Mining aeronautical data by using visualized driven rules extraction approach

International audienceData Mining aims at researching relevant information from a huge volume of data. It can be automatic thanks to algorithms, or manual, for instance by using visual exploration tools. An algorithm finds an exhaustive set of patterns matching specific measures. But, depending on measures thresholds, the volume of extracted information can be greater than the volume of initial data. The second approach is Visual Data Mining which helps the specialist to focus on specific areas of data that may describe interesting patterns. However it is generally limited by the difficulty to tackle a great number of multi dimensional data. In this paper, we propose both methods, by combining the use of algorithms with manual visual data mining. From a scatter plot visualization, an algorithm generates association rules, depending on the visual variables assignments. Thus they have a direct effect on the construction of the found rules. Then we characterize the visualization with the extracted association rules in order to show the involvement of the data in the rules, and then which data can be used for predictions. We illustrate our method on two databases. The first describes one month French air traffic and the second stems from a FAA database about delays and cancellations causes