Event Detection Demonstration here

Using deepleaning technique for real-time event detection
- CNN: identify informative data
- LSTM: temporal event detetection (earthquake data is used for demo)
Introduction We designed a Graphic User Interface usingPyQt. Both CNN model and LSTM model are built in Python using Keras deep learning library. Because the deep models can take a long time totrain, we should pre-train them for saving andthen load them again for use. We can use tabtraining CNN or RNN to retrain them. The following interfaces are tabs for setting and some modes of our application such as a configuration tab for keywords, set the language for streaming tweets, set name, and some other informations. Also, the display list of the informative tweet and other is the list of non-informative tweet coming with information about time. Thelast sub-figure is real-time plotting ofaccumulated frequency keyword-related tweets, predicted values, and earthquake candidate underlying time series data.

Requirement

python
tensorflow
keras
pyqt4

Running

Step1: clone this sourcecode to your local repository

git clone https://github.com/quanap5/eventDetection

Step2: Download word-embeddings for CNN model

Download from website: https://nlp.stanford.edu/projects/glove/
This configure for embedding with 200 of dimensionality namelly: glove_twitter_27B_200d.text
Add to folder /embedings/

Step3: Run with GUI

python try2.py (or python demo.try)

Step4: Configuration keywords on Setting tab

fill out the given blank
chose language
limitation with area keywords

Step5: Training CNN

Use tab

Step6: Training RNN

Use tab or eventDetection-master$ python rnn/lstm-anomaly-detection.py