Applicability of various wavelet families in fuzzy classification of access networks' telecommunication lines

摘要

The future of the smart society sets challenges for all types of existing telecommunication networks and links. For ensuring the optimal utilization of these networks precise performance predictions are necessary, especially in case of the symmetrical access networks with rather limited transmission capacity. It is also important to harness the already established infrastructure as long as it is technically possible, so that the use of the environmental resources would be minimal and the economical advantages would be maximal. In performance prediction of telecommunication links the high-dimensional input data, like the insertion loss spectrum, should be compessed. After reducing the dimension of the antecedent set, a fuzzy inference can be carried out for each of the lines. As the number of lines used for building the fuzzy sets is finite and the supports of the fuzzy set do not cover the whole space, a stabilized KH interpolation is used in the decision process. Wavelets constitute the basis of methods for compressing and analyzing data in many fields of science and technology. For the reduction of the input dimension, wavelets proved to be an effective tool. The applicability of various wavelet families with different sizes of filter coefficient sets are tested in the following considerations, with the result, that the wavelet type does not play an essential role as well as the length of the wavelets. Only the deepness of the wavelet transform influences essentially the goodness of the prediction: the remaining number of points should be 4 after the transformation.