Python 機器學習: 如何將if ~ else ~語法寫為一列,塞進lambda函數中, pandas.DataFrame 如何使用 .apply(func) 增加新的一欄? model = tensorflow.keras.models.Sequential() #均一化資料

lis_col = df_dp.columns.to_list()
for col in lis_col:
    max_value = df_dp[col].max()
    ratio = 1/max_value
    df_dp[col] =   df_dp[col]*ratio

# -*- coding: utf-8 -*-
"""
Created on Sun Nov 12 15:27:56 2023

@author: SavingKing
"""

import os
import pandas as pd
import numpy as np
dirname = r"P:\Python class\powen\AI人工智慧自然語言與語音語意辨識開發應用實務班\powen2_AI人工智慧自然語言與語音語意辨識開發應用實務班\day6\ch07-MLP-flower"
basename = "weather.csv"

fpath = os.path.join(dirname,basename)
#'C:\\Users\\SavingKing\\Downloads\\day6\\ch07-MLP-flower\\weather.csv'

df = pd.read_csv(fpath)

func = lambda strr: 1 if strr == "Yes" else 0

df["RainTomorrow_01"] = df["RainTomorrow"].apply( func )

lis_drop_col = ["WindGustDir", "WindDir9am","WindDir3pm","RainToday","RainTomorrow"]
#非數值資料的欄位先drop，只留下數值資料的欄位
#WindSpeed9am 資料含有NA

df_dp = df.drop(columns = lis_drop_col)
#[366 rows x 18 columns]


#資料中含有NA,需要先dropna做資料清洗,不然loss, predict會出現nan
df_dp.dropna(axis=0, how='any', subset=None, inplace=True)
#[354 rows x 18 columns]

lis_col = df_dp.columns.to_list()
for col in lis_col:
    max_value = df_dp[col].max()
    ratio = 1/max_value
    df_dp[col] =   df_dp[col]*ratio

X = df_dp.drop(columns="RainTomorrow_01").values
#[366 rows x 17 columns]
Y = df_dp.iloc[:,-1].values
#pandas.core.series.Series
"""
X.shape
Out[61]: (366, 17)

Y.shape
Out[62]: (366,)
"""

from sklearn.model_selection import train_test_split
import tensorflow as tf

category=2   #下雨/不下雨 分別用1 0 代表, 兩種結果
dim=X.shape[1] #有幾個特徵值?
x_train , x_test , y_train , y_test = train_test_split(X,Y,test_size=0.05)
y_train2=tf.keras.utils.to_categorical(y_train, num_classes=(category))
y_test2=tf.keras.utils.to_categorical(y_test, num_classes=(category))
#One-hot Encoding 單熱編碼
#y_train y_test 都是一維的資料

print("x_train[:4]",x_train[:4])
print("y_train[:4]",y_train[:4])
print("y_train2[:4]",y_train2[:4])

# 建立模型
model = tf.keras.models.Sequential()
model.add(tf.keras.layers.Dense(units=200,
    activation=tf.nn.relu,
    input_dim=dim)) #input_dim 有幾個特徵值不能錯
model.add(tf.keras.layers.Dense(units=200,
    activation=tf.nn.relu ))
model.add(tf.keras.layers.Dense(units=200,
    activation=tf.nn.relu ))
model.add(tf.keras.layers.Dense(units=200,
    activation=tf.nn.relu ))
model.add(tf.keras.layers.Dense(units=category,
    activation=tf.nn.softmax )) #最後一層units=category也不能錯
model.compile(optimizer='adam',
    loss=tf.keras.losses.categorical_crossentropy,
    metrics=['accuracy'])
model.fit(x_train, y_train2,
          epochs=200,
          batch_size=64)

#測試
model.summary()

score = model.evaluate(x_test, y_test2)
print("score:",score)

predict = model.predict(x_test)
print("predict:",predict)
lis_ans = [np.argmax(predict[i]) for i in range(predict.shape[0]) ]
print("Ans:\t",lis_ans)
# print("Ans:",np.argmax(predict[0]),np.argmax(predict[1]),np.argmax(predict[2]),np.argmax(predict[3]))

# =============================================================================
# predict2 = model.predict_classes(x_test)
# print("predict_classes:",predict2)
# print("y_test",y_test[:])
# =============================================================================

lis_col = df_dp.columns.to_list()
for col in lis_col:
    max_value = df_dp[col].max()
    min_value = df_dp[col].min()
    rng = max_value-min_value
    
    df_dp[col] =   (df_dp[col]-min_value)/rng
df_dp.to_excel("./df_dp.xlsx")