Wavelet-based Space Partitioning for Symbolic Time Series Analysis

A Symbolic Representation Approach for the Diagnosis of Broken Rotor Bars in Induction Motors

“…In this work the Maximum Entropy Partitioning (MEP) is adopted [41]. As its name implies, MEP creates symbols in such a way as to maximize the entropy of the created string.…”

“…As its name implies, MEP creates symbols in such a way as to maximize the entropy of the created string. For a discrete time continuous valued signal with w samples and for an alphabet with cardinality  the MEP procedure is as follows [41], [42]: a) sort the samples of the original discrete time continuous valued signal in an ascending order, b) define the partitioning limits starting from the first point of the sorted signal from the first step such as w      data points will lie within each section and c) use the partition obtained in the second step, to transform the original discrete time continuous valued signal to a string of symbols (if the value of a data point lies within or on the lower bound of a partition, it is coded with the symbol associated with that partition).…”

Symbolic time series analysis of ultrasonic data for early detection of fatigue damage

“…This paper has adopted a wavelet-based partitioning approach [14] for construction of symbol sequences from the time series data. The wavelet transform [26] largely alleviates these shortcomings and is particularly effective with noisy data from high-dimensional dynamical systems.…”

“…Furthermore, dealing with high dimensions might lead to spurious results and dimension reduction may lead to loss of vital information. To alleviate these difficulties, this paper has adopted a novel method of wavelet-based partitioning [13,14]. Based on this partitioning, the pertinent information is extracted from time series data sets in the form of probability distributions.…”

Uma metodologia para extração de conhecimento em séries temporais por meio da identificação de motifs e da extração de características

“…Sendo assim, o particionamento não Uma solução para esse problemaé apresentada em (Rajagopalan & Ray, 2005) e consiste em dividir o intervalo em regiões de acordo com a informação que elas representam, istoé, ao invés de considerar as partições como sendo de mesmo tamanho, considerar a probabilidade de cada símbolo ocorrer como sendo a mesma. Nesse método, a máxima entropiaé obtida por meio da partição que produzir uma distribuição mais uniforme dos símbolos do alfabeto.…”