Icon

Workflow - Book Recommendation1.knwf

Combining Books RatingsData & Books Metadata Making Predictions for new products.We are getting the error 'Unsupported missing values in predictioncolumn' in the 'Spark Numeric Scorer' node because theRecommendation System model could not make predictions correspondingto product ids (book_id) present in the Test data as these ids were notpresent in the training data. The model predicts them as null values (whichare basically empty predictions shown in the format NaN).In such a situation when a new product is launched, we can recommendthem to users by using Content based Recommeder system models. Handling the missing valuesIn order to calculate the performance scores we willuse the 'Spark Missing Value' node to remove rowscorresponding to the Predictions having null values ormissing values. Observing the recommedations for a particular UserWe can select a particular user id and observe the predictedratings correponding to different books. We can arrange output in Descending order of ratings and seewhich are the top book recommendeds for the user. Model Building Node 6Node 7Node 9Creating Big Data ContextNode 14Node 15Node 16Removing Null PredictionsNode 22Books Ratings DataNode 24Filtering out 'id' column as it is same as 'book_id'Books DataNode 27 Table to Spark Spark Partitioning Spark Predictor(MLlib) Create Local BigData Environment Spark Row Filter Spark NumericScorer Spark CollaborativeFiltering Learner (MLlib) Spark Missing Value Spark NumericScorer CSV Reader Joiner Column Filter CSV Reader Math Formula(Multi Column) Combining Books RatingsData & Books Metadata Making Predictions for new products.We are getting the error 'Unsupported missing values in predictioncolumn' in the 'Spark Numeric Scorer' node because theRecommendation System model could not make predictions correspondingto product ids (book_id) present in the Test data as these ids were notpresent in the training data. The model predicts them as null values (whichare basically empty predictions shown in the format NaN).In such a situation when a new product is launched, we can recommendthem to users by using Content based Recommeder system models. Handling the missing valuesIn order to calculate the performance scores we willuse the 'Spark Missing Value' node to remove rowscorresponding to the Predictions having null values ormissing values. Observing the recommedations for a particular UserWe can select a particular user id and observe the predictedratings correponding to different books. We can arrange output in Descending order of ratings and seewhich are the top book recommendeds for the user. Model Building Node 6Node 7Node 9Creating Big Data ContextNode 14Node 15Node 16Removing Null PredictionsNode 22Books Ratings DataNode 24Filtering out 'id' column as it is same as 'book_id'Books DataNode 27 Table to Spark Spark Partitioning Spark Predictor(MLlib) Create Local BigData Environment Spark Row Filter Spark NumericScorer Spark CollaborativeFiltering Learner (MLlib) Spark Missing Value Spark NumericScorer CSV Reader Joiner Column Filter CSV Reader Math Formula(Multi Column)

Nodes

Extensions

Links