Data Every Day: マーベル vs DC

tldr

KggleのMarvel vs DCをMarvel or DC??? - Data Every Day #035に沿ってやっていきます。

実行環境はGoogle Colaboratorです。

インポート

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

import sklearn.preprocessing as sp
from sklearn.model_selection import train_test_split
import sklearn.linear_model as slm

import tensorflow as tf

データのダウンロード

Google Driveをマウントします。

from google.colab import drive
drive.mount('/content/drive')

Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount("/content/drive", force_remount=True).

KaggleのAPIクライアントを初期化し、認証します。
認証情報はGoogle Drive内（/content/drive/My Drive/Colab Notebooks/Kaggle）にkaggle.jsonとして置いてあります。

import os
kaggle_path = "/content/drive/My Drive/Colab Notebooks/Kaggle"
os.environ['KAGGLE_CONFIG_DIR'] = kaggle_path

from kaggle.api.kaggle_api_extended import KaggleApi
api = KaggleApi()
api.authenticate() 

Kaggle APIを使ってデータをダウンロードします。

dataset_id = 'leonardopena/marvel-vs-dc'
dataset = api.dataset_list_files(dataset_id)
file_name = dataset.files[0].name
file_path = os.path.join(api.get_default_download_dir(), file_name)
file_path

Warning: Looks like you're using an outdated API Version, please consider updating (server 1.5.10 / client 1.5.9)





'/content/db.csv'

api.dataset_download_file(dataset_id, file_name, force=True, quiet=False)

100%|██████████| 3.70k/3.70k [00:00<00:00, 1.02MB/s]

Downloading db.csv to /content









True

データの読み込み

Padasを使ってダウンロードしてきたCSVファイルを読み込みます。

data = pd.read_csv(file_path, encoding='latin-1')

