Implementation of Density Based Spatial Clustering of Applications with Noise (DBSCAN) and Multi-objective Co-variance Based Artificial Bee Colony (M-CABC) Algorithm on Portfolio Optimization Problem

In order to prepare planned and unplanned needs in the future, investments must be made. In investing, an investor is faced with the problem of determining the number of assets and the proportion of optimal capital in each asset in building an investment portfolio. This problem is called the portfolio optimization problem. In building a portfolio, diversification is required, i.e. combining different characteristics of assets with the aim of reducing investment risk. Clustering can be used as a diversification strategy. The aim of this study is to investigate asset diversification strategy on portfolio using DBSCAN and to select assets and determine the optimal capital proportion of each asset that builds the portfolio using the metaheuristic algorithm M-CABC. DBSCAN is a density-based clustering algorithm designed to form clusters and find noise in arbitrarily shaped data. M-CABC algorithm is a development of the Artificial Bee Colony (ABC) algorithm by adding statistical covariance concepts to accelerate convergence. The assets used in the study are stocks. The implementation is carried out with three different methods: optimization without DBSCAN, optimization with DBSCAN but without noise, and optimization with DBSCAN including noise. The result shows that the size of the proportions of stocks with negative mean returns had an effect on the selection of methods used to obtain the portfolio which carries the least risk.