Vendor recommendation engine

Question

I am working on portal where I have PO with details of ordered items and Vendor we are going to deal with.

I am trying to make a item based Vendor recommendation engine for the next quotation so that buyer can have idea before dealing.

Right now I have Vendor, Item and few Parameters to evaluate as mentioned below,

• How much quantity he sent on every order? For example, ordered 100 but delivered only 80

• How is the quality of item? Any damage? For example, 20 damage from ordered 100.

• What is the actual delivery date based on expected delivery date? For example, expecting 26th May but receiving item on 28th May.

• and few others will be added in future.

And all the parameters have some sort of importance and result comes out of 100%.

I came to know about recommendation feature of Apache Mahout.

https://mahout.apache.org/users/recommender/recommender-documentation.html

I have few questions mentioned below :

1. Does Apache Mahout provides recommendation on multiple parameter?
2. How we can deal with this situation? Guidance please.
3. Is Apache Mahout right choice for this kind of situation? Any other option?
4. Is this content based or collaborative filtering?

Answer 1

The page you reference is for the old style mapreduce recommenders that are being deprecated. See this page for newer tech https://mahout.apache.org/users/algorithms/recommender-overview.html

Mahout can create the model for recommenders but you need a whole system of data input processing and serving. We have an end-to-end solution built on Apache PredictionIO that uses the Mahout code above but wraps it in all the other services you will need.

The Universal Recommender here: http://actionml.com/universal-recommender

Answer 2

In my case, formula and complexity is very high.

So, I don't find Mahout as right choice. Then, I have decided to use the PostgreSQL database and application is running fine without any problem on production.