Kaggle_tweet_emotion_w2v_rnn

最新推荐文章于 2020-05-25 22:04:45 发布
最新推荐文章于 2020-05-25 22:04:45 发布
阅读量219 收藏
点赞数
分类专栏： 比赛
版权声明：本文为博主原创文章，遵循 CC 4.0 BY 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/weixin_42681868/article/details/105719787
版权 
import pandas as pd
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
import os
from gensim.models.word2vec import Word2Vec
import tqdm
import tensorflow as tf
print(tf.__version__)
from tensorflow import keras
from sklearn.model_selection import train_test_split
from tensorflow.keras import Sequential as Sequential
import nltk
from nltk.corpus import stopwords


2.0.0

# set config 
BATCH_SIZE = 64
SEQUENCE_LENGTH = 22
WORD_SIZE = 200

log

1:w2v + 1 layer lstm : 0.77(valid data)

2:w2v + 2 layer lstm : 0.77

3:w2v + 4 layer lstm : 0.77

4:w2v + 1 layer bi-lstm : 0.76

5:w2v + 4 layer bi-lstm : 0.77

6:w2v + 4 layer bi-lstm + dropout(0.5 层间): 0.77

7:w2v + 4 layer bi-lstm + dropout(0.5 层间): 0.77

8:w2v + 8 layer bi-lstm + dropout(0.5 层间): 0.76

9:w2v + 3 layer bi-lstm + dropout(0.4 input gate 之前):0.76

10:w2v + 3 layer bi-lstm + dropout(0.3 input gate 之前):0.76

11:w2v + 3 layer bi-lstm + dropout(0.3 input gate 之前) + batch = 32(加倍batch):0.76

12:w2v + 3 layer bi-lstm + dropout(0.1 input gate 之前) + batch = 64 :0.76

小结:bidirection 在这个案例中无效

#load data
path_home = "/home/lowry/pro/kaggle_tweets/kaggle_tweets_emotion"
data_train = pd.read_csv(os.path.join(path_home,"train.csv"))
data_test = pd.read_csv(os.path.join(path_home,"test.csv"))
data_submit = pd.read_csv(os.path.join(path_home,"sample_submission.csv"))

data_train.head()
idkeywordlocationtexttarget
01NaNNaNOur Deeds are the Reason of this #earthquake M...1
14NaNNaNForest fire near La Ronge Sask. Canada1
25NaNNaNAll residents asked to 'shelter in place' are ...1
36NaNNaN13,000 people receive #wildfires evacuation or...1
47NaNNaNJust got sent this photo from Ruby #Alaska as ...1
# data clearn

stopwords_english = stopwords.words("english")

import re
def cleanword(s):
    s = s.lower()
    s = " ".join([word for word in s.split(" ") if word not in stopwords_english])
    temp = re.findall("http\S*",s)  
    for deletStr in temp:
        if deletStr != "":
            s = s.replace(deletStr," ")
    temp = re.findall("@\S*",s)
    for deletStr in temp:
        if deletStr != "":
            s = s.replace(deletStr," ")
    temp = re.findall("\d*",s)
    for deletStr in temp:
        if deletStr != "":
            s = s.replace(deletStr," ")
    temp = re.findall("\x89\S*",s)
    for deletStr in temp:
        if deletStr != "":
            s = s.replace(deletStr[:5]," ")

    s = s.replace("\n"," ")
    s = s.replace(","," ")
    s = s.replace("?"," ")
    s = s.replace("..."," ")
    s = s.replace("."," ")
    s = s.replace("["," ")
    s = s.replace("]"," ")
    s = s.replace("!"," ")
    s = s.replace(":"," ")
    s = s.replace("-"," ")
    s = s.replace("#"," ")
    s = s.replace("|"," ")
    s = s.replace("("," ")
    s = s.replace(")"," ")
    s = s.replace(";"," ")
    s = s.replace("="," ")
    s = s.replace(">"," ")
    s = s.replace("<"," ")
    s = s.replace("/"," ")
    

