How to Design Retail Recommendation Engines with Neo4j

[@tobiasjakob4338]

Hi, I have a question regarding the difference between a neo4j recommendation engine and traditional solutions: Normally big recommendation architectures work with offline computation of models and offline usage of these to compute recommendations that then can be requested online / in realtime. (At least this is how I understand traditional approaches) So how does this work with neo4j. Is everything online when working with the graph approach or will there be a offline calculation as well. I am really interested in understanding the differences and its effect on the architectural part. It would be AWSOME if someone could give me some clarification and background-information. Thank you!

[@sapphirecloud1705]

Thank you, found exactly what I was looking for. Please suggest me sources to study this further.

[@B1TCH35K1LL3R]

Hi! great video! i have a question, how is it that the MERGE statements perform its duty on the individual rows, since the query is "returning" all rows at the sql database? it is not very clear for me... no need for any kind of loop

[@lyonwj]

Hi Farid - the MERGE statements are executed for each row yielded from the call to apoc.load.jdbc. In general, Cypher statements execute for each path (row) built up in the traversal at that point. More info here: neo4j.com/developer/guide-importing-data-and-etl/

[@satyamvijayvargiya8616]

can anybody give me the link of the mysql database which is use in this video , i want to practice the use case

[@lyonwj]

Hi Satyam - I don't have the sql database, but the first tab of this spreadsheet is the data used in the "products" table. So if you load that into mysql you can use the same Cypher statement to load the data into Neo4j using apoc.load.jdbc docs.google.com/spreadsheets/d/1AL4uijztdNowNitO7H1aPJO1ZTxgpujyi7acRAA69FE/edit?usp=sharing

[@satyamvijayvargiya8616]

yeah got your point thanks for replying