Data mining techniques on satellite images for discovery of risk areas

International audience The high rates of cholera epidemic mortality in less developed countries is a challenge for health fa- cilities to which it is necessary to equip itself with the epidemiological surveillance. To strengthen the capacity of epidemiological surveillance, this paper focuses on remote sensing satellite data processing using data mining methods to discover risk areas of the epidemic disease by connecting the environ- ment, climate and health. These satellite data are combined with field data collected during the same set of periods in order to explain and deduct the causes of the epidemic evolution from one period to another in relation to the environment. The existing technical (algorithms) for processing satellite im- ages are mature and efficient, so the challenge today is to provide the most suitable means allowing the best interpretation of obtained results. For that, we focus on supervised classification algorithm to process a set of satellite images from the same area but on different periods. A novel research method- ology (describing pre-treatment, data mining, and post-treatment) is proposed to ensure suitable means for transforming data, generating information and extracting knowledge. This methodology consists of six phases: (1.A) Acquisition of information from the field about epidemic, (1.B) Satellite data acquisition, (2) Selection and transformation of data (Data derived from images), (3) Remote sensing measurements, (4) Discretization of data, (5) Data treatment, and (6) Interpretation of results. The main contributions of the paper are: to establish the nature of links between the environment and the epidemic, and to highlight those risky environments when the public awareness of the problem and the prevention policies are absolutely necessary for mitigation of the propagation and emergence of the epidemic. This will allow national governments, local authorities and the public health officials to effective management according to risk areas. The case study concerns the knowledge discovery in databases related to risk areas of the cholera epidemic in Mopti region, Mali (West Africa). The results generate from data mining association rules indicate that the level of the Niger River in the wintering periods and some societal factors have an impact on the variation of cholera epidemic rate in Mopti town. More the river level is high, at 66% the rate of contamination is high.