The competition can be found via the link: https://www.kaggle.com/c/quora-question-pairs

• Download pre-trained word vectors googlenews-vectors-negative300.bin which is available here https://drive.google.com/file/d/0B7XkCwpI5KDYNlNUTTlSS21pQmM/edit
• Download pre-trained word vectors glove.840B.300d which is available here https://nlp.stanford.edu/projects/glove/
• Download the train and test data from https://www.kaggle.com/c/quora-question-pairs/data. Create a folder named "data" and put them in.

• Python 3.5
• Xgboost 0.6
• lightGBM 0.2
• numpy 1.11.3
• pandas 0.19.3
• sklearn 0.18.1
• jupyter 1.0.1

• This code is written in Python 3.5 and tested on a machine with Intel E5-1630 V4 processor and Nvidia GeForce GTX 1060. Keras is used with Tensorflow backend and GPU support
• Go to the [feature] and run Feature_1 Feature_2 Feature_3 Feature_4 and Generate the csv files in the feature [train] & the [test]
• Run LSTM and xgboost in [model]
• IF you got want to have a better performance,go to [feature] and run feature_engineering.py

we got the Top4% in the competition ,The code show here may will have a better performance,SO,Try this!