A Clustering of Investors' Behavior According to Their Financial, Behavioral, and Demographic Characteristics (an Application of K-means Algorithm)

One of the issues that has a great impact on how people invest is the behavioral characteristics of investors. Given the importance of this issue, investors should be able to categorize investors into different classes and for each class, recommend investment appropriate to the personality type of the same class. One of the solutions that can be used for this purpose is clustering. Clustering is one of the unsupervised learning methods and has a descriptive nature. In this method, the data are allocated based on a similarity criterion so that the data in each cluster are most similar to each other and the least similar to the data in other clusters.

In this study, using K-means clustering and Affinity propagation clustering, we identify a group of investors with similar ability and willingness to accept risk. We also show how to effectively allocate assets using investor characteristics using clustering techniques.

Use silhouette coefficient to evaluate two clustering methods to select the best method for data clustering. The k-means coefficient was equal to 0.17 and the Affinity propagation clustering was equal to 0.097. Therefore, we choose the k-means method as the optimal clustering method. Using the K-means clustering method, we cluster investors based on financial, behavioral and demographic characteristics, and according to the clustering results, we divide individuals into seven categories with low to high risk acceptance.

All calculations in this study were performed by Python 3.8. The results of this study can be used by investment managers and stock advisors.