Knowledge Graph For Data Mining

Matrix factorisation based recommendation systems have some limitations:

• The performance of such recommendation systems is inversely proportional to the sparsity of user item matrix.
• Item latency or cold start for newly added items.
• User cold start.

A solution to deal with the shortcomings of Matrix Factorisation based recommendation systems is to use items’ (and users’) metadata and develop algorithms which can exploit this newly added data.

Generally, this data is connected and can be expressed as Knowledge Graphs.