Big Data Analytics

Recent advances in sensor technology, internet connectivity, and an ever accumulating amount of data in both production and business-oriented applications have led to an exponential growth of data and data availability. The term Big Data is often used to denote data sets of a size that cannot be stored on a single computer and for which traditional data processing and algorithms are inadequate. Big Data Analytics comprise methods from statistics, applied mathematics, machine learning, high performance computing and computer science such as classification and prediction but also advanced visualization and presentation of results with the objective to detect relations and underlying patterns for turning data into knowledge.

The big data analytics landscape is highly competitive including tailor made on-line solutions with real time processing requirements on streaming data in the car industry to off-line batch analysis of daily, weekly, monthly sales statistics for retail corporations. In the automotive industry, big data analytics challenges include how to feed back information about actual car usage to the development of future models, how to collect, analyze, and visualize a mix of structured and unstructured information related to advanced engineering, and how to classify road conditions or road profiles for improved safety or optimizing fuel consumption based on streaming vehicle-vehicle or vehicle-cloud data.

In the process industry both quality assurance and optimized process control are key ingredients for sustainable operations and a profitable business. Here, big data analytics in combination with traditional modeling, simulation and optimization will provide a leading edge and help making the most out of new sensor technology and traditional process control.

FCC has worked together with Volvo Cars in projects where the objectives have been to identify patterns of car usage and classify typical driver types with the aim to adapt the design of functions and improve customer satisfaction. Within the projects specialized data mining techniques such as clustering and classification on big data have been developed and employed.

Quite a number of commercial Big Data solutions are already on the market but the topic still needs and will profit from new mathematical research. Topics like quantification of uncertainties in the data, robustness of prediction, and computational efficiency really need mathematical input although many people see the field from pure computer science perspective. Our ambition in the big data analytics area is to build knowledge around key algorithmic components such as clustering for classification and regression for prediction as well as means for deploying these solutions to own or customer owned platforms. We strive to be known for our analytics capabilities and deep statistical knowledge providing high quality analysis of customer data.

© 2014-2017  Fraunhofer-Chalmers Centre