def test_vertical_slope(self):
"""Attempt to calculate a vertical slope.
Assert
:class:`movie_recommender.exceptions.VerticalLineOfBestFitGraphError`
is raised.
"""
graph = Graph((Point(1, 1), Point(1, 2)))
with self.assertRaises(exceptions.VerticalLineOfBestFitGraphError):
graph.slope # pylint:disable=pointless-statement
def setUpClass(cls):
"""Define a non-trivial graph."""
# The line of best fit is `y = -1x + 10`.
cls.graph = Graph((
# Define a line 2 above the line of best fit.
Point(-2, 14),
Point(2, 10),
# Define a line 2 below the line of best fit.
Point(-2, 10),
Point(2, 6),
))
def make_genre_predictor(genre, user_id, forbidden_movie=None):
"""Make a predictor for the given genre for the given user.
:param genre: A genre name, as a string.
:param user_id: A user ID. The user for which a predictor is being created.
:param forbidden_movie: A movie ID. A movie to ignore when creating the
predictor.
:return: A function which accepts a movie ID and returns a predicted
rating.
"""
query = """
SELECT movies.genres, ratings.rating
FROM movies JOIN ratings USING (movieId)
WHERE ratings.userId == ?
"""
params = [user_id]
if forbidden_movie:
query += 'AND movieId != ?'
params.append(forbidden_movie)
# Iterate through movies this user has rated. For each movie, create a
# Cartesian point, where X is whether the move has the given genre, and Y
# is the rating this user has given to this movie.
points = []
with common.get_db_conn() as conn:
for row in conn.execute(query, params):
genres = row[0].split('|')
genre_present = 1 if genre in genres else 0
rating = row[1]
points.append(Point(genre_present, rating))
graph = Graph(points)
def predictor(movie_id):
"""Predict a user's rating for the given movie.
:param movie_id: A movie ID.
:return: A predicted rating for the given movie.
"""
genres = read.genres(movie_id)
genre_present = 1 if genre in genres else 0
try:
rating = graph.predict_y(genre_present)
except exceptions.VerticalLineOfBestFitGraphError:
rating = graph.avg_point.y
return clamp_rating(rating)
return predictor
def get_points(input_stream, columns, *, header_rows=0):
"""Consume an input stream, yielding an x,y point for each line.
:param input_stream: A `text I/O`_ stream, where eacah line is CSV data in
"x,y" format.
:param Columns columns: The columns to read from the CSV file.
:param header_rows: The number of header rows in the input file. Header
rows are ignored.
:returns: A generator yielding :class:`movie_recommender.graph.Point`
instances.
:raises: A ``TypeError`` if an input line doesn't have exactly two fields.
.. _text I/O: https://docs.python.org/3/library/io.html#text-i-o
"""
current_line = 0
reader = csv.reader(input_stream)
for row in reader:
current_line += 1
if current_line <= header_rows:
continue
x = float(row[columns.x])
y = float(row[columns.y])
yield Point(x, y)
def test_positive_intercept(self):
"""Verify the method can calculate a positive y-intercept."""
graph = Graph((Point(1, 11), Point(2, 12)))
self.assertEqual(graph.y_intercept, 10)
def test_flat_slope(self):
"""Verify the method can calculate a flat slope."""
graph = Graph((Point(1, 1), Point(2, 1)))
self.assertEqual(graph.slope, 0)
def test_negative_slope(self):
"""Verify the method can calculate a negative slope."""
graph = Graph((Point(1, 2), Point(2, 1)))
self.assertEqual(graph.slope, -1)
def test_positive_slope(self):
"""Verify the method can calculate a positive slope."""
graph = Graph((Point(1, 1), Point(2, 2)))
self.assertEqual(graph.slope, 1)
def test_two_points(self):
"""Assert the method behaves correctly when given two points."""
graph = Graph((Point(10, -1), Point(20, -2)))
self.assertEqual(graph.avg_point, Point(15, -1.5))
def test_two_points(self):
"""Create a graph with two points."""
points = (Point(1.2, 3.4), Point(5.6, 7.8))
graph = Graph(points)
self.assertEqual(graph.points, points)
def test_one_point(self):
"""Create a graph with one point."""
points = (Point(-1, 1), )
graph = Graph(points)
self.assertEqual(graph.points, points)
def setUpClass(cls):
"""Define a trivial graph."""
# y = 0.5x + 1
cls.graph = Graph((Point(0, 1), Point(2, 2)))
def make_year_predictor(user_id, forbidden_movie=None):
"""Make a year-based predictor for the given user.
If year information can't be extracted from a movie's title, then that
movie is skipped when generating a predictor. This is done because so few
movies have this issue. See:
`class:`movie_recommender.exceptions.NoMovieYearError`.
:param user_id: A user ID. The user for which a predictor is being created.
:param forbidden_movie: A movie ID. A movie to ignore when creating the
predictor.
:return: A function which accepts a movie ID and returns a predicted
rating.
"""
query = """
SELECT movies.title, ratings.rating
FROM movies JOIN ratings USING (movieId)
WHERE ratings.userId == ?
"""
params = [user_id]
if forbidden_movie:
query += 'AND movieId != ?'
params.append(forbidden_movie)
# Iterate through movies this user has rated. For each movie, create a
# Cartesian point, where X is the movie's year, and Y is the rating this
# user has given to this movie.
points = []
with common.get_db_conn() as conn:
for row in conn.execute(query, params):
try:
year = read.year(row[0])
except exceptions.NoMovieYearError:
continue
rating = row[1]
points.append(Point(year, rating))
graph = Graph(points)
def predictor(movie_id):
"""Predict a user's rating for the given movie.
:param movie_id: A movie ID.
:return: A predicted rating for the given movie.
:raise movie_recommender.exceptions.NoMovieYearError: If the given
movie's title doesn't include a year, and this predictor makes use
of year data.
:raise movie_recommender.exceptions.EmptyGraphError: If this predictor
can't predict movie ratings at all, due to a lack of relevant data.
For example, this will occur if the given movie's title does
include a year, but all of the movies this user has rated lack a
year.
"""
title = read.title(movie_id)
year = read.year(title)
try:
rating = graph.predict_y(year)
except exceptions.VerticalLineOfBestFitGraphError:
rating = graph.avg_point.y
return clamp_rating(rating)
return predictor
def test_negative_intercept(self):
"""Verify the method can calculate a positive y-intercept."""
graph = Graph((Point(1, -9), Point(2, -8)))
self.assertEqual(graph.y_intercept, -10)
def test_one_point(self):
"""Assert the method behaves correctly when given one point."""
graph = Graph((Point(25, -1.5), ))
self.assertEqual(graph.avg_point, graph.points[0])
def test_zero_intercept(self):
"""Verify the method can calculate a zero y-intercept."""
graph = Graph((Point(1, 0.5), Point(2, 1)))
self.assertEqual(graph.y_intercept, 0)
def setUpClass(cls):
"""Set class-wide variables."""
cls.points = (Point(0.25, 4.4), Point(0.265, 4.62), Point(0.25, 3.84),
Point(0.256, 4.16), Point(0.27, 4.28), Point(0.25, 4.5),
Point(0.269, 4.47))