#https://qiita.com/Nelca/items/8c0aaaf2a8452352ffe4
#https://ichi.pro/post/218892206272728
#https://keras.io/ja/getting-started/sequential-model-guide/
#https://www.tcom242242.net/entry/ai-2/deeplearning/%E3%80%90%E8%B6%85%E5%85%A5%E9%96%80%E3%80%81%E3%83%8B%E3%83%A5%E3%83%BC%E3%83%A9%E3%83%AB%E3%83%8D%E3%83%83%E3%83%88%E3%83%AF%E3%83%BC%E3%82%AF%E3%80%91keras%E3%81%A7%EF%BC%92%E5%80%A4%E5%88%86/


import numpy as np
from scipy import signal
import matplotlib.pyplot as plt
#import pandas as pd
#from pandas import DataFrame


data = np.loadtxt(
    fname="daondata.csv",    # 読み込みたいファイルのパス
    dtype="float",       # 要素の方を指定
    delimiter=",",       # ファイルの区切り文字
)

[N,M]=data.shape

dt = 0.00005
fs = 1/dt
t=np.array(range(0,N-2))*dt

hakei=1#1の場合波形　それ以外はパワースペクトルで分類

if hakei==1:
    for i in range(0,M):
        if i==0:
            b1=data.T[0][0:N-2]
            xx=b1
            y=data.T[0][N-1]
        else:
            b1=data.T[i][0:N-2]
            xx=np.vstack((xx, b1))
            y=np.append(y,data.T[i][N-2])
            
    plt.figure(figsize=(12, 8))
    # １つ目のプロット
    plt.subplot(2,1,1)
    plt.plot(t,xx[0], label="ukinashi")
    plt.xlabel('time[sec]')
    plt.ylabel('roudness[dB]')
    plt.legend()

    # 2つ目のプロット
    plt.subplot(2,1,2)
    plt.plot(t, xx[302], label="ukiari")
    plt.xlabel('time[sec]')
    plt.ylabel('roudness[dB]')
    plt.legend()
    plt.show()

else:
    for i in range(0,M):
        if i==0:
            a1=data.T[0][0:N-2]
            freq, b1 = signal.welch(a1, fs)
            xx=b1
            y=data.T[0][N-1]
        else:
            a1=data.T[i][0:N-2]
            freq, b1 = signal.welch(a1, fs)
            xx=np.vstack((xx, b1))
            y=np.append(y,data.T[i][N-2])
            
    plt.figure(figsize=(12, 8))
    # １つ目のプロット
    plt.subplot(2,1,1)
    plt.plot(freq,xx[0], label="ukinashi")
    plt.xlabel('freq[Hz]')
    plt.ylabel('psp')
    plt.legend()

    # 2つ目のプロット
    plt.subplot(2,1,2)
    plt.plot(freq,xx[302], label="ukiari")
    plt.xlabel('freq[Hz]')
    plt.ylabel('psp')
    plt.legend()
    plt.show()

    
from sklearn import model_selection

import keras
from keras.models import Model, Sequential
from keras.layers import Input, Dense, Dropout, Activation
from keras.layers import Conv2D, GlobalAveragePooling2D
from keras.layers import BatchNormalization, Add
from keras.callbacks import EarlyStopping, ModelCheckpoint


#データの作成
randomseed=1
x_train, x_test, y_train, y_test = model_selection.train_test_split(xx, y, test_size=0.25, stratify=y,random_state=randomseed)
#x_train, x_test, y_train, y_test = model_selection.train_test_split(xx, y, test_size=0.25, stratify=y,shuffle=True)
print("x train:{0}\ny train:{1}\nx test:{2}\ny test:{3}".format(len(x_train),
                                                                len(y_train), 
                                                                len(x_test), 
                                                                len(y_test)))

#from keras.utils import np_utils
#Y_one_hot = np_utils.to_categorical(y_train)

classes = 2
y_train = keras.utils.to_categorical(y_train, classes)
y_test = keras.utils.to_categorical(y_test, classes)


# モデルの作成
model = Sequential()
model.add(Dense(10, input_dim=len(x_train[0])))    # 入力層2ノード, 隠れ層に3ノード, 全結合
model.add(Activation("sigmoid"))    # 活性化関数はsigmoid
#model.add(Activation("relu"))    # 活性化関数はrelu
model.add(Dense(10)) # 層の追加
model.add(Dense(2)) # 出力層2ノード,全結合
model.add(Activation("sigmoid"))

    
# 2値分類問題の場合
model.compile(optimizer='rmsprop',
#model.compile(optimizer="adam",
              loss='binary_crossentropy',
              metrics=['accuracy'])


# 訓練
history = model.fit(x_train, y_train, nb_epoch=20, batch_size=1)

plt.plot(history.epoch, history.history["acc"], label="acc")
plt.plot(history.epoch, history.history["loss"], label="loss")
plt.xlabel("epoch")
plt.legend()
plt.show()

# 評価
score = model.evaluate(x_test, y_test, verbose=1)

print("Test score", score[0])#Test scoreが誤差
print("Test accuracy", score[1])#Test accuracy正答率

classes = model.predict_classes(x_test, batch_size=1)
prob=model.predict_proba(x_test, batch_size=1)