Linear Regression Approach For Stock Chart Pattern Recognition

Finding stock chart patterns, like channels, triangles, head and shoulders, etc., in high-dimensional stock data is arduous. Technical analysis experts can visualize these patterns in stock candlestick charts based on experience and predict stock price movements. The subjectivity of visualizing these stock patterns by different experts increases the possibility of drawing different variations of the same patterns based on scale and time. Stock market pattern recognition is a complex challenge due to the volatile nature of the stock price. Various machine-learning algorithms can be used to detect stock chart patterns. A linear regression approach for forming these patterns can minimize human errors by automating the process. In this paper, we have implemented linear regression, which is an efficient approach for handling dynamic stock data and drawing a triangle pattern on charts. It has an average accuracy of 98.60% when the top ten stocks of the Information Technology sector and Pharmaceutical sector in terms of market capitalization in the Indian National Stock Exchange are analyzed.

I. Introduction

Today, stock markets are attracting many investors, as calculated investments facilitate financial growth. Since 2019, the number of Demat accounts[1] in India has doubled[2]. Analyzing stock charts is considered to be a prime skill for investors, as it is an important tool for the technical analysis of a stock. These charts conveniently visualize price action by putting the underlying financial instrument’s historical market data on a graph. This graphic depiction makes it easier to identify typical and uncommon price trends. Technical analysis attempts to address questions about price action. These charts contain two axes: vertical and horizontal. The horizontal axis depicts historical periods, and the vertical axis depicts the stock price or trading volume for each period. There are many types of charts that can be incorporated for technical analysis. However, the four most prevalent varieties are line charts, bar charts, point and figure charts, and candlestick charts[3]. Among these four, the candlestick chart is the most popular one[4] because of which we have used it in our implementation.