    #delet conntinue " "
    s_new_list = [word for word in s.split(" ") if word != " "]
    s_new = " ".join(s_new_list)
            
    s_new = s_new.strip()
    return s_new
data_test['text'] = data_test['text'].apply(cleanword)
data_train['text'] = data_train['text'].apply(cleanword)


data_train_copy = data_train.copy()
data_train_copy["text_length"] = data_train_copy["text"].apply(lambda s : len(s.split(" ")))
data_test_copy = data_train.copy()
data_test_copy["text_length"] = data_train_copy["text"].apply(lambda s : len(s.split(" ")))
print("train_text_words_length(95%) = ",np.percentile(data_train_copy["text_length"].tolist(),95))
print("test_text_words_length(95%) = ",np.percentile(data_test_copy["text_length"].tolist(),95))

train_text_words_length(95%) =  22.0
test_text_words_length(95%) =  22.0

# print(data_train_copy)
# print(data_train_copy.info(memory_usage=True))

1

#train word2vec model 


# sentences = []
# for line in data_train['text'].values:
#     sentences.append(list(line.split(" ")))
# for line in data_test['text'].values:
#     sentences.append(list(line.split(" ")))
# print(len(sentences))
# path_model_w2v = os.path.join(path_home,"w2v_model.model")
# model_w2v = Word2Vec(
#         sentences=sentences,
#         size=200,#维度
#         alpha=0.025, #默认
#         window=5, #默认
#         min_count=2,#2，3
#         sample=0.001,#
#         seed=2018, #
#         workers=11, #线程
#         min_alpha=0.0001, 
#         sg=0, #cbow
#         hs=0, #负采样
#         negative=5,#负采样个数
#         ns_exponent=0.75, 
#         cbow_mean=1,#求和再取平均
#         iter=10 #10到15
#         )
# model_w2v.save(path_model_w2v)




# test: 
# model_w2v = Word2Vec.load(path_model_w2v)
# print(model_w2v)
# model_w2v.wv["our"].shape
# for word , wordInfo in model_w2v.wv.vocab.items():
#     print("word = ",word)
#     print("wordInfo = ",wordInfo )
#     break


10876

#build word2vec map
path_model_w2v = os.path.join(path_home,"w2v_model.model")
model_w2v = Word2Vec.load(path_model_w2v)
vocab_list = [word for word, Vocab in model_w2v.wv.vocab.items()]# 存储 所有的 词语
word_index = {" ": 0}# 初始化 `[word : token]` ，后期 tokenize 语料库就是用该词典 (把词语改成下表)。
word_vector = {} # 初始化`[word : vector]`字典
# 初始化存储所有向量的大矩阵，留意其中多一位（首行），词向量全为 0，用于 padding补零。
# 行数 为 所有单词数+1 比如 10000+1 ； 列数为 词向量“维度”比如100。
embedding_matrix = np.zeros( (len(vocab_list)+1,model_w2v.vector_size) )  #embedding 矩阵每一列都是一个词向量
print(embedding_matrix.shape)

#填充上述字典和大矩阵
for i in range(len(vocab_list)):
    word = vocab_list[i]   #每个词语
    word_index[word] = i+1 #词语->序号
    word_vector[word] = model_w2v.wv[word] #词语->词向量
    embedding_matrix[i+1] = model_w2v.wv[word] #序号->词向量

(8010, 200)

def tokenizer(texts, word_index):
    data = []
    for sentence in texts:  #遍历每一行
        new_txt = []
        for word in sentence.split(" "):
            try:
                new_txt.append(word_index[word])  # 把句子中的 词语转化为index
            except:
                pass
        
        new_txt = new_txt[:SEQUENCE_LENGTH]
        padding_length = SEQUENCE_LENGTH - len(new_txt)
        if padding_length > 0:
            new_txt += [0]*padding_length
        data.append(new_txt)
    return np.array(data)

