Predicting Dividend Omission Behaviour of Indian Firms Using Machine Learning Algorithms

Abstract

The life-cycle theory of dividends suggests that dividend omissions may indicate significant strategic changes in the firm's life cycle. Such behaviors at the same time have implications for investor perception as dividend omissions may signal weak operating performance or financial distress situation. A firm's preference for dividend payments relative to omitting dividend payments is also used to cater to investor time-varying preferences. This paper aims to test the prediction models of dividend omission behaviors of firms in India. The financial data of 12942 farmwear observations from 2013 to 2018 Indica the 55 percent dividend omissions. The paper uses five classes of machine learning algorithms to predict these behaviors. The multi-layer perceptron (MLP)AN approach using the R Prop algorithm achieves a predictive accuracy of 82.36 percent with an ROC (area under the curve) of 0. 901.The feat ture set relating to the financial parameters of a firm contribute to the prediction accuracy.