Forecasting Multivariate Time Series Data Using Neural Networks
Abstract
Over the last few years, neural networks have become extremely popular, and their usageis increasing rapidly. This project has investigated the use of neural networks for one-steptime series forecasting on highly random data. Multi-layer perceptron (MLP), convolutionalneural networks (CNN), recurrent neural networks (RNN), and long short-termmemory (LSTM) cells are tested to see if they can give a binary classification accuracyabove 50% using this data. The assignment focuses on designing a small embedded neuralnetwork with low latency.
The different neural network architectures are built using a deep learning library inPython, called Keras. This is a high-level software framework, built on top of either Tensorflow or Theano, for fast and easy prototyping of neural networks.
The conclusion of the study is that only the CNN satisfied the requirements of theassignment during the work of this thesis. None of the other architectures showed signof learning generalized patterns and structures from the dataset in question. The CNNshowed the most promising results, being able to extract information about the trainingset that increased the classification accuracy of the test. This leads the way for furtherdevelopment and an eventual hardware implementation of the inference phase reducingthe run-time latency.