DynamicCF/prepare_data.py at master · AlbMLpy/DynamicCF · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
import json

import pandas as pd
import numpy as np
import matplotlib as mpl

from source.plotting import PARAMS, get_fig_ax, set_ax, add_plot_ax

from source.data_preparation import (
    USER_ID,
    ITEM_ID,
    DATE_DAYS,
    RELEVANCE_COLUMN,
    TIMESTAMP,
    print_recsys_df_stats,
)
from source.experiments.prepare_data import prepare_data_for_experiment
from source.experiments.hyperparameters_search import import_config
from load_data import (
    RESULTS_DIR,
    DATA_DIR_AMZ_B,
    DATA_DIR_AMZ_G,
    DATA_DIR_ML20M,
    DATA_DIR_STEAM,
    DIR_ML20M_NAME,
    DIR_AMZ_B_NAME,
    DIR_AMZ_G_NAME,
    DIR_STEAM_NAME,
)

def encode_column(data: pd.DataFrame, column: str) -> pd.DataFrame:
    return (
        data
        .assign(_temp=pd.Categorical(data[column]).codes)
        .drop([column], axis=1)
        .rename({'_temp': column}, axis=1)
    )

def prepare_ml20m():
    ratio = 0.2
    data = (
        pd.read_csv(DATA_DIR_ML20M / 'ratings.csv')
        .rename({'userId': USER_ID, 'movieId': ITEM_ID}, axis=1)
        [[USER_ID, ITEM_ID, TIMESTAMP]]
    )
    data = data.sort_values(by=TIMESTAMP)
    data = data.tail(int(data.shape[0] * ratio)).reset_index(drop=True)
    data[TIMESTAMP] = pd.to_datetime(data[TIMESTAMP], unit='s', utc=True)
    data['day'] = data[TIMESTAMP].round('d')

    data[RELEVANCE_COLUMN] = np.ones(len(data))

    print_recsys_df_stats(data, DATA_DIR_ML20M / 'data_stats.csv')
    data.to_csv(DATA_DIR_ML20M / 'prepared_data.csv')

def prepare_amazon(file_dir, file_name):
    data = (
        pd.read_csv(file_dir / file_name)
        .rename(
            {'reviewerID': USER_ID, 'asin': ITEM_ID, 'unixReviewTime': TIMESTAMP},
            axis=1
        )
        [[USER_ID, ITEM_ID, TIMESTAMP]]
    )
    data = data.sort_values(by=TIMESTAMP)
    data = data.reset_index(drop=True)
    data[TIMESTAMP] = pd.to_datetime(data[TIMESTAMP], unit='s', utc=True)
    data['day'] = data[TIMESTAMP].round('d')

    data[RELEVANCE_COLUMN] = np.ones(len(data))
    data = encode_column(data, USER_ID)
    data = encode_column(data, ITEM_ID)

    print_recsys_df_stats(data, file_dir / 'data_stats.csv')
    data.to_csv(file_dir / 'prepared_data.csv')

def prepare_steam():
    data = (
        pd.read_csv(DATA_DIR_STEAM / 'steam.csv')
        .rename(
            {'username': USER_ID, 'product_id': ITEM_ID},
            axis=1
        )
        [[USER_ID, ITEM_ID, TIMESTAMP]]
    )
    data = data.sort_values(by=TIMESTAMP)
    data = data.reset_index(drop=True)
    data[TIMESTAMP] = pd.to_datetime(data[TIMESTAMP], unit='s', utc=True)
    data['day'] = data[TIMESTAMP].round('d')

    data[RELEVANCE_COLUMN] = np.ones(len(data))
    data = encode_column(data, USER_ID)
    data = encode_column(data, ITEM_ID)

    print_recsys_df_stats(data, DATA_DIR_STEAM / 'data_stats.csv')
    data.to_csv(DATA_DIR_STEAM / 'prepared_data.csv')

def prepare_data():
    prepare_ml20m()
    prepare_amazon(DATA_DIR_AMZ_B, 'amz_b.gz')
    prepare_amazon(DATA_DIR_AMZ_G, 'amz_g.gz')
    prepare_steam()

def save_data_dynamics():
    for dataset in [DIR_ML20M_NAME, DIR_AMZ_B_NAME, DIR_AMZ_G_NAME, DIR_STEAM_NAME]:
        conf = import_config(dataset)
        _, _, left_data = prepare_data_for_experiment(
            conf.prepared_data_path,
            conf.init_ratio,
            conf.hm_actions_min_stream,
        )

        target_day = np.sort(left_data[DATE_DAYS].unique())[:conf.how_many_iterations][-1]
        left_data = left_data[left_data[DATE_DAYS] < target_day]

        uniq_users = left_data.groupby(DATE_DAYS)[USER_ID].nunique().sort_index().values
        uniq_interactions = left_data.groupby(DATE_DAYS)[USER_ID].count().sort_index().values
        res = {
            'n_users_dynamics': uniq_users.tolist(),
            'n_interactions_dynamics': uniq_interactions.tolist(),
        }

        with open(RESULTS_DIR / dataset / 'metric_dynamics' / 'data_dynamics.json', 'w') as f:
            json.dump(res, f)

if __name__ == '__main__':
    prepare_data()
    save_data_dynamics()