Big data goes big time
From business to science, economics and even finance – big data is everywhere. “With big data, financial analysts can instantly process data on hundreds of stocks, retailers can keep records of what you bought during your last shopping spree, and economists can provide real-time forecasts on the state of the economy,” says Ines Wilms, a researcher at Maastricht University. However, as Wilms points out, before any of this can be done, the right tools are needed. “While big data has been around for years, we are now just starting to understand the importance of having an appropriate statistical toolkit for extracting the information we need,” she adds. With the support of the EU-funded BigTime project, Wilms is working to develop the statistical tools economists need to make more confident decisions using big data.
A better way of dealing with big data problems
Using a combination of statistics, machine learning and econometrics, the project developed statistical learning methods and corresponding software for dealing with big data problems. Specifically, the toolkit is meant to help researchers, data analysts and students make reliable estimations and accurate forecasts, quantify their uncertainty and make inferences on causal effects. “With these methods, a user could, for example, understand how a central bank’s massive information set can be used to forecast a country’s gross domestic product both this and next quarter,” explains Wilms. “The model could also help determine the effects of a central bank’s policy over time.” The software programme, called bigtime, which is now publicly available, was designed to meet the needs of various users. “For expert users, we offer all the computer code as open source and combine it with user-friendly documentation so the user can easily navigate through it,” notes Wilms. “For novice users, we provide web-based applications that guide them through basic time-series concepts and simple use cases.”
A big hit
According to Wilms, bigtime is already a big hit. “The software has been downloaded more than 25 000 times,” she says. “I hope to see these numbers continue to grow as more and more researchers, business analysts and students discover the rich area of big data.” Even with this success, Wilms is nowhere near being done. “This Marie Skłodowska-Curie Action made me realise there is much more to do and discover in the exciting area of causal analysis for high-dimensional time series,” she adds. Currently, Wilms is focusing her work on using statistical learning methods for causal analysis, along with expanding the software toolbox. “I hope to bring our software to a broader audience, outside of academia, to interested industry and business partners,” concludes Wilms. “To this end, we have developed a course on time-series analysis and forecasting for the BlueCourses platform that showcases some of the methods developed during the BigTime project.”
Keywords
BigTime, big data, statistical tools, economists, financial, economy, machine learning, software, open source, time series
