Abstract - This study aims at understanding the relation between various factors and a student's mood and how their personality traits affect this relation.Data used for this study was collected through a smartphone app called i-Log, using questionnaires as well as smartphone sensors. A GPS sensor was used in this case. Data was collected over a month period The datasets were cleaned, pre-processed and merged together to form a master dataset. Statistical tests were performed to check if there exists any difference between the mood over weekdays and over weekends and also between students of various personalities. Machine Learning methods such as Random Forest and Gradient Boosting were used to predict mood. Results showed that the most contributing predictors were varying based on personalities of the students and the time of the week. These findings provide insights into how contextual factors, sensor data and personality differences interact to shape a student's mood.
Keywords: Random Forest, Gradient Boosting, Feature Importance, i-Log, GPS, ANOVA, t-test, One-Hot Encoding
Online data collection methods have transformed research methodologies. Through smartphone apps a larger and diverse population can be reached in order to collect data anytime anywhere which has never been possible through traditional methods. One such application is i-Log which researchers at the University of Trento have developed in order to collect data through questionnaires as well as mobile sensors. Our study made use of the data collected through i-Log application over a month period between November 2022 and December 2022. Data was collected by asking student participants contextual questions like where they are at the moment, who they are with, what they were doing at that moment and how they feel (mood). Simultaneously the app sends sensor data such as light sensors, proximity, gyro, gps coordinates and so on. Certain socio-demographic and personality trait information was collected from participants in the beginning such as, which department they study in, what degree they are enrolled in, nationality and region they live in and estimations were also made on where they stand in terms of big five personality traits. For my specific study I made use of the gps sensor data along with other datasets mentioned above. The purpose was to study the mood of respondents, what are the predictors of the mood among students with different personality types and how do these predictors vary between different time diaries.
- Data Cleaning
- Data Pre-processing
- Exploratory Analysis
- Data Visualization
- Hypothesis Testing
- Machine Learning Models
- Feature Importance
- Evaluation
For detailed analysis please refer to the report document published in this repo.
The project aimed to understand the relationship between mood and various predictors in the students of University of Trento. Through Exploratory Data Analysis, several insights were gained on the impact of the environment on mood. For example one insight was that, the law students exhibit relatively lower levels of happiness when they are with their partners. Time of the day, specified as hour, day of the week and level of conscientiousness were found to be some of the most important factors across all groups of students. Additionally the day type was also found to play a significant role in the variation of mood, with distinction made between weekdays and weekends. These findings highlight the importance of considering personality traits as well as contexts in which students experience their mood and provide valuable insights for future research in this area. Use of sensor data has been instrumental in this research as such data have the least chances of error as compared to self reported data. Ultimately, this work has the potential to inform educational policies and strategies aimed at promoting the well-being of university students.