Clustering and anomalies of USA stock market volatility index data

Abstract

Actuality. Investing money is an important way to improve the financial condition of both an individual and society as a whole. The problem of understanding financial data and making decisions regarding the investment of money in a certain project at this moment in time is relevant. The object of the study is the process of establishing the dependence of the value of the asset on time. The subject of research is mathematical models of data processing, data clustering and anomaly detection. The purpose of this work is to develop a method for effective investment of money using data processing and analysis methods for CVOE volatility index values in the USA, determination of clusters based on actions with assets, as well as checking the presence of anomalous data. Research results. Official data of volatility index values were selected and prepared for further analysis by removing incomplete sets and further normalization. Clustering of time series was carried out and the array was divided into five homogeneous groups. Clusters determine the ranges of the volatility index, which reflect the different sentiments of investors in the market and encourage appropriate actions with assets: to sell, to wait of index increasing, to buy, to remove money from developing projects and invest in stable ones, to wait of index decreasing. Segmentation of data, application of window function, centroids for segments were determined and signal reconstruction was carried out. Data anomaly points were identified. A comparative analysis was carried out based on the results of constructed initial data, reconstructed data and reconstruction error.