I'm trying to build a travel itinerary recommender system. I'm using ML.NET Builder
Scenario Recommendation
Data The data used is .csv file training.csv
Data Column to predict (Rating): RatingRounded User column: UserId Item column: ItemId Time to train: 10
Train But each time I start the training, the Best quality (RSquared) is always 0.0000 Train
Evaluate RSquared: 0.0000 Model: Matrix Factorization
when I predict my model, it always return Nan Predicted Rating
Did i miss something? why does it keep returning me NaN, I have about 2k+ data on my csv
I tried searching for issue, it says that my dataset is not enough. But I have around 2k+ data is it really not enough or am I missing something? Should not use the ML.NET Model Builder?
The problem may lay with your variable types. Your userID and itemID are long alphanumerics. If they represent distinct travel itineraries and users, then select 'categorical' for the data type.
If it is assuming they are numerical or text variables, then it will be trying to fit a regression to what's contained within in each itemID, when really the special sequence of letters and numbers in each ID is not the meaningful bit - it's just an ID (a label for a category).
The RSquared of 0 means at the moment it is finding no relationship between the contents of the ID's and rating.