Automatic machine learning applied to time series forecasting for novice users in small to medium-sized businesses : a review of how companies accumulate and use data along with an interface for data preparation as well as easy and powerful prediction analysis capable of providing valuable insight

Abstract

Data analytics is gradually becoming one of the most essential tools and sources of competitive advantage for modern companies. There is a multitude of analytical services and solutions on the market, and the effect of data analytics is both well documented and significant. One of the more underutilized aspects of data analytics, especially for small to medium-size businesses, is the making of time series predictions or forecasts based on all available and relevant data. Such companies do not often have their own data scientists and limited resources to invest in learning or developing data analytical competence (Henke et al., 2016). However, there have been great developments in the field of automatic machine learning, making it much easier to create high-quality models without the need for expertly customizing a model to the data. In this thesis, I develop an interface for both data preparation and automatic machine learning that lets novice users apply the full power of H2O AutoML for easy data analytical insight into the unobserved future.