Processing Measure Uncertainty into Fuzzy Classifier

Machine learning such as data based classification is a diagnosis solution useful to monitor complex systems when designing a model is a long and expensive process. When used for process monitoring the processed data are available thanks to sensors. But in many situations it is hard to get an exact measure from these sensors. Indeed measure is done with a lot of noise that can be caused by the environment, a bad use of the sensor or even the conversion from analogic to numerical measure. In this paper we propose a framework based on a fuzzy logic classifier to model the uncertainty on the data by the use of crisp (non fuzzy) or fuzzy intervals. Our objective is to increase the number of good classification results in the presence of noisy data. The classifier is named LAMDA (Learning Algorithm for Multivariate Data Analysis) and can perform machine learning and clustering on different kind of data like numerical values, symbols or interval values.