LSTM model using text data

Deep learning example code with text data




# Searching letters
df['column'][df['column'].str.contains('[^a-z ]')].sum()

# Replace letters
df['column'] = df['column'].str.replace('[^a-z ]','')

# delete sentence's head and tail's empty word
df['column'] = df['column'].str.strip()

# search duplicate data
df['column'].duplicated().sum()

# drop duplicate data
df.drop_duplicates(subset=['column'], inplace=True)

# Graph using value_counts()
df['column'].value_counts().plot(kind='bar')

# Split X and Y
features = df['column'].values
labels = df['column'].values

# TF-IDF
from sklearn.feature_extraction.text import TfidfVectorizer
tfidf = TfidfVectorizer()
x_train_v = tfidf.fit_transform(x_train)
x_test_v = tfidf.transform(x_test)

# Tokenizing
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
tokenizer = Tokenizer()
tokenizer.fit_on_texts(x_train)

# word:number mapping
tokenizer.index_word

# total word count
len(tokenizer.index_word)

# change sentence to array
x_train_seq = tokenizer.texts_to_sequences(x_train)
x_test_seq = tokenizer.texts_to_sequences(x_test)

# sentence's maximun lengh
max(len(line) for line in x_train_seq)

# Set all sentence length to maximun length
x_train_pad = pad_sequences(x_train_seq, maxlen=38)
x_test_pad = pad_sequences(x_test_seq, maxlen=38)

# HyperParameter
max_words = 47646 + 1    # add padding
max_len = 38~~ 
embedding_dim = 32~~

# model's first layer embedding 
model.add(Embedding(max_words, embedding_dim, input_length=max_len))
# LSTM layer (after 1st input layer)
model.add(LSTM(16, return_sequences=True))
model.add(LSTM(16, return_sequences=True))
model.add(Flatten())
model.add(Dense(128, activation='swish'))
model.add(Dense(32, activation='swish'))
model.add(Dense(6, activation='softmax'))

# Model Compile 
model.compile (loss = 'sparse_categorical_crossentropy',
              optimizer = 'adam',
              metrics = 'accuracy')
model.summary()

# EarlyStopping 
es = EarlyStopping(monitor='val_loss', patience=10, verbose=1)

# Set ModelCheckpoint
checkpoint_path = 'tmp_checkpoint.ckpt'
cp = ModelCheckpoint(checkpoint_path, monitor='val_loss', verbose=1, save_best_only=True)

# model fit
history = model.fit(x_train_pad, y_train, epochs=50, batch_size=512,
                      validation_split=0.2, verbose =1, callbacks=[es, cp])

# model evaluate using test dataset
model.evaluate(x_test_pad, y_test)

# Predict using model
predict = model.predict(x_test_pad[:1])

Comments

Post a Comment

Popular posts from this blog

Voting Ensemble Model

By using the [voting ensemble model], we can learn multiple models respectively and compare the results. from sklearn.linear_model import LinearRegression as lr from xgboost import XGBRegressor as xgb from sklearn.ensemble import RandomForestRegressor as rfr from sklearn.ensemble import GradientBoostingRegressor as grb from sklearn.ensemble import VotingRegressor import joblib import time model_list=[lr(), rfr(), grb(), xgb()] # Convert mutli dimension array to single dimension array. train_y = train_y.to_numpy().flatten() # Save several models after learning sequence. model_result = [] for i in range(len(model_list)): model = model_list[i] model.fit(train_x, train_y) pred_y = model.predict(test_x) model_result.append(model) # Models list voting_models = [ ('linear_reg', model_rslt[0]), ('randForest', model_rslt[1]), ('gradBoost', model_rslt[2]), ('xgboost', model_rslt[3]) ] # Run VotingRegressor voting_
Read more

CNN

Load image files and make the Convolutional Deep learnig model. images = os.listdir(~~~path) # image load and print img = Image.open("PATH AND FILE").resize((?, ?)) plt.title('NAME') plt.imshow(img) plt.show() # image to array and scaling img = Image.open ("PATH AND FILE").resize((?, ?)) img = np.array(img)/255. # after work, chage array again list_arr = np.array(img_list) # handle as x, y # or img2 = image.img_to_array(img_origin) img2 = img2.reshape((-1, ?, ?, 3)) img2 = preprocess_input(img2) features = model.predict(img2) #print(decode_predictions(features, top=3)) # Conv Model from tensorflow.keras.models import Sequential from tensorflow.keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPooling2D model = Sequential() model.add(Conv2D(32, kernel_size=(5,5), strides=(1,1), padding='same', activation='relu', input_shape=(?,)))) model.add(MaxPooling2D(pool_size=(2,2), strides=(2,2))) model.add(Dropout(0.2)) model.add(Flat
Read more

Graph

Several modules that will work with Python offer a variety of functions. Among them, I drew and compared graphs using the matplotlib library that helps me draw graphs efficiently and Pandas' own plot function. # default bar graph df['column'].value_counts().plot(kind='bar') # sub graph plt.subplot(100) df['column1'].plot(kind='hist') plt.subplot(150) plt.hist(df['column2']) plt.show() # seaborn histogram graph sns.histplot(data=df, x='column') # seaborn histogram graph with hue color sns.histplot(data=df, x='column', hue='column color') # seaborn histogram with round end sns.kdeplot(data=df, x='column', hue='column color')
Read more