Predicting Ozone Peaks : A combined CBR and cell mapping approach

In this paper we present a new approach for predicting ozone peaks when monitoring atmospheric pollution. Our main idea is that atmospheric pollution is closely monitored and over the past 5 to 10 years we have built up fairly extensive databases concerning it. Thus, rather than looking for a physical model in the first instance, maybe we should look for patterns and repeatability in the historical data. The approach presented is a hybrid one based on case based reasoning and cell mapping. Essentially, we forget all about a model in the first instance and simply search historical data to see if we can construct cases. We then extend this idea to include notions from cell mapping to see if we can build a cell map from the cases. The method is applied to real data coming from the Rhône-Alpes region of France.