Predict on Kaggle’s disaster tweets using Keras’ Bidirectional LSTM

A few years ago, when I began my machine learning journey, I studied Python’s machine learning library, SciKit learn, affectionately called sklearn. Sklearn is a great all round library for machine learning because it has many methods that are required to carry out a machine learning project.

As time went on, however, I realised that I would need to become proficient in other Python libraries if I intended to hone my skills. In fact, on one occasion I received a message from someone on Medium asking me if I was familiar with LSTM, which I was not at that time. That message drove home the notion that I needed to increase my proficiency in the field of machine learning.

I therefore decided to study other libraries, and Keras was just one of them. Keras is an open source Python library designed to create neural networks. It is built on top of the Tensorflow library, which provides a computational backend. Keras supports many neural network architectures and training techniques, including convolutional networks (CNNs), recurrent networks (RNNs), and autoencoders. It is a versatile tool for deep learning tasks, such as image recognition, natural language processing, and more.

I decided to try Keras’ Bidirectional Long Short Term Memory (LSTM) model out and see how it performs against…