Twitter Data Sets For Disaster Detection And Tracking

Abstract

Nowadays lots of information we can see about disasters either man made or natural, but we don’t have any tracking system which can be tracked or identify the disaster in previous before coming to a particular place. Here I have applied the BERT approach of machine learning algorithm which will classify the datasets after building a model with good accuracy of prediction. Also, we have taken datasets from www.iswsm.com website and divide www.iswsm.com disaster based datasets into training and testing sets. After pre processing and tokenized the data and build a model with the help of BERT and CNN deep learning algorithm and trained this model with less loss of data and optimized those models with ADAM to enhance the accuracy and efficiency of the model after deployment in the real life. Compare both models CNN and BERT the accuracy of BERT is high. This system also can be helpful in the area of medical, tourism and weather forecast to give prediction before anything happening.