X_train_tokenizer = tokenizer(data_train["text"].values,word_index)
X_test_tokenizer = tokenizer(data_test["text"].values,word_index)



x_train ,x_valid , y_train , y_valid = train_test_split(X_train_tokenizer,data_train["target"].values)

x_train.shape

(5709, 22)

def lstm(BATCH_SIZE,SEQUENCE_LENGTH,WORD_SIZE):
    model = keras.Sequential()
    #输入层
    model.add(    
        keras.layers.Embedding(
            input_dim = len(embedding_matrix),
            output_dim = WORD_SIZE,
            weights=[embedding_matrix],     #预训练的词向量系数
            input_length = SEQUENCE_LENGTH,
            trainable = False                #是否在训练过程中更新词向量
        ) 
    )
    

    
    
    #隐层
    for i in range(3):
#         model.add(keras.layers.LSTM(64,activation='tanh',return_sequences=True,dropout=0.1))
        model.add(keras.layers.LSTM(64,activation='tanh',return_sequences=True))
#         model.add(keras.layers.Bidirectional(keras.layers.LSTM(64,activation='tanh',return_sequences=True,dropout=0.1)))
#         model.add(keras.layers.Dropout(0.5))
#         model.add(keras.layers.BatchNormalization())

    model.add(keras.layers.LSTM(64,activation='tanh'))
    
    
    #输出层
    model.add(keras.layers.Dense(1,activation="sigmoid"))
    
    model.compile(
        loss="binary_crossentropy",
        optimizer="adam",
        metrics=["accuracy"]
    )
    return model

model = lstm(BATCH_SIZE,SEQUENCE_LENGTH,WORD_SIZE)
model.summary()

Model: "sequential_21"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
embedding_21 (Embedding)     (None, 22, 200)           1602000   
_________________________________________________________________
lstm_102 (LSTM)              (None, 22, 64)            67840     
_________________________________________________________________
lstm_103 (LSTM)              (None, 22, 64)            33024     
_________________________________________________________________
lstm_104 (LSTM)              (None, 22, 64)            33024     
_________________________________________________________________
lstm_105 (LSTM)              (None, 64)                33024     
_________________________________________________________________
dense_21 (Dense)             (None, 1)                 65        
=================================================================
Total params: 1,768,977
Trainable params: 166,977
Non-trainable params: 1,602,000
_________________________________________________________________

history = model.fit(
    x_train,
    y_train,
    validation_data=(x_valid,y_valid),
    epochs = 60,
    batch_size= BATCH_SIZE,
)

