롤 승패 예측 모델 만들기

롤 승부 예측 모델

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
import numpy as np  

Data_set = np.loadtxt("/content/Challenger_Ranked_Games.csv", delimiter=",", skiprows=1)
X = Data_set[:, [1, 3, 5, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18]]
y = Data_set[:,2]

model = Sequential()
model.add(Dense(30, input_dim=14, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
history = model.fit(X, y, epochs=1000, batch_size=90)

데이터에서 게임 시간, 골드, 킬 등을 넣어서 학습을 시켰다.

Epoch 997/1000
299/299 [==============================] - 1s 4ms/step - loss: 0.8367 - accuracy: 0.9461
Epoch 998/1000
299/299 [==============================] - 1s 4ms/step - loss: 0.6947 - accuracy: 0.9535
Epoch 999/1000
299/299 [==============================] - 1s 4ms/step - loss: 0.9827 - accuracy: 0.9405
Epoch 1000/1000
299/299 [==============================] - 1s 4ms/step - loss: 1.1175 - accuracy: 0.9377

학습 결과로 0.9377이라는 정확도가 나옴

테스트 해보기

Data_set1 = np.loadtxt("/content/GrandMaster_Ranked_Games.csv", delimiter=",", skiprows=1)
X1 = Data_set[:, [1, 3, 5, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18]]
y1 = Data_set[:,2]
score = model.evaluate(X1, y1)
print('Test accuracy: ', score[1])

그랜드마스터 티어의 데이터로 테스트 진행

841/841 [==============================] - 2s 2ms/step - loss: 1.1704 - accuracy: 0.9384
Test accuracy: 0.9384106397628784

다음과 같은 결과가 나옴