Flood prediction based on the fuzzy subset data fusion approach

The development of the Internet of Things has contributed to better predicting natural phenomena using heterogeneous smart objects. The data fusion technique is widely used to manage the massive volume of imperfect data generated by the latter. This paper addresses the prediction of flooding risks problem using the fuzzy subsets data fusion approach to deal with imprecision data and conflict between sensors. First, the supervised area is divided into regions where were deployed sensor nodes able to collect factors impacting the flooding phenomenon namely water level, rainfall rate, and wind speed. Then, the measurements collected are modeled using an intuitionistic trapezoidal fuzzy number to capture their variability. After that, the data is fused locally within each region and then, the obtained results are fused globally. At this step, we have made an improvement of a combination operator to consider the conflicts between information sources. Finally, the decision-making step is used to decide whether or not a team should intervene in the different regions monitored. The results obtained on the generated dataset show that the proposed approach with the improved fusion operator is better than the original and Mean fusion operators.