Organizations are facing a rapidly growing amount of both internally generated as well as externally available structured and unstructured data. The sources and types of data are getting more complex and various which makes it difficult to process them by using traditional tools and methods. Most importantly, the processing should often be done in real time since this can be of crucial importance to react timely. Such situation triggers a strong need for developing innovative concepts to collect and analyze these so called Big Data. The main goal of advanced business analytics is to enable an automated information extraction, information aggregation, and information analysis using intelligent systems. The derived insights are used for various fields of application, e.g., e-health and retail. Data may be gatheres from various sources like the World Wide Web (www), sensors, business databases, etc.
“Web Sources” cover log files which are automatically generated by web servers as well as data extracted from social networks. These data are mainly used to understand behavior and needs of certain groups of users, e.g., users of specific internet services. Data being recorded by sensors may comprise images or videos as well as temperature or humidity time series. These data are used, e. g., to track customers and their behavior at the Point of Sale. Business databases contain structured data, e.g., about customers, suppliers or sales. These data are mainly used to optimize internal processes and to allocate enterprise resources. Self-reported data like surveys, questionnaires and polls can consist of various types of structured and unstructured data.
Modern instruments such as swarm intelligence, fuzzy logic, semantic networks as well as text and image mining methods are used to analyze huge volumes of data. Descriptive models enable aggregated views on the analyzed data, revealing specific characteristics. They look into historical data and uncover relationships or characteristics that are further used for developing predictive models.
Predictive models allow forecasting of specific events or behaviors. They are based on the relationships identified by the descriptive models, but go deeper by trying to identify associations of those relationships. For example, based on the purchasing history, a predictive model can answer questions like: “Which of the products will have a high demand at a certain time or date?” or “What customers will most likely become premium customers”, etc.
Decision models not only predict outcomes but also provide guidance for decision makers or analysts on how to react. This in turn contributes to minimization of potential negative effects and maximization of the potential positive effects of a decision.