Data Mining and Knowledge Engineering

R Subathra ¹

Affiliations
1 Department of Statistics in Government Arts College (Autonomous), Salem-7, IN

Abstract

Any trading strategy which is a combination of technical indicators will have rolling averages as a natural choice. While it is true that the convergence of short and long rolling average is a widely used concept to analyze the pattern of price changes, the methodology of computing the rolling averages has always been a subject of discussion. In this study, the credibility of simple and weighted rolling averages in predicting the price patterns is analyzed using Bayesian classifier and the accuracy of the strategy which is assessed with back testing is used as a proxy to analyze the risk. The empirical analysis is carried out using the five stocks from NIFTY 50 selected at random. The study proposes two strategies which are: 1. SMA-RSI-ATR which is a combination of MACD computed with simple moving averages along with Relative Strength Index (RSI) and Average True Range (ATR). 2. EMA-RSI-ATR which is a combination of MACD computed with Exponential moving averages along with Relative Strength Index(RSI) and Average True Range(ATR) Among these two strategies, the study concludes that the strategy SMA-RSI-ATR is associated with larger returns as compared with EMA-RSI-ATR. But the risk associated with SMA-RSI-ATR is higher than that of EMA-RSI-ATR as revealed by the accuracy of the model.

Keywords

Average True Range (ATR), Confusion Matrix, Momentum Indicator, Moving Average Convergence and Divergence (MACD), Naïve Bayes Classifier, Relative Strength Index (RSI), Sensitivity, Specificity, Trend Indicator, Volatility Indicator.

Full Text