Train on 5709 samples, validate on 1904 samples
Epoch 1/60
5709/5709 [==============================] - 10s 2ms/sample - loss: 0.6199 - accuracy: 0.6614 - val_loss: 0.5872 - val_accuracy: 0.6949
Epoch 2/60
5709/5709 [==============================] - 1s 249us/sample - loss: 0.6039 - accuracy: 0.6744 - val_loss: 0.6035 - val_accuracy: 0.6702
Epoch 3/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5763 - accuracy: 0.7043 - val_loss: 0.5729 - val_accuracy: 0.7048
Epoch 4/60
5709/5709 [==============================] - 1s 249us/sample - loss: 0.5697 - accuracy: 0.7084 - val_loss: 0.5840 - val_accuracy: 0.6843
Epoch 5/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5733 - accuracy: 0.7057 - val_loss: 0.5598 - val_accuracy: 0.7159
Epoch 6/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.5620 - accuracy: 0.7183 - val_loss: 0.5613 - val_accuracy: 0.7164
Epoch 7/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5538 - accuracy: 0.7208 - val_loss: 0.5454 - val_accuracy: 0.7227
Epoch 8/60
5709/5709 [==============================] - 1s 249us/sample - loss: 0.5453 - accuracy: 0.7334 - val_loss: 0.5382 - val_accuracy: 0.7327
Epoch 9/60
5709/5709 [==============================] - 1s 249us/sample - loss: 0.5509 - accuracy: 0.7255 - val_loss: 0.5407 - val_accuracy: 0.7311
Epoch 10/60
5709/5709 [==============================] - 1s 252us/sample - loss: 0.5544 - accuracy: 0.7194 - val_loss: 0.5543 - val_accuracy: 0.7337
Epoch 11/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.5406 - accuracy: 0.7274 - val_loss: 0.5313 - val_accuracy: 0.7390
Epoch 12/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5377 - accuracy: 0.7325 - val_loss: 0.5391 - val_accuracy: 0.7321
Epoch 13/60
5709/5709 [==============================] - 1s 250us/sample - loss: 0.5265 - accuracy: 0.7465 - val_loss: 0.5458 - val_accuracy: 0.7321
Epoch 14/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5335 - accuracy: 0.7352 - val_loss: 0.5556 - val_accuracy: 0.7269
Epoch 15/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5299 - accuracy: 0.7437 - val_loss: 0.5301 - val_accuracy: 0.7405
Epoch 16/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.5247 - accuracy: 0.7462 - val_loss: 0.5677 - val_accuracy: 0.7274
Epoch 17/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5229 - accuracy: 0.7485 - val_loss: 0.5201 - val_accuracy: 0.7426
Epoch 18/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.5315 - accuracy: 0.7359 - val_loss: 0.5246 - val_accuracy: 0.7463
Epoch 19/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.5240 - accuracy: 0.7437 - val_loss: 0.5345 - val_accuracy: 0.7316
Epoch 20/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5256 - accuracy: 0.7399 - val_loss: 0.5279 - val_accuracy: 0.7468
Epoch 21/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.5160 - accuracy: 0.7551 - val_loss: 0.5166 - val_accuracy: 0.7521
Epoch 22/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5224 - accuracy: 0.7462 - val_loss: 0.5210 - val_accuracy: 0.7521
Epoch 23/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5236 - accuracy: 0.7444 - val_loss: 0.5187 - val_accuracy: 0.7500
Epoch 24/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5285 - accuracy: 0.7409 - val_loss: 0.5789 - val_accuracy: 0.7447
Epoch 25/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.5236 - accuracy: 0.7493 - val_loss: 0.5223 - val_accuracy: 0.7484
Epoch 26/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5126 - accuracy: 0.7511 - val_loss: 0.5415 - val_accuracy: 0.7337
Epoch 27/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5133 - accuracy: 0.7499 - val_loss: 0.5087 - val_accuracy: 0.7558
Epoch 28/60
5709/5709 [==============================] - 1s 250us/sample - loss: 0.5091 - accuracy: 0.7502 - val_loss: 0.5145 - val_accuracy: 0.7489
Epoch 29/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5125 - accuracy: 0.7493 - val_loss: 0.5141 - val_accuracy: 0.7505
Epoch 30/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.5104 - accuracy: 0.7514 - val_loss: 0.5357 - val_accuracy: 0.7285
