top of page

Asia-Pacific Technology Gems: Portfolio Construction with Genetic Algorithms

Writer's picture: Tian Khean NgTian Khean Ng

Data as of 18 December 2024




Introduction

Attached .xlsx is a spreadsheet of 59 Asia-Pacific Technology Gem stocks  from Singapore, India, Malaysia, Indonesia and Thailand. #Hong Kong   stocks did not meet the screening criteria. The screening criteria (Filters) were:

1.       Sector: Information Technology

2.       Return on Common Equity >10%

3.       Return on Invested Capital  >10%

4.       Unlevered Free Cash Flow >0%

5.       EBITDA Margin >10%

Objective

Our objective is to develop a points scoring system that has been optimized by Genetic Algorithms, allowing us to deviate from an equal- weighted portfolio  as well giving  us the decision-making information  for holding  a smaller portfolio instead of including all the 59 stocks.

What are Genetic Algorithms?

Genetic Algorithms (GA) are a form of Artificial Intelligence that mimics Evolution- the process in Nature that over multiple generations improve the fitness of a species. The basic unit is the Chromosome, and the process involves Selection, Crossover and Mutation of Genes.  

Understanding Crossover and Mutation

Crossover biological analogy: Just like how offspring inherit traits from both parents in biology, crossover combines weights from two parent chromosomes. Benefit: Encourages the combination of good traits (weights) from different chromosomes, potentially leading to better-performing offspring.

Mutation Biological Analogy: Mutations introduce random genetic variations.

Benefit: Prevents the population from becoming too similar (converging prematurely), helping the algorithm explore a broader search space for optimal solutions.

How do we begin?

We first designed a Chromosome that has 5 segments, each segment representing one of the 5 Filters above.  We start the process by initializing a population of 100 randomized Chromosomes (that is, each Chromosome is different by having different values in their 5 segments) .

Objective

The GA must have an Objective and a Fitness Function.

The objective of the GA is to find the optimal set of weights for the stock evaluation criteria (Market Cap, ROE, ROIC, Free Cash Flow, and EBITDA Margin) such that the total scores calculated for each stock maximize their variance.

Why Variance? Variance measures how spread out the scores are. By maximizing variance, we ensure that the scoring system effectively differentiates stocks based on their performance across the criteria. This helps highlight the standout performers while minimizing overlaps between similar scores. We started with multiple criteria (e.g., market capitalization, profitability ratios, cash flow) and assigned initial random weights to each.

Using the Genetic Algorithm, we iteratively adjusted these weights to find the combination that resulted in the widest spread (highest variance) in the total scores.

This process mimics natural selection: weights that led to better differentiation were kept and improved over multiple rounds, while others were discarded.

The Fitness Function is as in the image below:



The Constraints are:




The Exchange on which the stock is listed is given. You can see that the score if we used equal weighting (Total Score) is very different from the score generated by GA (Total Score (GA))

How many stocks for your portfolio?



In the chart above, you can see the slope of the GA Score begins to moderate after the 20th stock.

Thus, with the GA-optimized score you may  want to hold 15- 20 stocks in your Asia-Pacific Technology Gems  portfolio instead of the full list of 59 stocks. Since the stocks are from 5 country markets (India, Malaysia, Singapore, Thailand) there is a certain degree of diversification of risk. From the spreadsheet You can also count the number of stocks from each country thus giving you an insight into the Information Technology sector in each country.

8 views0 comments

Recent Posts

See All

Comentarios


bottom of page