REPO: Honours-Thesis-Cluster-Analysis-for-Portfolio-Selection

R Code

Honours Thesis: Comparing the Hierarchical Risk Parity Algorithm and Mean-Variance Portfolio Selection

Author: Nina Matthews

Author: Siphesihle Cele

Supervisor: A/Prof. Tim Gebbie

University of Cape Town

############################################################################# #############################################################################

Needed Files per Section:

3. Data Wrangling Section:

Already prepared data: Dataset 1: PT-TAA.RData

Cleaning of Dataset 2: Main File: Dataset2ETFDataWrangling.r

Requires:

• ETF-DATA-2001-2021.xlsx

4. Data Science and function validation:

Main File: HRP - comp and tests -1.R

Requires:

• x_output.csv
• Lopez_data.csv
• corMat.csv
• covMat.csv

5. Compatibility: Algorithms and the Data Section:

Main File: HRP implemented for ETF data.R

Requires:

• DataPostCov2.xlsx

6. Dynamic Backtest Section:

Main File: All_Ports_Rolling_Windows.R

Requires:

• HRP Fn.R
• PT-TAA.RData

7. Static Backtest Section:

Main File: IS_SR vs OS_SR.R

Requires:

• PT-TAA.RData

############################################################################# #############################################################################

Project Summary:

The Hierarchical Risk Parity Algorithm in Comparison To Mean-Variance for Portfolio Selection.

Data collection and processing

1. Sourced from Bloomberg
2. Data cleaning
3. Winsorizing
4. Geometrically compounding returns
5. Obtaining Covariance matrix

Data Science

1. Construction of the HRP algorithm
2. Construction of the Mean-Variance SR Maximising algorithm

Functions for:

1. Equally Weighted Port
2. SR Maximising Port
3. Buy-Hold
4. HRP Port
5. Constant Mix Port

Backtest Simulation

1. Overlapping Rolling window
2. Growing Window
3. Static IS v OOS

Performance measures theory

1. Probabilistic SR
2. Deflated SR
3. Probability of Backtest overfitting

#############################################################################