Analyzing email for prediction of job performance

Abstract

Email is one of the major computer mediated communication tools. Email has been used as a business application, so it contains work related and behavior of users. Even though there are several researches about user behaviors in email have been studied for many years, previous researches focus on reducing email overload by studying the usability of email interfaces based on survey data, or visualizing data. In this research, I propose performance index (PI) factors that indicate job performance in the workplace through an analysis of email-user interaction data. I illustrate our analysis with examples from the Enron Corporation’s email archive. Using binary logistic regression on the data set, I tested the factors I developed. The analysis provides insight into which variables are significant predictors of performance at work. The best predictors were the ratio of email replies sent to total emails sent, the ratio of multiple-recipient emails received to multiple-recipient emails sent, and the ratio of multiple-recipient emails received to total emails received.