This thesis presents a comprehensive study on understanding human behavior as well as predicting personality traits using smartphone sensor data and app usage data. By leveraging Supervised Machine Learning algorithms and extensive data analysis, we aim to model the Big Five personality framework effectively. This work encapsulates data collection, feature extraction, exploratory analysis, and predictive modeling, shedding light on the intricate relationship between smartphone usage patterns and personality traits.
- Python 3.x
- Pandas
- Scikit-learn
- XGBoost
- Matplotlib
- Seaborn
- Plotly
- openai
- Power BI
I extend my gratitude to Dr. Ivano Bison, Dr. Bruno Lepri, and Dr. Fausto Giunchiglia for their guidance and supervision. This research was supported by the University of Trento and the Bruno Kessler Foundation and conducted as part of the requirements for obtaining a Master’s degree in Data Science.
This study has showcased the remarkable potential to discern an individual’s personality traits over the course of several days, all through the subtle analysis of behavioral data gathered from everyday smartphones. This innovative approach introduces a fresh paradigm, utilizing smartphone usage data (smartphone sensor data, app usage data) as the key to unlocking one’s personality, all without the need to delve into app-specific and social media content.
Furthermore, the inclusion of a broader array of sensors and the consideration of specific factors like gender and educational background have significantly enhanced the performance compared to previous findings in the literature. This research holds promise for a wide spectrum of applications, spanning the realms of social sciences and public health, where it can serve as a means to automatically detect potential confounding variables. Additionally, in the field of marketing, it offers a streamlined method for extracting more comprehensive consumer behaviors, while in the service sectors, it can be harnessed to discern preferences for personalized experiences. This work presents an innovative approach of utilizing mobile sensing for psychological research.
Although smartphones currently stand as the most prevalent devices boasting considerable sensing capabilities, it’s worth noting that they are not the sole technology at our disposal for such investigations. A myriad of other devices are gaining increased acceptance, some of which are purposefully engineered for ongoing data collection. For instance, smartwatches and fitness trackers, explicitly designed for the continuous monitoring of sleep patterns and physical activity, are now becoming more commonplace. As technology continues to evolve, this research has the potential to expand its horizons, encompassing a wider array of unobtrusive sensing techniques.
While this work marks a promising initial step and a proof of concept, it underscores the need for extensive further research to expand, validate, and apply these findings across a diverse range of disciplines.
If you utilize this research or data in a downstream work, please consider citing it with:
@misc{Personality_Prediction_2023,
title = {Cracking The Personality Code: A New Frontier In Personality Prediction},
url = {https://github.com/KSwaviman/Cracking-The-Personality-Code-A-Behavioral-Research},
author = {Kumar, Swaviman and Bison, Ivano and Giunchiglia, Fausto and Lepri, Bruno},
date = {2023},
publisher = {GitHub},
journal = {GitHub repository},
}