preprocess.py 文件源码

def remove_rare_elements(data, min_user_activity, min_item_popularity):
    '''Removes user and items that appears in too few interactions.
    min_user_activity is the minimum number of interaction that a user should have.
    min_item_popularity is the minimum number of interaction that an item should have.
    NB: the constraint on item might not be strictly satisfied because rare users and items are removed in alternance, 
    and the last removal of inactive users might create new rare items.
    '''

    print('Remove inactive users and rare items...')

    #Remove inactive users a first time
    user_activity = data.groupby('u').size()
    data = data[np.in1d(data.u, user_activity[user_activity >= min_user_activity].index)]
    #Remove unpopular items
    item_popularity = data.groupby('i').size()
    data = data[np.in1d(data.i, item_popularity[item_popularity >= min_item_popularity].index)]
    #Remove users that might have passed below the activity threshold due to the removal of rare items
    user_activity = data.groupby('u').size()
    data = data[np.in1d(data.u, user_activity[user_activity >= min_user_activity].index)]

    return data