IDENTIFIKASI KEPRIBADIAN MENGGUNAKAN TRANSFER LEARNING MODEL OPENFACE BERBASIS VIDEO

PERSONALITY IDENTIFICATION USING TRANSFER LEARNING MODEL OPENFACE BASED ON VIDEO 

Kepribadian adalah faktor krusial yang berperan dalam menentukan pilihan hidup, jalur karier, performa kerja, kondisi kesehatan, dan preferensi setiap individu. Metode konvensional seperti kuesioner sering kali memakan waktu dan rentan subjektivitas, terutama dalam proses rekrutmen pegawai. Penelitian ini bertujuan mengembangkan sistem prediksi kepribadian otomatis berbasis video. Model menggunakan transfer learning OpenFace yang ditambahkan dengan LSTM untuk ekstraksi fitur wajah dari video, kemudian dioptimasi dengan RMSProp untuk memprediksi lima dimensi Big Five Personality (Openness, Conscientiousness, Extraversion, Agreeableness, dan Neuroticism). Model dari penelitian ini memiliki hasil akurasi sebesar 88% dan system dibangun menggunakan flask. Temuan ini tidak hanya menunjukkan efektivitas pendekatan berbasis computer vision dalam assessment kepribadian, tetapi juga membuka peluang pengembangan alat rekruitmen pegawai berbasis kecerdasan buatan yang lebih efisien dan minim bias, meskipun implementasinya perlu mempertimbangkan tantangan teknis seperti variasi kualitas frame video dan kondisi pencahayaan. 

Personality is a crucial factor that plays a role in determining life choices, career paths, work performance, health conditions, and preferences of each individual. Conventional methods such as questionnaires are often time-consuming and prone to subjectivity, especially in the employee recruitment process. This research aims to develop a video-based automatic personality prediction system. The model uses OpenFace transfer learning augmented with LSTM to extract facial features from videos, then optimized with RMSProp to predict the five dimensions of Big Five Personality (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism). The model from this study has an accuracy result of 88% and the system was built using flask. These findings not only demonstrate the effectiveness of computer vision-based approaches in personality assessment, but also open up opportunities for the development of more efficient and less biased artificial intelligence-based employee recruitment tools, although their implementation needs to consider technical challenges such as variations in video frame quality and lighting conditions.