Text and Predictive Analytics; Classification of On-line Customer Opinion Surveys

Abstract

Thanks to computers and other data storage technologies, today we can easily save and rapid access to enormous amounts of data. According to results a report published by Andrew Harbison & Pearse Ryan, the information stored in electronic platform consist of more than 80% from written text (unstructured data). Text analytics can be very useful tool in converting these large amounts of unstructured data into regular structured data and then actionable information. Predictive models are very useful tools in text mining. The results of the models may give important information that can be used in decision making. Feature construction from pre-existing features and feature selection techniques involve creating the predictive models. Research has shown that by applying different feature selection techniques in creating predictive models can increase the model’s accuracy rate. In this thesis we review text and predictive analytical methods, and apply them to a cast satisfaction.