Prediction of Selected Personality Traits Based on Text Messages from Instant Messenger

Abstract

The aim of the work was to check effectiveness of machine learning to predict personality traits based on technical parameters of text messages. At the beginning, personality traits of test group (based on Big Five model) were determined. Each person provided a collection of text messages from the instant messenger. The authorial analyzer was used to aggregate the technical parameters: (number of emoji used, average message length, number of punctuation characters used, etc.) During the analysis, the most important parameters of text messages were identified, with the help of which it was possible to predict personality traits. In addition, based on the collected data and conducted analysis, a proprietary system to predict personality traits was created. To this end, methods of supervised machine learning were used. Finally, tests were carried out on the implemented solution, its prediction effectiveness was verified, and conclusions were drawn.