recommendation-engine | Building a very simple recommendation engine | Recommender System library

When you run python generate_recommendations.py this puts the script's directory on the path (sys.path which is searched for modules when importing). When you use from core.rating import Rating in ratingDAO.py then it will search the path for a package called core but since the dao directory is not on the path it cannot be found.

A solution is to use a relative import in the ratingDAO.py module:

Why don't you use _TABLE_SUFFIX?

I simplified the query by deleting unnecessary columns here:

You cannot use table aliases in the where clause.

Instead, just use the expressions:

You already have transactionId on each row. Your query generates rows where each row has transactionId and some array. The BigQuery GUI just formats array elements to separate rows. You probably want to join products directly in from clause, then you obtain one row per product, with transactionId from hit.

Ok, bros. I think there is no option method to filter aggregation significant terms bucket list result using different field.

Based on elasticsearch documentation Significant Terms Aggregation Parameters, which refers to Terms Aggregation Filtering Value. There is no other option than filter using partition expression and filter values with exact values (which i have been using as above, "exclude" param).

So i create other way around by getting the comic ids which i want to exclude and store it as excludeComics variable in array. Then use the excludeComics var in exclude param. And boom, there you go. Filtered significant terms aggregation bucket list result.

Sounds like you need pivot

tl;dr: you cannot use Scala 2.12 because Spark does not support it yet and you also need to use %% when specifying dependencies to avoid problems with incorrect binary versions. Read below for more explanation.

Scala versions like 2.x are binary incompatible, therefore all libraries have to be compiled separately for each such release (2.10, 2.11 and 2.12 being the currently used ones, although 2.10 is on its route to being legacy). That's what _2.12 an _2.11 suffixes are about.

Naturally, you cannot use libraries compiled for a different version of Scala than the one you're currently using. So if you set your scalaVersion to, say, 2.12.1, you cannot use libraries with names suffixed by _2.11. This is why it is possible to write either "groupName" % "artifactName" and "groupName" %% "artifactName": in the latter case, when you use double percent sign, the current Scala binary version will be appended to the name automatically:

I'm the author of this post :)

FM will "simply" fill the missing values in the recommendation matrix. What you could do is batch-predict all movies for all users, sort the results by descending score and store the top 10 results for each user in a cache, why not. That would make it easy to retrieve results in real-time from any kind of app. I suppose you would also retrain periodically to account for new user recos.

Hope this helps.

step-deploy depends on step-train's datas written at file://tmp or hdfs://tmp.