Tabular

The collaborative filtering problem we will be doing is one we have done before in lesson 8. But, what if try doing it using a tabular model. Note: We learned tabular modeling in lesson 9!

from fastai.collab import *
from fastai.tabular.all import *
from sklearn.ensemble import RandomForestRegressor

Data

Lets grab the data, like before, and view it

path = untar_data(URLs.ML_100k)

ratings = pd.read_csv(path/'u.data', delimiter='\t', header=None,
                      names=['user','movie','rating','timestamp'])

ratings.head()
user movie rating timestamp
0 196 242 3 881250949
1 186 302 3 891717742
2 22 377 1 878887116
3 244 51 2 880606923
4 166 346 1 886397596

Grabbing data

Let's initialize some variables we need to create our tabular object.

user_name = ratings.columns[0] #user col
item_name = ratings.columns[1] #movie col
rating_name = ratings.columns[2] #label (rating col)

cat_names = [user_name,item_name] #category col
splits = RandomSplitter()(range_of(ratings)) #split
procs = [Categorify, FillMissing, Normalize]

Tabular Object

Lets now take everything above and create a tabular object.

to = TabularCollab(ratings, procs, cat_names, y_names=[rating_name], y_block=TransformBlock(), splits=splits, reduce_memory=False)
dls = to.dataloaders()

dls.show_batch()
user movie rating
0 614 476 3
1 582 676 2
2 792 124 4
3 733 762 4
4 344 815 2
5 36 269 3
6 110 779 3
7 752 302 5
8 409 214 4
9 916 50 5

Training

Now let's find the best lr and train our model.

learn = tabular_learner(dls, y_range=(0,5.5), layers=[500,250],
                        n_out=1, loss_func=MSELossFlat())

learn.lr_find() #find best lr

SuggestedLRs(lr_min=0.00831763744354248, lr_steep=0.0008317637839354575)
learn.fit_one_cycle(4, 1e-3, wd=.01)
epoch train_loss valid_loss time
0 0.973944 0.977132 00:12
1 0.875915 0.895781 00:12
2 0.807531 0.848947 00:12
3 0.727740 0.856915 00:12

If you compare these results with lesson 8, you will notice they are similer.

Random Forest

We can also create a random forest, which we also learned in lesson 9!

def r_mse(pred,y): return round(math.sqrt(((pred-y)**2).mean()), 6)
def m_rmse(m, xs, y): return r_mse(m.predict(xs), y)

Method below creates our random forest and fits it

def rf(xs, y, n_estimators=40, max_samples=80000, max_features=0.5, min_samples_leaf=5, **kwargs):
    return RandomForestRegressor(n_jobs=-1, n_estimators=n_estimators,
                                max_samples=max_samples, max_features=max_features,
                                min_samples_leaf=min_samples_leaf, oob_score=True).fit(xs, y)

Tabular object

As we did above, lets create our tabular object.

procs = [Categorify, FillMissing]

to = TabularCollab(ratings, procs, cat_names, y_names=[rating_name], y_block=TransformBlock(), splits=splits, reduce_memory=False)

xs,y = to.train.xs, to.train.y 
valid_xs,valid_y = to.valid.xs, to.valid.y

m = rf(xs, y) #Fitting

m_rmse(m, xs, y), m_rmse(m, valid_xs, valid_y)  

(0.848744, 1.004573)