Data Science & Machine Learning in Finance (ACCFIN5246)
Instruction
— This is an individual assessment. Answer all questions listed in Section 4: Q.1, Q.2, Q.3, Q.4.
— Submission to be made electronically via the course Moodle page.
— This submission includes a report including brief accounts on the logical steps taken, nu merical answers, diagrams, and comments (each part in Section 4 specifies how answers are required to be presented).
— Submit only the main report (no additional spreadsheet, nor software routine, etc.).
— Clearly number each part in your report and structure the answers to follow the same order in Section 4: Q.1, Q.2, Q.3, Q.4. The project outline is presented across four sections to
describe the implementations, estimation and arrangement of the results.
— Each part, within the assignment’s overall grade, carries a weight described below:
— Results should be reported in a clear format. Avoid reporting numbers in the ‘scientific format’ e.g. 7.2031e-06. All reported numbers should be rounded to two decimal points.
For example, report 0.00 in place of 7.2031e-06.
— The project includes an Appendix (X1) which provides additional data and their descrip tions.
Section 1: Models
Consider the standard and two extended capital asset pricing models characterised below in equa tions (1)-(3), denoted as Model (A), Model (B) and Model (C), rspectively:
Each model aims at interrelating the real excess equity log-return as the outcome variable, on a given asset rt − rf,t where rf,t is the risk-free log-return, to the right-hand-side explanatory vari ables, where:
[1.1] Time (daily frequency) is denoted by t = 1, . . . , T and the window index (consecutive days) is denoted by w = 1, . . . , W.
[1.2] Each x-covariate denotes an additional financial, market-level or macroeconomic variable, described in Appendix (X1).
[1.3] x-covariates are indexed by i and j in Model B and Model C, respectively, where i = 1, . . . , K and K describes the number of available covariates in Appendix (X1), whereas j ∈ F(w)
refers to a sub-selection of all available covariates. The selected covariates included in set F(w) is carried out based on the learning method described in Section (3). Note that the
selection outcome in F(w) is permitted to vary based on the rolling window. [1.4] Specification error terms are denoted by {ϵ
A t
, ϵB t
, ϵC t } across Models A-C, respectively. [1.5] Model constants and CAPM parameters are described by {α A w , α B w , α w } and {βw
A, βw
B, β C w }, re spectively, in Model A, Model B and Model C. The additional x-covariates are parameterised according to ψi,w and θj,w, in Model B and Model C, respectively. The object of interest in each model is the time-varying coefficients described in [1.5]. Each model provides a suggested specification, ultimately aiming at uncovering the true but unobserved α and β.
my wechat:_0206girl
Don't hesitate to contact me