ECOM193 Statistical Machine Learning in Finance
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
May Examination Period 2021–22
ECOM193 Statistical Machine Learning in Finance
When writing formulas, please note the following:
· It is acceptable to use the standard alphabet in place of Greek letters. The following are
recommended: a for ↵ , b for β, d for 6, D for ∆ , l (lowercase l) for λ , m for µ , v for v, s for σ , S for ⌃ .
· Use + for addition, - for subtraction, * for multiplication and / for division.
· Where appropriate use an underscore to indicate a subscript, e.g. x i for xi .
· Use the ˆ character for power, e.g. x^2 for x2 , x^0.5 for ^x.
· When referring to the following functions use log(x) or ln(x) for loge x, logb(x) for logb x, exp(x) for e北 , cos(x) for cos x.
· Use infty for 1.
· Use Sum to denote summation of terms, e.g. Sum i=1^n x i for P xi .
· Use Prod to denote product of terms, e.g. Prod i=1^n x i for Q xi .
· Use D for derivative, e.g. D(x^2) = 2x.
· Use Int for integral, e.g. Int a^b (x) dx for Ra(b) xdx.
· Use cap for \ and cup for n when referring to sets.
· Where it is not obvious that an estimate is implied then state this in full, e.g. ‘a suitable
estimate of b is 0.125’ or more simply (and equally acceptable) ‘est.b = 0.125’ .
· Use brackets as necessary. To make your answer clearer use diferent types of bracket pairs where
appropriate, e.g. (), [], {}.
Use obvious choices for any other mathematical symbols not listed above that you may require.
Question 1
a) Explain how the bootstrap procedure may be used with a suitable sample to estimate the bias and the standard error of a statistical estimator. [12 marks]
b) The following ordinal credit score data were obtained on eight individuals.
589 845 701 842 599 913 749 845
A bootstrap analysis of the data was thought to be sensible and the following four bootstrap samples were obtained from the above sample in the usual way.
Bootstrap Sample
1 2 3 4
589 589 589 599
589 589 589 599
589 599 701 599
599 749 749 701
842 749 842 749
864 842 864 842
864 845 864 842
864 845 913 864
From the data in the table above:
i. Calculate the bootstrap estimate of the mean.
ii. Calculate the bootstrap estimate of the median.
iii. What is the bootstrap estimate of the bias of the mean? What do you conclude?
iv. What is the bootstrap estimate of the bias of the median? How does this compare with your preceding answer for the bias of the mean?
v. What is the bootstrap estimate of the standard error of the mean? What do you conclude about this estimate? [13 marks]
Question 2
a) What is an ensemble method in machine learning? [5 marks]
b) What are their primary advantages of ensemble methods compared with more classical data modelling approaches? [6 marks]
c) Describe one ensemble method and its general procedural implementation. (You do not need to derive the particular method you choose from first principles but you must give a clear description of the method.) [8 marks]
d) Why are out- of- bag samples relevant in an ensemble method context? [6 marks]
Question 3
a) Describe the method of Principal Component Analysis (PCA) its general aims and advantages. [10 marks]
b) Why is it advisable to ensure the input variables to PCA are generally measured on comparable scales? What would you do if this wasn’t the case? [4 marks]
c) The following correlation matrix is estimated from the daily returns of five currency pairs for the year 2020:
USDAUD USDBRL USDCOP USDMXN USDMYR
USDAUD 1 .0000 0 .3429 0 .4760 0 .5764 0 .3251
USDBRL 0 .3429 1 .0000 0 .3993 0 .5502 0 .1394
USDCOP 0 .4760 0 .3993 1 .0000 0 .5316 0 .2980
USDMXN 0 .5764 0 .5502 0 .5316 1 .0000 0 .3316
USDMYR 0 .3251 0 .1394 0 .2980 0 .3316 1 .0000
(USD base currency).
a PCA analysis in R using this correlation matrix gives
Loadings (Principal Components):
USDAUD
USDBRL
USDCOP
USDMXN
USDMYR
Comp .1
0 .474
0 .418
0 .470
0 .526
0 .322
Comp .2
0 .103
-0 .546
-0 .135
0 .820
Comp .3
0 .567
-0 .598
0 .323
-0 .464
Comp .4 Comp .5
0 .530 0 .403
0 .407
-0 .818
0 .215 -0 .811
(small values are automatically suppressed)
Importance of components:
Comp .1 Comp .2 Comp .3 Comp .4 Comp .5
Standard deviation 1 .6233 0 .9397 0 .7768 0 .7268 0 .5917
Proportion of Variance 0 .5270 0 .1766 0 .1207 0 .1057 0 .0700
Cumulative Proportion 0 .5270 0 .7036 0 .8243 0 .9300 1 .0000
i. What do you observe about the correlation structure?
ii. What do you deduce from the components or loadings themselves?
ii. How many components might you choose to describe the data? Justify your answer.
iii. Explain how PCA scores can be obtained from the above analysis.
[11 marks]
Question 4
a) Describe what is meant by a generalized linear model (GLM). [12 marks]
b) Why is such a framework of use? [6 marks]
c) Briefly outline how one might estimate the parameters of a GLM. [4 marks]
d) Name three probability distributions for the response variable that the GLM framework can handle. [3 marks]
Question 5
a) What is a maximal margin classifier and why is it a largely theoretical starting point for support vector machines? [4 marks]
b) Explain some of the advantages a support vector machine has over linear discriminant analysis. Where might linear discriminant analysis do better? [6 marks]
c) Explain what is meant by a kernel in a support vector machine context. [3 marks]
d) What role does a cost function play in fitting support vector machines? In what form is this incorporated into the model? [5 marks]
e) Give three real world examples where you might apply a support vector machine. [3 marks]
f) You decide to fit a support vector machine using a polynomial kernel. How might you go about choosing its degree? What would you guard against? [4 marks]
Question 6
a) What is multiple logistic regression? [6 marks]
b) Explain how neural networks can be thought of as a flexible non-linear extension of multiple logistic regression. [6 marks]
c) What is the role of an activation function in a neural network? [3 marks]
d) Describe and compare two common neural network activation functions. [5 marks]
e) List the main advantages and disadvantages of neural network models compared with more classical parametrized statistical models. [5 marks]
2023-02-09