LSTM-based network churn classification from EDA phasic data

Abstract

Understanding television watching behavior of consumers can be useful in many contexts, such as evaluating the influence of a TV network, building recommendation systems, or providing insights regarding commercials for advertisers. Electrodermal activity (EDA) is a psychophysiological indicator of emotional arousal and attention that reflects the variation of the electrical properties of the skin. Given that it is a measure that reflects the emotional status of consumers and has advantages over self-report of emotions, it has been widely used in consumer research studies. In this study, we built a classification model using long-short term memory networks and EDA phasic signals to classify network switch/churn occurrence. The developed model had an accuracy of 71%, which demonstrates that EDA phasic activity is a good candidate to predict channel churn occurrence.