Dendritic Cell Algorithm with Fuzzy Inference System for Input Signal Generation

Dendritic cell algorithm (DCA) is a binary classification system based on the biological danger theory and functioning of human dendritic cells. In its pre-processing and initialization phase, DCA select the most relevant features from input training dataset and assign each selected attribute to a signal category of either as safe, danger or pathogenic associated molecular pattern (PAMP). Then, to get the value of each signal category, signals are represented as an aggregate of real-valued numbers such as average from the assigned features. To compute the average, data transformation is performed through linear normalisation in order to fit them in a uniform scale. If non-linear relationship exists between the selected attributes and the resulting signals, using the average technique to generate a specific signal value may negatively affect the classification results of the DCA. This study proposes an approach for aggregating the assigned features nonlinearly in order to get the value of each signal category using fuzzy inference method. Fuzzy inference system (FIS) is used to compute the value of each input signal from the assigned features using fuzzy rule control and rule base before applying them as input to the DCA. The proposed approached was evaluated and validated using the KDD99 data set widely used in the intrusion detection field. The classification results indicate that, the proposed approach performs better than the classical DCA and its counterparts.