Abstract
Clustering is an effective tool for diversifying investments, reducing risk and identifying new opportunities. Clustering plays a key role in data analysis, allowing you to group objects with similar characteristics. This study examines the application of clustering to the tasks of forming and optimizing investment portfolios. The paper presents two clustering methods: K-medoids and fuzzy clustering (C-means). K-medoids divides assets into clusters by correlation, and C-means allows assets to belong to several clusters with varying degrees. The article analyzes various clustering methods in the context of the stock market. Similarity measures for stock clustering are compared. The K-means algorithm for clustering companies by time series is presented. Chaotic cartographic clustering is used to analyze the dynamics of the stock market. The TreeGNG algorithm is used to identify stock market sectors. The HRK method has been developed to predict short-term changes in stock prices. The application of data mining methods to predict stock market trends is discussed. The K-means and C-means methods for clustering banking and energy companies are compared. The effectiveness of the SOM-SVR hybrid approach for predicting price dynamics and volatility is demonstrated. C-means combines with artificial neural networks to improve the accuracy of stock market forecasting. The study demonstrates the potential of various clustering methods in the context of compiling and optimizing investment portfolios.