Portfolio optimization (PO) entails deciding which assets to invest in order to maximize overall return while minimizing overall risk at the same time. To successfully mitigate loss risk, the assets within the portfolio need to be diversified among different instruments, industry, etc, so that the risk exposure to any one kind of asset is limited. Another challenge in PO is finding the optimum weight allocation of funds to the assets within Portfolio. Each investor has their unique risk appetite and earning goals, so apportioning of assets within portfolio is needed to balance the risk and reward as per individual investor’s goals. With an increase in the number of assets under a portfolio, the optimal asset weight allocation becomes more complex. It becomes more challenging when there is a vast number of assets to choose from. In recent years, nature inspired algorithms have been considered on a large scale as a first approach to PO in computational finance literature that provides approximate results which could be improved with deterministic techniques.

By incorporating nature-inspired algorithms, researchers have been able to overcome the limitations of traditional optimization techniques and achieve better performance in terms of PO. However, there are several issues and challenges associated with using swarm intelligence (a class of nature inspired algorithmic techniques) for PO. Swarm intelligence does not inherently enforce portfolio diversification. Without explicit constraints, these algorithms may generate portfolios that are concentrated towards similar type of assets. Swarm intelligence also face issues when dealing with large-scale portfolio weight optimization. The computational complexity and optimization time increases with the increasing number of assets within a portfolio. These issues hinder the practical applicability of swarm intelligence algorithms for real world portfolio optimization.

Novelty in this thesis is that we improved two different swarm intelligence algorithms and employed an integrative conjunction of these two algorithms: Ant Brood Clustering (ABC) and Particle Swarm Optimization (PSO) to overcome the issues mentioned above. We improved ABC method with co-integration measure of time series for portfolio diversification which gave clusters of similar stocks based on the co-integration of their closing prices. Subsequently, we fed several diversified portfolios to a novel design of a Set-Based PSO algorithm, synchronous set-based PSO for further stock selection and optimizing the weight allocation of assets within the portfolio.

The primary conclusions drawn from the studies suggest that the resulting portfolio contained diversified stocks and the weights allocated to those stocks minimized the risk while maximizing the returns. The proposed study presents a promising approach for investors who are aiming to build robust and resilient portfolios to potentially adopt these algorithms across diverse market conditions.