2026-04-17T17:53:52.749Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:1:8485 2022-02-14T19:04:04Z 1

Apparent personality analysis based on robust estimation 2021 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/1/8485 https://conferences.ulbsibiu.ro/icdd/2021/files/Proceedings_ICDD2021.pdf M. Vukojičić M. Veinović Apparent trait prediction is one of the hardest tasks in the domain of computer science. Multimodal approaches where the traits are extracted from various sources are giving the best prediction when we compare results to the Revised NEO Personality Inventory (NEO-PI-R model). Multimodal apparent trait personality based on text, images, audio, and handwriting can give us a wide range of estimation for each of the Big Five properties (extraversion, agreeableness, openness to experience, conscientiousness, neuroticism). This problem can be solved by using the aggregation function at the end of the model. The problem that this paper tries to overcome is to reduce the influence of the outliers on the final prediction of the model, by introducing nonlinearity to the final prediction of the model. Aggregation functions such as Max and Min are linear functions and they are creating outliers at the end of the model. We will introduce the nonlinear Huber function to get better results from the outliers and minor deviations from the output of NEO-PI-R model conferenceObject 145 154 Proceedings ICDD