Context-Aware Prediction Techniques with Applications in Mobile Broadband Networks

This thesis describes a number of context-based prediction techniques as a part of business intelligence (BI) solutions. Computerized systems have influenced development and expansion of different domains. Decreasing computer size and increasing computational power allowed developing sensors for different tasks. Continuous sensor performance monitoring at regular intervals resulted that huge data volumes were generated. Persistently stored domain data from sensors allow designing and implementing BI solutions. BI solutions improve efficiency and effectiveness of the domain related operations using historical data analysis, mining, prediction, and other techniques. Historical data analysis showed high context-oriented data potential. The developed prediction techniques implemented standard time series data transformation into context-oriented data format. This thesis investigates prediction technique development and applicability challenges and makes the following contributions.