Fig. 4

Symbolization using SAX with different parameters (\(\alpha = 6\) and \(\alpha = 3\)). The area under Gaussian distribution is divided into \(\alpha\) equal-sized areas, with each area associated with a unique symbol. When a value in a time series falls into a divided area, the value is converted to a symbol corresponding to the space. Comparison of breakpoints between \(\alpha = 6\) and \(\alpha = 3\) shows that the second biggest breakpoint when \(\alpha = 6\) is equal to the biggest breakpoint when \(\alpha = 3\), the second biggest breakpoint when \(\alpha = 6\) is equal to the biggest breakpoint when \(\alpha = 3\). It means that the data converted to a and b when \(\alpha = 6\) are converted to a when \(\alpha = 3\); the data converted to c and d when \(\alpha = 6\) are converted to b when \(\alpha = 3\); the data converted to e and f when \(\alpha = 6\) are converted to c when \(\alpha = 3\). Therefore, we can easily obtain symbol series when \(\alpha = 3\) by re-using that of \(\alpha = 6\)