Classification of stock value using visibility graph method

Abstract

The Price to Earnings (P/E) ratios are extensively used to assess the value of a company. However, the real-time analysis is affected due to the lag in calculating P/E Ratios. In this work, we investigate the visibility graph (VG) method to analyze the time series of stock price data for predicting stock values. The stock price is mapped to a network by the VG method, which can capture the underlying characteristics of the time series. Appropriate network measures are extracted from the network. These measures (network features) are then used to classify the stock value.We analyzed the stock price data for 330 stocks representing 11 sectors for a total of 10 years from 2010 to 2020 through Yahoo Finance's API interface. Two variants of the VG method, namely, natural visibility graph (NVG) and horizontal visibility graph (HVG), were used to transform the time series into network. The metrics of the network were used as features to classify the stock sector and P/E ratio values. The results show that the VG method can represent the underlying characteristics of the stock price to classify the stock sector and value. The eigenvalues of the stock price network can classify the stock sector with a maximum accuracy of 96%. The eigenvalues of the stock price network can classify the stock value based on the P/E ratio with a maximum accuracy of 79%. The proposed method can be used in real-time to classify a stock value and provides an impetus for applying complex network methods in analyzing financial time series.--Author's abstract