Page 293 - Contributed Paper Session (CPS) - Volume 2

CPS1863 La Gubu et al.
practitioners. One of the focuses of research in portfolio selection is the
efficiency of optimum portfolio selection time. This is understandable because
the greater the number of securities involved in portfolio selection, the more
likely the portfolio can be formed. The large number of securities involved in
portfolio selection can be tackled by grouping stock data using cluster
analysis. Cluster analysis is a statistical analysis that aims to separate objects
into clusters of similar characteristics.
In the past few years, a lot of research on the portfolios selection has used
cluster analysis. Guan and Jiang (2007) proposed a portfolio selection method
using clustering techniques to optimize portfolio selection. Tola et al. (2008)
proposed a portfolio optimization model based on two clustering techniques,
namely average lingkage and single lingkage clustering. Chen and Huang
(2009) proposed portfolio optimization using clustering techniques and fuzzy
optimization models. In this approach, clustering techniques are used to group
stock data into different groups, then a fuzzy optimization model is applied to
determine the optimal proportion of investment for each cluster. Nanda et al.
(2010) investigated portfolio selection models based on the clustering
algorithm and the Markowitz model. K-means clustering, fuzzy c-means (FCM)
techniques, and self organizing maps (SOM) are used to group stock data into
different clusters. Long et al. (2014) proposed a portfolio selection model
based on the FCM clustering algorithm and multi-objective genetic algorithm
(MOGA). However, none of the above literatures has considered the
robustness of the methods against the possibility of the outlier existence in
the data.
From the literature review, it can be concluded that for the time efficiency
to make the optimum portfolio, there are three steps that must be done. The
first stage is to group stoks into different clusters. The second stage is to
choose stocks that will form the optimum portfolio. The third stage is to
determine the weight of each stock that forms the optimum portfolio. In this
paper, as our new contribution, in the third stage we consider Fast Minimum
Covariance Determinant (FMCD) estimation method for the mean and the
variance of the data.
The remainder of this paper is organized as follows. In Section 2 we give
Methodology. Results are given in Section 3. Finally in Section 4 we give
Discussion and Conclusion.

2. Methodology
Based on the literature review, it can be seen that portfolio selection
problem can be resolved more efficiently by grouping stocks into clusters and
then selecting stocks in clusters to form efficient portfolios. The problem
framework in this study is shown in Figure 1.

282 | I S I W S C 2 0 1 9