search

macmania / recommendation-engine

1 stars 1 forks source link

Refactor code #3

Open macmania opened 9 years ago

macmania commented 9 years ago

refactor code such that it runs properly with the in file json

macmania commented 9 years ago

@giridhar49 - if we use weka, it's kinda hard to give recommendations to the user using a gui tool. Can you explain to me what the pipeline will look like when the user types in the topics or a few words that they are interested in?

giridhar49 commented 9 years ago

@macmania : This code will not work for recommendation system. This tool will classify documents (not real time).

For recommendation type of projects , we need to build model using collaborative filtering machine learning algorithms. The built model is used to compute recommendations for the users. Results are stored in Persistent back end store like Relational DB, Redis,NOSQL. Once you have required recommendations stored in back end, web application can be built on back end store to provide recommendations. Process is called batch processing which is not realtime.

However for real time recommendations, there are machine learning libraries like Trident ML for Apache Storm and MLLib for Apache Spark based systems.

Let me know if any concerns :)

macmania commented 9 years ago

@giridhar49, will check that out. As per our conversation, I will be focusing on improving the granularity of the mulit-classifier to improve accuracy.