs with C in a document or entity
...
However in case we have class which does not contain all all the features f1, f2, ..., fn then #E(f1, f2, ..., fn) = 0, this makes score infinitiveinfinite. So given that we have n input feature, and class C contain contains only l feature features (l <=n), the final score is computed as:
...