def generateLinks():
db = ds.MovieLens('ml-latest-small/')
links = db.links
links.drop("imdbId", 1)
return links
def _readDatasetFile(self):
self.log("Opening dataset...")
file = os.path.dirname(os.path.abspath(__file__)) + \
"/" + self.datasetFolder
self.movieDataset = ds.MovieLens(file)
We'll also use the LensKit API to implement our recommender systems algorithms.
***STEP 1***
**Step 1.1**
"""
!pip install lenskit
!pip install -U numba
import lenskit.datasets as ds
import pandas as pd
!git clone https://github.com/crash-course-ai/lab4-recommender-systems.git
data = ds.MovieLens('lab4-recommender-systems/')
print("Successfully installed dataset.")
"""It's important to understand how a dataset is structured and to make sure that the dataset imported correctly. Let's print out a few rows of the rating data.
As you see, MovieLens stores a user's ID number (the first row few rows look like they're all ratings from user 1), the item's ID (in this case each ID is a different movie), the rating the user gave this item, and a time stamp for when the rating was left.
**Step 1.2**
"""
# Top 10 "ratings" of the MovieLens files
rows_to_show = 10
data.ratings.head(rows_to_show)
"""A big aspect of recommender system datasets is how they handle missing data. Recommender systems usually have a LOT of missing data, because most users only rate a few movies and most movies only receive ratings from a few users.
"""
This module defines the data sets that we are going to work with.
"""
import pandas as pd
from lenskit import datasets as ds
ml20m = ds.MovieLens('data/ml-20m')
mlsmall = ds.MovieLens('data/ml-latest-small')
ml100k = ds.ML100K('data/ml-100k')
ml1m = ds.ML1M('data/ml-1m')
ml10m = ds.ML10M('data/ml-10M100K')
if hasattr(ds, 'BookCrossing'):
bx = ds.BookCrossing('data/bx')
def ds_diff(full, subset):
"Return the difference of two data sets."
mask = pd.Series(True, index=full.index)
mask[subset.index] = False
return full[mask]
def load_data(link):
repo = ds.MovieLens(link)
return repo