Epoch 31/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.5051 - accuracy: 0.7500 - val_loss: 0.5242 - val_accuracy: 0.7384
Epoch 32/60
5709/5709 [==============================] - 1s 250us/sample - loss: 0.5008 - accuracy: 0.7588 - val_loss: 0.5083 - val_accuracy: 0.7526
Epoch 33/60
5709/5709 [==============================] - 1s 249us/sample - loss: 0.5096 - accuracy: 0.7527 - val_loss: 0.5223 - val_accuracy: 0.7463
Epoch 34/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.4977 - accuracy: 0.7581 - val_loss: 0.5096 - val_accuracy: 0.7563
Epoch 35/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5061 - accuracy: 0.7555 - val_loss: 0.5132 - val_accuracy: 0.7574
Epoch 36/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.5076 - accuracy: 0.7544 - val_loss: 0.5090 - val_accuracy: 0.7558
Epoch 37/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.4993 - accuracy: 0.7613 - val_loss: 0.5265 - val_accuracy: 0.7468
Epoch 38/60
5709/5709 [==============================] - 1s 244us/sample - loss: 0.4971 - accuracy: 0.7630 - val_loss: 0.5162 - val_accuracy: 0.7579
Epoch 39/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4979 - accuracy: 0.7623 - val_loss: 0.5178 - val_accuracy: 0.7426
Epoch 40/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.4977 - accuracy: 0.7586 - val_loss: 0.5048 - val_accuracy: 0.7537
Epoch 41/60
5709/5709 [==============================] - 1s 248us/sample - loss: 0.4993 - accuracy: 0.7642 - val_loss: 0.5071 - val_accuracy: 0.7516
Epoch 42/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4915 - accuracy: 0.7628 - val_loss: 0.5114 - val_accuracy: 0.7600
Epoch 43/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.4857 - accuracy: 0.7705 - val_loss: 0.5226 - val_accuracy: 0.7558
Epoch 44/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4879 - accuracy: 0.7676 - val_loss: 0.5074 - val_accuracy: 0.7647
Epoch 45/60
5709/5709 [==============================] - 1s 244us/sample - loss: 0.4837 - accuracy: 0.7709 - val_loss: 0.5122 - val_accuracy: 0.7574
Epoch 46/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4837 - accuracy: 0.7681 - val_loss: 0.5023 - val_accuracy: 0.7558
Epoch 47/60
5709/5709 [==============================] - 1s 244us/sample - loss: 0.4820 - accuracy: 0.7688 - val_loss: 0.5541 - val_accuracy: 0.7279
Epoch 48/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4844 - accuracy: 0.7665 - val_loss: 0.5059 - val_accuracy: 0.7521
Epoch 49/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4790 - accuracy: 0.7739 - val_loss: 0.5055 - val_accuracy: 0.7621
Epoch 50/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.4820 - accuracy: 0.7716 - val_loss: 0.5064 - val_accuracy: 0.7579
Epoch 51/60
5709/5709 [==============================] - 1s 245us/sample - loss: 0.4840 - accuracy: 0.7672 - val_loss: 0.5160 - val_accuracy: 0.7521
Epoch 52/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4745 - accuracy: 0.7733 - val_loss: 0.5221 - val_accuracy: 0.7584
Epoch 53/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4749 - accuracy: 0.7758 - val_loss: 0.5186 - val_accuracy: 0.7474
Epoch 54/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4699 - accuracy: 0.7775 - val_loss: 0.5061 - val_accuracy: 0.7568
Epoch 55/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4683 - accuracy: 0.7767 - val_loss: 0.5178 - val_accuracy: 0.7432
Epoch 56/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4728 - accuracy: 0.7756 - val_loss: 0.5245 - val_accuracy: 0.7516
Epoch 57/60
5709/5709 [==============================] - 1s 247us/sample - loss: 0.4718 - accuracy: 0.7730 - val_loss: 0.5179 - val_accuracy: 0.7495
Epoch 58/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4681 - accuracy: 0.7781 - val_loss: 0.5209 - val_accuracy: 0.7547
Epoch 59/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4612 - accuracy: 0.7847 - val_loss: 0.5193 - val_accuracy: 0.7521
Epoch 60/60
5709/5709 [==============================] - 1s 246us/sample - loss: 0.4562 - accuracy: 0.7886 - val_loss: 0.5204 - val_accuracy: 0.7563

