Germán G. Creamer

Automated social hierarchy detection through email network analysis

This paper provides a novel algorithm for automatically extracting social hierarchy data from electronic communication behavior. The algorithm is based on data mining user behaviors to automatically analyze and catalog patterns of communications between entities in a email collection to extract social standing. The advantage to such automatic methods is that they extract relevancy between hierarchy levels and are dynamic over time. We illustrate the algorithms over real world data using the Enron corporations email archive. The results show great promise when compared to the corporations work chart and judicial proceeding analyzing the major players.

Learning a board Balanced Scorecard to improve corporate performance

The objective of this paper is to demonstrate how the boosting approach can be used to define a data-driven board Balanced Scorecard (BSC) with applications to S&P 500 companies. Using Adaboost, we can generate alternating decision trees (ADTs) that explain the relationship between corporate governance variables, and firm performance. We also propose an algorithm to build a representative ADT based on cross-validation experiments. The representative ADT selects the most important indicators for the board BSC. As a final result, we propose a partially automated strategic planning system combining Adaboost with the board BSC for board-level or investment decisions.

Automated Trading with Boosting and Expert Weighting

We propose a multi-stock automated trading system that relies on a layered structure consisting of a machine learning algorithm, an online learning utility, and a risk management overlay. Alternating decision tree (ADT), which is implemented with Logitboost, was chosen as the underlying algorithm. One of the strengths of our approach is that the algorithm is able to select the best combination of rules derived from well-known technical analysis indicators and is also able to select the best parameters of the technical indicators. Additionally, the online learning layer combines the output of several ADTs and suggests a short or long position. Finally, the risk management layer can validate the trading signal when it exceeds a specified non-zero threshold and limit the application of our trading strategy when it is not profitable. We test the expert weighting algorithm with data of 100 randomly selected companies of the S&P 500 index during the period 2003–2005. We find that this algorithm generates abnormal returns during the test period. Our experiments show that the boosting approach is able to improve the predictive capacity when indicators are combined and aggregated as a single predictor. Even more, the combination of indicators of different stocks demonstrated to be adequate in order to reduce the use of computational resources, and still maintain an adequate predictive capacity.

Segmentation and Automated Social Hierarchy Detection through Email Network Analysis

We present our work on automatically extracting social hierarchies from electronic communication data. Data mining based on user behavior can be leveraged to analyze and catalog patterns of communications between entities to rank relationships. The advantage is that the analysis can be done in an automatic fashion and can adopt itself to organizational changes over time. We illustrate the algorithms over real world data using the Enron corporations email archive. The results show great promise when compared to the corporations work chart and judicial proceeding analyzing the major players.

A temporal based forensic analysis of electronic communication

Previous work [1] reported on our research in developing a data mining environment for analyzing email communication data. In this paper, we describe our extensions to EMT for applying forensic discovery over temporal email data. The goal is to produce a semi-automatic system to aid in evidence discovery and a host of other applications. We describe our research on profile stability, temporal search and clustering, and new social network dynamic algorithms.

A link mining algorithm for earnings forecast and trading

The objective of this paper is to present and discuss a link mining algorithm called CorpInterlock and its application to the financial domain. This algorithm selects the largest strongly connected component of a social network and ranks its vertices using several indicators of distance and centrality. These indicators are merged with other relevant indicators in order to forecast new variables using a boosting algorithm. We applied the algorithm CorpInterlock to integrate the metrics of an extended corporate interlock (social network of directors and financial analysts) with corporate fundamental variables and analysts’ predictions (consensus). CorpInterlock used these metrics to forecast the trend of the cumulative abnormal return and earnings surprise of S&P 500 companies. 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. Following the CorpInterlock algorithm, we calculated a group of well-known social network metrics and integrated with economic variables using Logitboost. We used the results of the CorpInterlock algorithm to evaluate several trading strategies. We observed an improvement of the Sharpe ratio (risk-adjustment return) when we used “long only” trading strategies with the extended corporate interlock instead of the basic corporate interlock before the regulation Fair Disclosure (FD) was adopted (1998–2001). There was no major difference among the trading strategies after 2001. Additionally, the CorpInterlock algorithm implemented with Logitboost showed a significantly lower test error than when the CorpInterlock algorithm was implemented with logistic regression. We conclude that the CorpInterlock algorithm showed to be an effective forecasting algorithm and supported profitable trading strategies.

Impact of Dynamic Corporate News Networks on Asset Return and Volatility

This paper analyzes the relationship between asset return, volatility and the centrality indicators of a corporate news network conducting a longitudinal network analysis. We build a sequence of daily corporate news network for the period 2005-2011 using companies of the STOXX 50 index as nodes, the weights of the edges are the sum of the number of news items with the same topic by every pair of companies identified by the topic model methodology. The STOXX 50 includes the top 50 European companies by level of capitalization. We performed the Granger causality test and the Brownian distance covariance test of independence among several measures of centrality, return and volatility. We found that the average eigenvector centrality of the corporate news networks at different points of time has an impact on return and volatility of the STOXX 50 index. Likewise, return and volatility of the STOXX 50 index also has an effect on average eigenvector centrality. These results are more significant during the most important period of the recent financial crisis (January 2008-March 2009). So, we observe that there is a dynamic process that affects and is affected by return, volatility, and centrality. The causality tests suggest it is possible to improve the prediction of return and volatility by extracting and analyzing a network based on the common topics of news stories.

A Boosting Approach for Automated Trading

This article describes an algorithm for short-term technical trading. The algorithm was tested in the context of the Penn-Lehman Automated Trading (PLAT) competition. The algorithm is based on three main ideas. The first idea is to use a combination of technical indicators to predict the daily trend of the stock, the combination is optimized using a boosting algorithm. The second idea is to use the constant rebalanced portfolios within the day in order to take advantage of market volatility without increasing risk. The third idea is to use limit orders rather than market orders in order to minimize transaction costs.

Using Adaboost for Equity Investment Scorecards

The objective of this paper is to demonstrate how the boosting approach can be used to define a data-driven board balanced scorecard (BSC) with applications to S&P 500 companies. Using Adaboost, we can generate alternating decision trees (ADTs) that explain the relationship between corporate governance variables, and firm performance. We also propose an algorithm to build a representative ADT based on cross-validation experiments. The representative ADT selects the most important indicators for the board BSC. As a final result, we propose a partially automated strategic planning system combining Adaboost with the board BSC for board-level or investment decisions.

Open regionalism in the Andean Community: a trade flow analysis

This article evaluates the extent to which the establishment of the Andean Free Trade Zone (AFTZ) has led to an improvement of intra-regional trade, as promised by open regionalism, without reducing extra-regional trade. Open regionalism is a dynamic process in which economic agreements serve as intermediate steps towards integration with the world economy. The calculations of ex post income elasticities of import demand show that the establishment of the AFTZ in 1993 led to an increase in Andean Community trade, and not to a contraction of extra-regional trade. The intensity of trade index and the propensity to export intra- and extra-regionally confirm this finding. The article discusses these results in the context of the multilateral trading system and the trade relations between the Andean Community and the rest of the world.