import requests
from bs4 import BeautifulSoup
import re
import pandas as pd

def horse_page_url(url):
    req = requests.get(url)
    BeautifulSoup(req.content, 'html.parser')
    soup = BeautifulSoup(req.content, 'html.parser')
    table1 = soup.find('table', attrs={'class':'race_table_01 nk_tb_common'})
    table2 = soup.findAll('td', attrs={'class':'txt_l'})
    all_data_keiba=[]
    for a in table2:
        b=a.find('a')
        if b != None:
            c=b.get('href')
            if "horse" in c:
                print(c)
                data=race_page_url(c)
                for i in range(len(data)):
                    all_data_keiba.append(data[i])
    return all_data_keiba

def race_page_url(url):
    req = requests.get(url)
    BeautifulSoup(req.content, 'html.parser')
    soup = BeautifulSoup(req.content, 'html.parser')
    race_table = soup.findAll("table")[4]
    race_table1 = race_table.find_all("a")
    all_data=[]
    if race_table1==[]:
        race_table = soup.findAll("table")[3]
        race_table1=race_table.find_all("a")
    for a in race_table1:
        b=a.get('href')
        if "race" in b
            if "2018" in b:
                if "list" in b or "sum" in b or "movie" in b:
                    b=b
                else:
                    if b.islower() == True:
                        b='https://db.netkeiba.com' + b
                        race=race_data(b)
                        horse=horse_page_url2(b)
                        for i in range(len(horse)):
                            horse[i]=horse[i]+race
                        for i in range(len(horse)):
                            all_data.append(horse[i])
    return all_data

def race_data(url):
    req = requests.get(url)
    BeautifulSoup(req.content, 'html.parser')
    soup = BeautifulSoup(req.content, 'html.parser')
    race_data=[]
    race_name = soup.find("p",attrs={"class":"smalltxt"})
    print(race_name)
    race_table = soup.findAll("table")[0]
    race_table2= soup.findAll("span")[6]
    text=str(race_table2.text)
    post_code = re.findall(r'\d+', text)
    code=int(post_code[0])
    race_data.append(code)
    race=[]
    if "芝" in race_table2.text:
        race_data.append(0)
    elif "ダ" in race_table2.text:
        race_data.append(1)

    if "右" in race_table2.text:
        race_data.append(0)
    elif "左" in race_table2.text:
        race_data.append(1)
    if "障" in race_table2.text:
        race_data.append(2)

    if "晴" in race_table2.text:
        race_data.append(0)
    elif "曇" in race_table2.text:
        race_data.append(1)
    elif "雨" in race_table2.text:
        race_data.append(2)

    if "良" in race_table2.text:
        race_data.append(0)
    elif "稍" in race_table2.text:
        race_data.append(1)
    elif "重" in race_table2.text:
        race_data.append(2)
    
    if "中山" in race_table2.text:
        race_data.append(0)
    else:
        race_data.append(1)

    return race_data

def horse_page_url2(url):
    req = requests.get(url)
    BeautifulSoup(req.content, 'html.parser')
    soup = BeautifulSoup(req.content, 'html.parser')
    table1 = soup.find('table', attrs={'class':'race_table_01 nk_tb_common'})
    table2 = soup.findAll('td', attrs={'class':'txt_l'})
    all_data=[]
    data=[]
    i=0
    for a in table2:
        b=a.find('a')
        if b != None:
            c=b.get('href')
            if "horse" in c:
                c='https://db.netkeiba.com' + c
                all_data.append(horse_data(c))
                
    for i in range(len(all_data)-1):
        for k in range(i+1,len(all_data)):
            data.append(all_data[i]+all_data[k])
    
    all_data2=all_data[::-1]
    
    for i in range(len(all_data2)-1):
        for k in range(I+1,len(all_data2)):
            data.append(all_data2[i]+all_data2[k])
    return data

def horse_data(url):
    req = requests.get(url)
    BeautifulSoup(req.content, 'html.parser')
    soup = BeautifulSoup(req.content, 'html.parser')
    race_table2=soup.findAll("table")[0]
    horse_name=soup.find("div",attrs={'class':'horse_title'})
       race_table4=soup.find("p",attrs={'class':'rate'})
        a=float(race_table4.text)
        data=[]
        all_data=[]
        data.append(a)
        for i in range(25):
            if i % 5 == 1:
                race_table3=str(race_table2.findAll("img")[i])
                data.append(int(re.findall(r'\d+', race_table3)[1]))
    return data

from keras.models import Sequential
from keras.layers import Activation, Dense
from keras.utils.np_utils import to_categorical
from keras.models import Sequential
from keras.layers import Dense, Dropout
from keras.optimizers import Adam
from keras.layers.normalization import BatchNormalization
import numpy as np
import keras

X = np.array(X_list)
# Yは0と1からなるリスト
Y = to_categorical(Y_list)

# モデル
model = Sequential()
# 全結合層(18を500に)
model.add(Dense(input_dim=18, output_dim=500))
# 活性化関数(ReLu関数) 
model.add(BatchNormalization())
model.add(Activation("relu"))
# 全結合層(500を2に) 
model.add(Dense(output_dim=2))
# softmax関数
model.add(Activation("softmax"))
# コンパイル
optimizer = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)
model.compile(loss="categorical_crossentropy", optimizer=optimizer, metrics=["accuracy"])
# 実行
model.fit(X, Y, nb_epoch=300, batch_size=50,validation_split=0.1)

# 予測
results = model.predict_proba(np.array(X_test))
# 結果
print("Predict:\n", results)

# モデル
model = Sequential()
# 全結合層(18を500に)
model.add(Dense(input_dim=18, output_dim=500))
# 活性化関数(ReLu関数) ドロップアウトを0.4にする
model.add(BatchNormalization())
model.add(Activation("relu"))
model.add(Dropout(0.4))
# 全結合層(500を2に) ドロップアウトを0.2にする
model.add(Dense(output_dim=2))
model.add(Dropout(0.2))
# softmax関数
model.add(Activation("softmax"))
# コンパイル
optimizer = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)
model.compile(loss="categorical_crossentropy", optimizer=optimizer, metrics=["accuracy"])
# 実行
model.fit(X, Y, nb_epoch=300, batch_size=50,validation_split=0.1)

# 予測
results = model.predict_proba(np.array(X_test))
# 結果
print("Predict:\n", results)

l=16
q=[]
p=[]
results2=results[::-1]
for i in range(l):
    s1=0
    b1=0
    c1=0
    n1=i
    b1=(i)*(l-1)-sum(range(1, n1))
    c1=(i+1)*(l-1)-sum(range(1, n1 + 1))
    a1=l-1
    for k in range(b1,c1):
        s1+=results[k][1]
    for k in range(n1):
        s1+=results[n1-1][0]
        a1-=1
        n1+=a1
    q.append(s1)
print(q)

for i in range(l):
    s2=0
    b2=0
    c2=0
    n2=i
    b2=(i)*(l-1)-sum(range(1, n2))
    c2=(i+1)*(l-1)-sum(range(1, n2 + 1))
    a2=l-1
    for k in range(b2,c2):
        s2+=results2[k][1]
    for k in range(n2):
        s2+=results2[n2-1][0]
        a2-=1
        n2+=a2
    p.append(s2)
p=p[::-1]
print(p)
for i in range(l):
    print(i+1,q[i]+p[i])

from sklearn import preprocessing
model = preprocessing.StandardScaler()
X = model.fit_transform(X_train)
from sklearn import preprocessing
model = preprocessing.StandardScaler()
X_test = model.fit_transform(X_test)