Temporal outlier detection using fuzzy logic and evolutionary computation

Early detection of adverse climatic conditions saves life and prevents harmful health impacts. One of the factors that causes such climatic conditions is the changes in ozone layer. Hyper levels of ozone concentration may cause several lung diseases and respiratory disorders. Many data mining algorithms perform analysis on the given data only after performing discretization. Discretisation involves the process of partitioning the data intervals of the attributes with crisp boundaries which leads to over estimation or under estimation of the boundary values. Fuzzy logic overcomes this problem by considering the natural boundaries of attributes. Genetic algorithm overcomes this crisp boundary problem by considering the entire interval of the attribute and yields optimal boundaries. Temporal data mining involves the process of extracting useful information from temporal database. The proposed method identifies these unusual climatic conditions using fuzzy logic and Genetic algorithm. Our method aims to find out the unexpected parameters(in terms of climatological variables) that causes ozone day/normal day from ozone level detection database which is of quantitative, multidimensional and time series in nature by using temporal outliers. These parameters which causes ozone day/normal day are identified by using Temporal Association Rule mining. Then the temporal association rules are pruned by the unexpectedness measure, Leverage to generate temporal outliers. Finally our proposed approach is validated through a set of experiments which show its efficiency in terms of accuracy.