A Link Mining Algorithm for Earnings Forecast Using Boosting

Abstract:     
The objective of this paper is to present and discuss the results of a link mining algorithm called CorpInterlock that integrates the metrics of an extended corporate interlock (social network of directors and financial analysts) with corporate fundamental variables and analysts' predictions (consensus) in order to forecast the trend of the cumulative abnormal return and earnings surprise using the boosting approach. The rationality behind this approach is that the corporate interlock has a direct effect on future earnings and returns because these variables affect directors and managers' compensation. The financial analysts engage in what the agency theory calls the "earnings game": Managers want to meet the financial forecasts of the analysts and analysts want to increase their compensation or business of the company that they follow.

We found that the basic and extended corporate interlock of the US stock market has the properties of a "small world" network. Based on this, we calculated a group of well-known social network metrics and integrated with economic variables using alternating decision trees (ADTs) implemented with Logitboost. We observed a significant reduction of the test error of the experiments when we used the extended corporate interlock instead of either the basic corporate interlock with fundamental variables and consensus or when only fundamental variables and consensus were included, during a "bull" market (1997-2001). The basic corporate interlock showed to be more effective during a "bear" market (2002-2003).

Keywords: Link mining, link analysis, social network,machine learning,computational finance, boosting, time series,pattern analysis, data mining applications

JEL Classifications: C49, C63, G14

Working Paper Series