This article presents an original approach for the analysis of context information in ubiquitous environments. Large volumes of heterogeneous data are now collected, such as location, temperature, etc. This "environmental" context may be enriched by data related to users, e.g., their activities or applications. We propose a unified analysis and correlation of all these dimensions of context in order to measure their impact on user activities. Formal Concept Analysis and association rules are used to discover non-trivial relationships between context elements and activities, which, otherwise, could seem independent. Our goal is to make an optimal use of available data in order to understand user behavior and eventually make recommendations. In this paper, we describe our general methodology for context analysis and we illustrate it on an experiment conducted on real data collected by a capture system. Thanks to this methodology, it is possible to identify correlation between context elements and user applications, making possible to recommend such applications for user in similar situations.