Professional Documents
Culture Documents
Time Series Via Fractals Geometry
Time Series Via Fractals Geometry
FRACTALS
GEOMETRY
K.BHAVANI(MA14M003)
R.HEMA(MA14M012)
Why?
Where?
How?
The most direct link between chaos theory and the real world is the
analysis of time series of real systems in terms of nonlinear
dynamics. Sometimes time series are so complex that they could not
be examined by traditional methods. In such cases time series data is
analysed via fractals.
Where?
Above figure defines the shape and size of the P-QRS-T wave of the
electrocardiogram(ECG), this is essential to determine the classification
of heart failure from ECG. The characteristics of all interval (PR, QRS,
QT, RR) can change by specific heart failures. ECG signals are widely
examined signals, thinking the significance in humans life.
Decisions about ECG signals were made through handling the intervals
between P-QRS-T waves..
How
1.Correlation method
2.Hausdorff method
3.Box counting method
Fractal dimension:
Correlation method:
Autocorrelation:
When there is no correlation in the time series then the expectation value
of correlation function is zero. Autocorrelation function is computed by
the formula.
A()=-)(-)/-)^2
represents ith element of the series, is time delay (Optimal time lag)
and N is equal to length of the series.
Quantities varying with time often turn out to have fractal graphs. One
way in which their fractal nature is often manifested is by a power-law
behaviour of correlation between measurements separated by time h.
C(h) = ) )dt
Hausdorff method:
The
Steps for estimating the Hurst exponent after breaking the time series into
chunks:
The cumulative deviation of the series from the mean by summing up the mean
centered values,
The Range (R), which is the difference between the maximum value of the
cumulative deviation and the minimum value of the cumulative deviation,
Finally, average the rescaled range over all the chunks. i.e. H
DH= -H
where H- hurst exponent(0<H<1)
For the ECG time series graph given above the different size of the boxes
the no. of boxes tabulated below
data given above in the table will able to plot a regression line.
The
Taking
=
There
is a facility to find the optimal box number to calculate the boxcounting dimension.
simple linear regression, Least Square Method, which are an useful and
popular technique in the field of statistics, to find the box counting
dimension via the slope of the regression line.
where a and b are unknown coefficients while all and are given. To
obtain the least square error, the unknown coefficients a and b must yield
zero first derivatives.
Linear ()
3
Ln of box size
Box dimension:
SLOPE=1.8314
R
N(r)
234550
59010 14805
16
32
64
128
63
20
Chart Title
14
12
6
4
2
0
0
4
LN(R)
Linear ()
CONCLUSION:
FRACTAL DIMENSION TYPE
DIMENSION VALUE
CORRELATION
1.002
HAUSDORFF
1.008171
BOX
1.8314
SINUSOIDAL WAVE
Applications:
Correlation dimension:
Time lag =6
Correlation dimension
1.924 where embedded
Dimension is 2.
Hausdorff dimension
H value is 0.995760
Hausdorff dim=2-H=1.004240.
Box dimension:
THANK YOU