data

	Unnamed: 0	Original Title	Company	Rate	Metascore	Minutes	Release	Budget	Opening Weekend USA	Gross USA	Gross Worldwide
0	1	Iron Man	Marvel	7.9	79	126	2008	140000000	98618668	318604126	585366247
1	2	The Incredible Hulk	Marvel	6.7	61	112	2008	150000000	55414050	134806913	263427551
2	3	Iron Man 2	Marvel	7.0	57	124	2010	200000000	128122480	312433331	623933331
3	4	Thor	Marvel	7.0	57	115	2011	150000000	65723338	181030624	449326618
4	5	Captain America: The First Avenger	Marvel	6.9	66	124	2011	140000000	65058524	176654505	370569774
5	6	The Avengers	Marvel	8.0	69	143	2012	220000000	207438708	623357910	1518812988
6	7	Iron Man Three	Marvel	7.2	62	130	2013	200000000	174144585	409013994	1214811252
7	8	Thor: The Dark World	Marvel	6.9	54	112	2013	170000000	85737841	206362140	644783140
8	9	Captain America: The Winter Soldier	Marvel	7.7	70	136	2014	170000000	95023721	259766572	714421503
9	10	Guardians of the Galaxy	Marvel	8.0	76	121	2014	170000000	94320883	333176600	772776600
10	11	Avengers: Age of Ultron	Marvel	7.3	66	141	2015	250000000	191271109	459005868	1402805868
11	12	Ant-Man	Marvel	7.3	64	117	2015	130000000	57225526	180202163	519311965
12	13	Captain America: Civil War	Marvel	7.8	75	147	2016	250000000	179139142	408084349	1153296293
13	14	Doctor Strange	Marvel	7.5	72	115	2016	165000000	85058311	232641920	677718395
14	15	Guardians of the Galaxy Vol. 2	Marvel	7.6	67	136	2017	200000000	146510104	389813101	863756051
15	16	Spider-Man: Homecoming	Marvel	7.4	73	133	2017	175000000	117027503	334201140	880166924
16	17	Thor:Ragnarok	Marvel	7.9	74	130	2017	180000000	122744989	315058289	853977126
17	18	Black Panther	Marvel	7.3	88	134	2018	200000000	202003951	700059566	1346913161
18	19	Avengers: Infinity War	Marvel	8.5	68	149	2018	321000000	257698183	678815482	2048359754
19	20	Ant-Man and the Wasp	Marvel	7.1	70	118	2018	162000000	75812205	216648740	622674139
20	21	Captain Marve	Marvel	6.9	64	123	2019	175000000	153433423	426829839	1128274794
21	22	Avengers: Endgame	Marvel	8.5	78	181	2019	356000000	357115007	858373000	2797800564
22	23	Spider-Man: Far from Home	Marvel	7.6	69	129	2019	160000000	92579212	390532085	1131927996
23	24	Catwoman	DC	3.3	27	104	2004	100000000	16728411	40202379	82102379
24	25	Batman Begins	DC	8.2	70	140	2005	150000000	48745440	206852432	373413297
25	26	Superman Returns	DC	6.0	72	154	2006	270000000	52535096	200081192	391081192
26	27	The Dark Knight	DC	9.0	84	152	2008	185000000	158411483	535234033	1004934033
27	28	Watchmen	DC	7.6	56	162	2009	130000000	55214334	107509799	185258983
28	29	Jonah Hex	DC	4.7	33	81	2010	47000000	5379365	10547117	10903312
29	30	Green Lantern	DC	5.5	39	114	2011	200000000	53174303	116601172	219851172
30	31	The Dark Knight Rises	DC	8.4	78	164	2012	250000000	160887295	448139099	1081041287
31	32	Man of Steel	DC	7.1	55	143	2013	225000000	116619362	291045518	668045518
32	33	Batman v Superman: Dawn of Justice	DC	6.5	44	151	2016	250000000	166007347	330360194	873634919
33	34	Suicide Squad	DC	6.0	40	123	2016	175000000	133682248	325100054	746846894
34	35	Wonder Woman	DC	7.4	76	141	2017	149000000	103251471	412563408	821847012
35	36	Justice League	DC	6.4	45	120	2017	300000000	93842239	229024295	657924295
36	37	Aquaman	DC	7.0	55	143	2018	160000000	67873522	335061807	1148161807
37	38	Shazam!	DC	7.1	71	132	2019	100000000	53505326	140371656	364571656
38	39	Joker	DC	8.7	59	122	2019	55000000	96202337	333204580	1060504580

下準備

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 39 entries, 0 to 38
Data columns (total 11 columns):
 #   Column               Non-Null Count  Dtype  
---  ------               --------------  -----  
 0   Unnamed: 0           39 non-null     int64  
 1   Original Title       39 non-null     object 
 2   Company              39 non-null     object 
 3   Rate                 39 non-null     float64
 4   Metascore            39 non-null     int64  
 5   Minutes              39 non-null     object 
 6   Release              39 non-null     int64  
 7   Budget               39 non-null     object 
 8   Opening Weekend USA  39 non-null     int64  
 9   Gross USA            39 non-null     int64  
 10  Gross Worldwide      39 non-null     int64  
dtypes: float64(1), int64(6), object(4)
memory usage: 3.5+ KB

不要な列の削除

data = data.drop(['Unnamed: 0', 'Original Title'], axis=1)

エンコード

encoder = sp.LabelEncoder()
data['Company'] = encoder.fit_transform(data['Company'])

可視化

data['Company'] = encoder.fit_transform(data['Company'])
plt.figure(figsize=(12, 10))
sns.heatmap(data.corr(), annot=True, vmax=1, vmin=-1)
plt.show()

スケーリングとデータの分割

y = data['Company']
X = data.drop('Company', axis=1)

スケーリング

scaler = sp.StandardScaler()
X = scaler.fit_transform(X)

分割

X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.7)

トレーニング

model = slm.LogisticRegression()
model.fit(X_train, y_train)
model.score(X_test, y_test)

0.8333333333333334

同じモデルで何回かトレーニングしてみましたが、かなり精度に差がありました。
データが少ないからしょうがないかなー