{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"#Import needed libraries\nfrom nltk.tokenize import word_tokenize, wordpunct_tokenize\nimport itertools\nimport pandas as pd\nimport numpy as np\nfrom nltk.corpus import stopwords\nimport matplotlib.pyplot as plt\nfrom wordcloud import WordCloud\nfrom sklearn.metrics import f1_score\nfrom sklearn.model_selection import train_test_split\nfrom keras.models import Sequential\nfrom keras.layers import Dense,Activation\nfrom keras.layers import Flatten, Dropout, Convolution1D, Bidirectional, LSTM, CuDNNLSTM\nfrom keras.layers.embeddings import Embedding\nfrom sklearn.linear_model import LogisticRegression\nfrom keras.preprocessing.sequence import pad_sequences\nfrom keras.preprocessing.text import Tokenizer","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true},"cell_type":"code","source":"#Import data\nprint('Importing data...')\ndf_train = pd.read_csv(\"../input/train.csv\")\ndf_test = pd.read_csv(\"../input/test.csv\")\n\nprint(\"Train shape : \",df_train.shape)\nprint(\"Test shape : \",df_test.shape)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"0e403baf439f6a90cbb318d00a50b23d17be49be"},"cell_type":"code","source":"#Shuffle data\ndf_train = df_train.sample(frac=1).reset_index(drop=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"5a70e69dfdb08c5e0f41e9b4ab8094655c6a0210"},"cell_type":"code","source":"max_features = 80000\nmax_len = 40","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"410404b901726f2287a482ed686472a2899cf502"},"cell_type":"code","source":"#Tokenize sentences\ntokenizer = Tokenizer(num_words=max_features, filters='#$%&()*+,-./:;<=>@[\\]^_`{|}~',)\n\ntext_train = df_train[\"question_text\"].fillna(\"_na_\").values\ntext_test = df_test[\"question_text\"].fillna(\"_na_\").values\n\ntokenizer.fit_on_texts(list(text_train)+list(text_test))\n\nprint('Tokenizing train...')\ntokenized_text_train = tokenizer.texts_to_sequences(text_train)\nprint('Tokenizing test...')\ntokenized_text_test = tokenizer.texts_to_sequences(text_test)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"2505567ac4e4a3cf884f4b84491c6d0796b8e7fe"},"cell_type":"code","source":"#Pad sentences\nprint('Padding train...')\ntokenized_text_train = pad_sequences(tokenized_text_train, maxlen=max_len)\nprint('Padding test...')\ntokenized_text_test = pad_sequences(tokenized_text_test, maxlen=max_len)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"00e2721ff04570024b070b76e3e862d4ded168c4"},"cell_type":"code","source":"#Split in train and validation\ntrain_x, valid_x, train_y, valid_y = train_test_split(tokenized_text_train, df_train['target'], test_size=0.15, random_state=42)\n\ntest_x = np.array(tokenized_text_test)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"3953371d576ff2acbfa151bfb39917560d35cf58"},"cell_type":"code","source":"#Build LSTM Network model\nmodel_lstm = Sequential()\nmodel_lstm.add(Embedding(max_features, 96, input_length=max_len))\nmodel_lstm.add(Bidirectional(CuDNNLSTM(96)))\nmodel_lstm.add(Dropout(0.2))\nmodel_lstm.add(Dense(4, activation='elu'))\nmodel_lstm.add(Dropout(0.1))\nmodel_lstm.add(Dense(1, activation='sigmoid'))\nmodel_lstm.compile(loss='binary_crossentropy', optimizer='adagrad', metrics=['accuracy'])\n\n\n# Fit the model\nmodel_lstm.fit(train_x, train_y, validation_split= 0.2, epochs=2, batch_size=256)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"e744e12a6562534959488c6e4d35e78cdb14db99"},"cell_type":"code","source":"# Evaluation of the model on validation set\nvalid_pred_prob_lstm = model_lstm.predict(valid_x)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"1717ca7cf097f0ed6cb62af0c03bdc88b58d0d54"},"cell_type":"code","source":"valid_pred_01_lstm = np.where(valid_pred_prob_lstm>0.27,1,0)\nvalid_pred_01_lstm = [int(item) for item in valid_pred_01_lstm]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"3587cc54ac4c90949deef8e3974e6cef51d3e2e6"},"cell_type":"code","source":"#Calculate F1 score on validation data\nf1_quora  = f1_score(valid_y, valid_pred_01_lstm)\nprint('Validation F1 score LSTM:', f1_quora)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"d18d8d6f43443161777ff2580e440f07cdf4c693"},"cell_type":"code","source":"# Final evaluation of the ensemble\ntest_pred_prob_lstm = model_lstm.predict(test_x)\ntest_pred_01_lstm = np.where(test_pred_prob_lstm>0.27,1,0)\ntest_pred_01_lstm = [int(item) for item in test_pred_01_lstm]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"4897152dfc433dde1a3c53071091357747fc3aba"},"cell_type":"code","source":"#Write to dataframe\nsubmit_df = pd.DataFrame({'qid':df_test['qid'].values, 'prediction': test_pred_01_lstm})\nsubmit_df.to_csv(\"submission.csv\", index=False)","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}