Financial time series_forecasting_svm

36
Tunisia Polytechnic School Data Mining project Presented by Mohamed DHAOUI 3 rd year engineering student (contact@Mohamed - dhaoui.com) Academic Year : 2015 - 2016 Financial time series forecasting using support vector machines

Transcript of Financial time series_forecasting_svm

Page 1: Financial time series_forecasting_svm

Tunisia Polytechnic School

Data Mining project

Presented by

Mohamed DHAOUI

3rd year engineering student

([email protected]) Academic Year : 2015-2016

Financial time series forecasting using support vector machines

Page 2: Financial time series_forecasting_svm

2

Page 3: Financial time series_forecasting_svm

3

Page 4: Financial time series_forecasting_svm

4

Page 5: Financial time series_forecasting_svm

5

Page 6: Financial time series_forecasting_svm

6

Page 7: Financial time series_forecasting_svm

7

Page 8: Financial time series_forecasting_svm

8

Page 9: Financial time series_forecasting_svm

9

Page 10: Financial time series_forecasting_svm

10

Page 11: Financial time series_forecasting_svm

11

Page 12: Financial time series_forecasting_svm

12

Page 13: Financial time series_forecasting_svm

13

Page 14: Financial time series_forecasting_svm

a weight parameter, which needs to be carefully set

14

Page 15: Financial time series_forecasting_svm

15

Page 16: Financial time series_forecasting_svm

16

Page 17: Financial time series_forecasting_svm

17

Page 18: Financial time series_forecasting_svm

18

Page 19: Financial time series_forecasting_svm

19

Page 20: Financial time series_forecasting_svm

20

Page 21: Financial time series_forecasting_svm

21

Page 22: Financial time series_forecasting_svm

22

Page 23: Financial time series_forecasting_svm

23

Page 24: Financial time series_forecasting_svm

24

Page 25: Financial time series_forecasting_svm

25

Page 26: Financial time series_forecasting_svm

26

Page 27: Financial time series_forecasting_svm

27

Page 28: Financial time series_forecasting_svm

Backpropagation, an abbreviation for "backward propagation of errors", is a common method of training artificial neural networks used in conjunction with an optimasition method such as gradient descent. The method calculates the gradient of a loss function with respect to all the weights in the network and try to update these weights,

28

Page 29: Financial time series_forecasting_svm

29

Page 30: Financial time series_forecasting_svm

Algorithm:

initialize network weight (randomly)

Do

forEach training example ex

prediction = neural-net-output(network, ex)

actual = teacher-output(ex)

compute error (prediction - actual) at the output units

compute for all weights

update network weights

until all examples classified correctly or another stopping criterion satisfied

return the network

30

Page 31: Financial time series_forecasting_svm

Weights updating

Δwt = -e* E + α Δwt-1

=H* δ0

e: learning rate

α :momentun

Wh,o

Hidden layer

Output layer

O

HE= actual-ideal

δ0= -E*f’(o)

δk= f’(h)*Wh,o *δ0

31

Page 32: Financial time series_forecasting_svm

Weaknesses

• Gradient descent with backpropagation is notguaranteed to find the global minimum.

• There is no rule for selecting the bestlearning rate and the momentum.

• Slow algorithm that need a computationalresources.

32

Page 33: Financial time series_forecasting_svm

SVM perfermance

• Too small value for C caused underfit the training data while too large a value of C caused overfit the training data

33

Page 34: Financial time series_forecasting_svm

the best prediction performance of the holdout data is recorded when delta is 25 and C is 78

SVM perfermance

34

Page 35: Financial time series_forecasting_svm

BP perfermance

• The best prediction performance for the holdout data is produced when the number of hidden processing elements are 24 and the stopping criteria is 146 400 epochs.

• The prediction performance of the holdout data is 54.7332% and that of the training data is 58.5217%.

35

Page 36: Financial time series_forecasting_svm

Comparison

SVM outperforms BPN and CBR by 3.0981% and 5.852% for the holdout data, respectively

For the training data, SVM has higher prediction accuracy than BPN by 6.2309%

SVM performs better than CBR at 5% statistical significance level

SVM does not significantly outperform BP BP and CBR do not significantly outperform each

other

36