Neural Network Machine Learning in Financial Application

Neural network is a form of computing systems framework for many different machine learning algorithms to work together and process complex data inputs.  It was inspired by the biological neural networks of animal brains.  There are many types of neural networks, one of the more popular neural networks is the recurrent neural network (RNN).

Within the RNN classes, there is one popular algorithm called Long Short-Term Memory (LSTM).  As the name implies, LSTM will “learn” from the input data and then incorporate both “long-term” and “short-term” information into the memory.  Thus, the predicted outcome comprises both long and short-term information.

LSTM has vast applications in speech recognition, image caption and even time-series analysis.  In this article, we will demonstrate a simple example of using LSTM to forecast airport passenger traffic.  The source code of the generic LSTM written in Python could be found in MachineLearningMastery website (Read more here).    The source code was then modified to analyse and forecast passenger traffic.  The passenger traffic data was taken from MAVCOM (Read more here).

The blue curve in the following graph is the monthly historical data from Jan 2013 to June 2018.  The orange curve is the predicted data using LSTM algorithm, based on historical data upto December 2017.  The predicted outcome is decent and could be used as draft guidance for future planning.

Gif Image Source: https://towardsdatascience.com/animated-rnn-lstm-and-gru-ef124d06cf45