def draw(history):
    data = history.history
    data = pd.DataFrame(data)
    data.plot()
    plt.show()
draw(history)

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-FVgzpJGV-1587653490846)(output_16_0.png)]

在这里插入图片描述

wenqiang su
关注
专栏目录
Tweet with Disaster（Kaggle NLP项目实战）
Steve_Zhao_Gan的博客
05-27 4089
Tweet with Disaster（Kaggle NLP项目实战）欢迎使用Markdown编辑器新的改变功能快捷键合理的创建标题，有助于目录的生成如何改变文本的样式插入链接与图片如何插入一段漂亮的代码片生成一个适合你的列表创建一个表格设定内容居中、居左、居右SmartyPants创建一个自定义列表如何创建一个注脚注释也是必不可少的KaTeX数学公式新的甘特图功能，丰富你的文章UML 图表FLowchart流程图导出与导入导出导入 欢迎使用Markdown编辑器 你好！ 这是你第一次使用 Markdown
Keras中的Embedding和Word2vec的区别
weixin_39963132的博客
12-29 6796
最近在接触深度学习，在自然语言处理的应用中，不可避免的要把文本转换成词向量，那么如何转化它呢？以及目前几种主要转化方式是有什么不一样呢？ 接下来借助我最近做的文本分类的问题加以解释说明，主要理清word2vec和keras提供的Embedding层的差别，这也是一度使我比较困惑的地方。 一、one-hot 一切要从one-hot编码开始。我们在对文本进行预处理时，一般会根据自己的训练文档来构...
Kaggle_tweet_emotion_bert
weixin_42681868的博客
04-27 351
import pandas as pd import numpy as np import matplotlib import matplotlib.pyplot as plt import seaborn as sns import os import tqdm import tensorflow as tf print(tf.__version__) import re from tensor...
在Keras的Embedding层中使用预训练的word2vec词向量
和而不流
05-15 2万+
文章目录1 准备工作1.1 什么是词向量?1.2 获取词向量2 转化词向量为keras所需格式2.1 获取所有词语word和词向量2.2 构造“词语-词向量”字典2.3 填充字典和矩阵2.4 在 keras的Embedding层中使用 预训练词向量*2.5 不使用“预训练”而直接生成词向量3 整体代码：在Keras模型中使用预训练的词向量3.1 读取数据3.2 句子分词3.3 *构造词向量字典3....
keras 用LSTM搭建一个语言情感分类器 - keras 进阶教程
蓝一潇的博客
05-27 2463
鉴于很多人最近问我LSTM的API在keras中怎么用，怎么总是报错，我觉得写一个博客，搭建一个最小化的语言情感分类模型，这样大家就能一通百通了。 这个讲解会主要侧重于LSTM中的数据维度处理。 这里就不谈LSTM的原理了，大家肯定没少看到过那个经典的LSTM cell结构图。 那么，第一步做什么呢？ 首先，我们需要一个数据集，用word2vec来做word embeding（即将词转化为向量的形...
PredictSurvival.rar_XGBoost kaggle_kaggle_python_xgboost_xgboost
07-15
本压缩包“PredictSurvival.rar_XGBoost kaggle_kaggle_python_xgboost_xgboost”提供了一个入门级别的Kaggle竞赛案例，主要关注的是存活预测问题。在这个项目中，参赛者使用了Python编程语言，并选择了强大的机器...
kaggle-avazu-rank2.zip_kaggle_kaggle rank_kaggle比赛_rank2_亚马逊
09-21
在本项目中，我们关注的是一个名为 "kaggle-avazu-rank2.zip" 的压缩包文件，它与Kaggle上的一个亚马逊广告效果预估竞赛相关，并且在这个比赛中获得了Rank2的好成绩。Kaggle是全球知名的机器学习和数据分析竞赛平台...
kaggle叶子分类.zip_KAGGLE叶子_kaggle_kaggle 叶子分类_quicklyrn9_一维卷积网络
07-15
利用一维卷积神经网络将叶子进行分类，里面包含的有数据
kaggle_house_pred.zip
05-08
标题 "kaggle_house_pred.zip" 提供的信息表明这是一个与Kaggle竞赛相关的数据集，主要目标可能是预测房价。Kaggle是一个著名的数据科学和机器学习竞赛平台，参与者使用提供的数据来训练模型，并预测未知数据的结果...
Kaggle_tweet_emotion_bert_transformers
weixin_42681868的博客
05-25 1026
