評価の別れた映画を抽出した

Author: ksomemo

ch02/movies_evalution.py

@@ -0,0 +1,53 @@
+# -*- coding: utf-8 -*-
+
+import pandas as pd
+import study
+
+def main():
+    metadata = {
+        'users': {
+            'names': ['user_id', 'gender', 'age', 'occupation', 'zip'],
+            'dat': 'users.dat'
+        },
+        'ratings': {
+            'names': ['user_id', 'movie_id', 'rating', 'timestamp'],
+            'dat': 'ratings.dat'
+        },
+        'movies': {
+            'names': ['movie_id', 'title', 'genre'],
+            'dat': 'movies.dat'
+        }
+    }
+
+    # data loading
+    tables = {}
+    for k, v in metadata.items():
+       tables[k] = read_table(study.DATA_DIR + 'ch02/movielens/' + v['dat'], v['names'])
+    # join(ratings.user_id = users.user_id, ratings.movie_id = movies.movie_id)
+    data = pd.merge(pd.merge(tables['ratings'], tables['users']), tables['movies'])
+
+    # u'タイトル別評価件数のうち、件数が上位である映画に対する女性の平均評価
+    mean_ratings = data.pivot_table(
+        'rating', rows='title', cols='gender', aggfunc='mean'
+    )
+    ratings_by_title = data.groupby('title').size()
+    active_titles = ratings_by_title.index[ratings_by_title >= 250]
+    mean_ratings = mean_ratings.ix[active_titles]
+    top_female_ratings = mean_ratings.sort_index(by='F', ascending=False)
+    print top_female_ratings['F'][:10]
+
+    # calculate for each row
+    mean_ratings['diff'] = mean_ratings['M'] - mean_ratings['F']
+    sorted_by_diff = mean_ratings.sort_index(by='diff')
+    print sorted_by_diff[:15]
+
+    # u'評価の別れた映画TOP10
+    ratings_std_by_title = data.groupby('title')['rating'].std()
+    ratings_std_by_title = ratings_std_by_title.ix[active_titles]
+    print ratings_std_by_title.order(ascending=False)[:10]
+
+def read_table(file_path, names, sep='::', header=None):
+    return pd.read_table(file_path, sep=sep, header=header, names=names)
+
+if __name__ == '__main__':
+    print main()