Predicting Negative Emotions Based on Mobile Phone Usage Patterns: An Exploratory Study.

Area of Science:

• Digital health
• Computational psychiatry
• Machine learning applications

Background:

• Early detection of negative emotions is vital for managing depression.
• Mobile technology offers a feasible platform for real-time emotion monitoring and intervention.

Purpose of the Study:

• To adapt machine learning methods for analyzing mobile phone usage patterns.
• To predict concurrent emotional states, specifically negative emotions.

Main Methods:

• An Android app collected emotional states and phone usage data (calls, app use).
• Participants used visual analog scales (VASs) to report depression, anxiety, and stress.
• Machine learning models were trained and validated using feature selection and rank product scoring.

Main Results:

• Data from 28 participants were analyzed, with 30% women, mean age 29.2 years.
• A prediction model combining 2-hour time slots, greedy forward selection, and Naïve Bayes was selected.
• The system achieved an overall predictive accuracy of 86.17% for negative emotions.

Conclusions:

• A system was developed to predict negative emotions from mobile phone usage.
• This system shows promise for ecological momentary intervention (EMI) in depressive disorders.
• Early recognition of negative emotions can facilitate preventive treatment before clinical escalation.