My model from "river.reco" does not find 'user' and 'item' arguments in the MovieLens dataset

[TxusLopez]

Hi,

I am trying to follow the example of "Matrix factorization for recommender systems", but I get this error in the Part 1 for the Baseline model:

TypeError: predict_one() missing 2 required positional arguments: 'user' and 'item'

Here the code:

import json
import itertools
import pandas as pd

from river import datasets
from river import metrics
from river import evaluate
from river import dummy
from river import stats
from river import preprocessing
from river import optim
from river import reco
from river import compose
from river import facto
from sklearn.preprocessing import StandardScaler


#MovieLens
for x, y in datasets.MovieLens100K():
    print(f'x = {json.dumps(x, indent=4)}')
    print(f'y = {y}')
    break

X_y = datasets.MovieLens100K(unpack_user_and_item=False)

############################ Dummy #########################################

dmmy_metrics = metrics.MAE() + metrics.RMSE()
#MovieLens
dummy_model = dummy.StatisticRegressor(stats.Mean())
dummy_steps = evaluate.iter_progressive_val_score(dataset=X_y, model=dummy_model, metric=dmmy_metrics, step=1)

for dstep in itertools.islice(dummy_steps, 100, 105):
   print(dstep)

############################ Baseline #########################################

baseline_metrics = metrics.MAE() + metrics.RMSE()
baseline_params = {
    'optimizer': optim.SGD(0.025),
    'l2': 0.,
    'initializer': optim.initializers.Zeros()
}

#MovieLens
baseline_model = preprocessing.PredClipper(
    regressor=reco.Baseline(**baseline_params),
    y_min=1,
    y_max=5
)

baseline_steps = evaluate.iter_progressive_val_score(dataset=X_y, model=baseline_model, metric=baseline_metrics, step=1)

for bstep in itertools.islice(baseline_steps, 100, 105):
   print(bstep)

Thanks in advance!

[gbolmier]

Hey @TxusLopez, according to the docs river.reco models learn_one and predict_one methods take a user and an item inputs.

Try loading your dataset this way in your code before calling evaluate.iter_progressive_val_score and it should work:

X_y = datasets.MovieLens100K(unpack_user_and_item=True)

[TxusLopez]

Thanks @gbolmier for your response. I get the same error with that. But I was trying to do the same without using the evaluate.iter_progressive_val_score by doing the process manually with a simple iteration and detailing the params x,y,user,item:

#Training
for index, row in X_train.iterrows():
    user = row['user']
    item = row['item']
    feats = row.to_dict()
    targ = y_train[index]

    baseline_model = baseline_model.learn_one(x=feats, y=targ, user=user, item=item)

, being X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=42) and

X = df_movielens.drop('rating', axis=1)  
y = df_movielens['rating']       

, where df_movielens is just a DataFrame of the MovieLens dataset. And then I got a code without that error.

[gbolmier]

Works on my machine though 😉