Kaggle challenge Axa Driver Telematics Analysis. (https://www.kaggle.com/c/axa-driver-telematics-analysis)
In the era of Big Data and scalable data analytics, we observe a rapidly developing space of both supervised and unsupervised algorithms for various different purposes within the domain of machine learning. Incremental progress in some spaces and rebirth along with signficant improvements for whole families of algorithms like neural networks mark the last few years of research in this area. Those constant improvements of algorithms together with the ubiquity of digital sensors and thus data, allow for application of algorithms in the real world yielding valuable insights for the respective user of those algorithms. Remarkable progress has been made in particular, in the family of deep learning algorithms for different kinds of applications, primarily though, in image recognition as well as natural language processing.
In this work, we investigate the performance of a specific algorithm from the family of deep learning, originally designed for a different purpose, in the domain of unsupervised anomaly detection. Throughout this investigation, we try to provide insights into the usability as well as the suitabiliy of the algorithm for this problem domain, by applying an existing implementation oft the algorithm to a publicly available dataset.