Recent technological advances made a great step towards the possibility of forecasting trends based on the semantic enrichment of unstructured patterns. With the expansion of the Internet new sources of mostly unstructured data constantly arise. However, the field of methodological analysis of these data with the goal of pattern recognition and trend forecasting is still underdeveloped. The results of trend forecasting based on simple searches in search engines are surprising and show that the potential is huge. With the help of Google Trends, researchers showed the correlation between searches by company names and the turnover on the stock exchange; and between the content of the searches and the movements of countries GDP. Moreover, they also developed approaches for determining the expected inflation rate and unemployment, for evaluating the current sales volume at the state level, etc.

On the other hand, a significant progress in the field of analysis of large amounts of unstructured data has contributed to the successful extraction of formal knowledge from

this data. Due to abundance of such data and the absence of adequate methodological support, pattern recognition and trend forecasting is still too demanding of both time and financial terms. At the same time advances in cloud computing, processing large amounts of data (Big Data) and large number of transactions (XTP Extreme Transaction Processing) allows the development of such solutions without building costly data centres. We believe that it is possible to develop an automated model that will not only recognize patterns, but will also be able to use them to forecast trends within a particular domain by leveraging methods of data acquisition, analysis and data sampling from heterogeneous data sources. This will include web site and portal search results, email messages and posts in social networks. The model in an innovative way uses the methods of statistical analysis and business intelligence, especially OLAP (OnLine Analytical Processing) and data mining, supported by cloud computing, Big Data and XTP. Our approach exploits the existing models for obtaining formal knowledge, introduces an innovative consensus-based decision model for pattern recognition using methods of artificial intelligence and an innovative mathematical model for trend forecasting. The proposed common solution can be adapted to particular domains, which can provide greater relevancy and accuracy of forecasts in shorter time with fewer resources as the proposed approach uses domainspecific input data.

The main objectives of the project are to develop the following deliverables, which also serve as the basis for assessing success of the project: (1) common contextual model forpattern recognition, (2) common contextual model for trend forecasting, (3) system prototype to demonstrate the feasibility of the proposed approach in domain-specific environments. The proposed project addresses fundamental challenges of business intelligence on data generated both within organization and in widersociety. Analysis of existing work has shown that the proposed project presents a unique and innovative approach on the world level and builds on lessons from EU/FP7 projects in which both applicants and collaborating partners from EU have participated. For the purpose of development, verification and validation of the proposed model we will leverage the computer cloud at the UL FRI Cloud Computing Centre.