import pandas as pd import numpy as np import matplotlib import matplotlib.pyplot as plt import seaborn as sns import os from tqdm import tqdm import re import inspect import tensorflow as tf from tensorflow import keras # import tensorflow.keras.backend
keras中的keras.utils.to_categorical方法
Elvirangel的博客
07-04 1645
to_categorical(y, num_classes=None, dtype='float32') 将整型标签转为onehot。y为int数组，num_classes为标签类别总数，大于max(y)（标签从0开始的）。 返回：如果num_classes=None，返回len(y) * [max(y)+1]（维度，m*n表示m行n列矩阵，下同），否则为len(y) * num_classe...
Keras实现textCNN文本分类
vivian_ll的博客
07-24 1万+
CNN的基本结构包括两层，其一为特征提取层，每个神经元的输入与前一层的局部接受域相连，并提取该局部的特征。一旦该局部特征被提取后，它与其它特征间的位置关系也随之确定下来；其二是特征映射层，网络的每个计算层由多个特征映射组成，每个特征映射是一个平面，平面上所有神经元的权值相等。特征映射结构采用影响函数核小的sigmoid函数作为卷积网络的激活函数，使得特征映射具有位移不变性。此外，由于一个映射面上的...
kaggle入门——NLP with Disaster Tweets
swiee的博客
02-09 1552
Contents数据集 数据集 training set： 数据7613条； id - a unique identifier for each tweet text - the text of the tweet location - the location the tweet was sent from (may be blank) keyword - a particular keywor...
2020-1 推特灾难推送真假
Yansixiliang的博客
02-17 1330
kaggle-nlp competition——Real or Not? NLP with Disaster Tweets. 判断推特内容是否真的在描述灾难。 Getting start tutorial kaggle kernal--0.77505 简述：tweet中包含的单词与disaster有关系。第一步，建立词袋表示法，把每一个tweet都用一个向量表示出来。第二步，由于特定单词的出...
keras句子分类 keras_demo_for_sentence_classification (simplified version)
Shingle_的博客
06-24 1468
''' This script loads pre-trained word embeddings(word2vec embeddings) into a Keras Embedding layer, and uses it to train a text classification model on a customized dataset. '''from __future__ import
kaggle----NLP线性回归---Real or Not? NLP with Disaster Tweets
查数菇乐园
05-06 1354
相关性检测 ridge regression回归模型 F1检验
W2V原理(一）
Rudy95的博客
07-16 7609
W2V
基于Keras的word2vec词向量训练和embeding
weixin_43025542的博客
01-13 1657
这里写自定义目录标题欢迎使用Markdown编辑器新的改变功能快捷键合理的创建标题，有助于目录的生成如何改变文本的样式插入链接与图片如何插入一段漂亮的代码片生成一个适合你的列表创建一个表格设定内容居中、居左、居右SmartyPants创建一个自定义列表如何创建一个注脚注释也是必不可少的KaTeX数学公式新的甘特图功能，丰富你的文章UML 图表FLowchart流程图导出与导入导出导入 欢迎使用Ma...
Kaggle_tweent_tfidf
热门推荐
weixin_42681868的博客
04-16 4万+
import pandas as pd import numpy as np import matplotlib import matplotlib.pyplot as plt import seaborn as sns import os from sklearn.feature_extraction.text import CountVectorizer, TfidfVectorizer,Tf...
No module named 'kaggle_environments'
最新发布
12-07
根据您提供的引用内容，您遇到了一个名为“ImportError”的错误，该错误提示您的Python环境中缺少名为“kaggle_environments”的模块。这通常意味着您需要安装该模块才能在您的代码中使用它。以下是解决此问题的步骤...
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值