Abstract:The adverse effects of loneliness on both physical and mental well-being are profound. Although previous research has utilized mobile sensing techniques to detect mental health issues, few studies have utilized state-of-the-art wearable devices to forecast loneliness and estimate the physiological manifestations of loneliness and its predictive nature. The primary objective of this study is to examine the feasibility of forecasting loneliness by employing wearable devices, such as smart rings and watches, to monitor early physiological indicators of loneliness. Furthermore, smartphones are employed to capture initial behavioral signs of loneliness. To accomplish this, we employed personalized machine learning techniques, leveraging a comprehensive dataset comprising physiological and behavioral information obtained during our study involving the monitoring of college students. Through the development of personalized models, we achieved a notable accuracy of 0.82 and an F-1 score of 0.82 in forecasting loneliness levels seven days in advance. Additionally, the application of Shapley values facilitated model explainability. The wealth of data provided by this study, coupled with the forecasting methodology employed, possesses the potential to augment interventions and facilitate the early identification of loneliness within populations at risk.

What problem does this paper attempt to address?

The problem this paper attempts to address is the prediction of loneliness in daily environments through multimodal wearable devices and mobile sensing technology. Specifically, the main goal of the study is to explore the feasibility of using wearable devices such as smart rings and watches to monitor early physiological indicators, and smartphones to capture behavioral cues to predict loneliness. The study aims to develop personalized machine learning models to accurately predict loneliness levels for the next 7 days and to enhance the interpretability of the models through Shapley values. ### Main Issues: 1. **Prediction of Loneliness**: Most current research focuses on real-time detection of loneliness or offline analysis, lacking studies on long-term prediction of loneliness using time-series data. This paper attempts to fill this gap by proposing a method for predicting loneliness based on multimodal time-series data. 2. **Integration of Physiological and Behavioral Data**: The study explores the relationship between loneliness and data by integrating physiological (e.g., heart rate, heart rate variability) and behavioral (e.g., smartphone usage patterns) data. 3. **Development of Personalized Models**: The study develops personalized machine learning models to improve prediction accuracy. These models can make more precise predictions based on individual-specific conditions. ### Research Background: - **Impact of Loneliness**: Loneliness has significant negative impacts on physical and mental health, including increased morbidity and mortality rates. - **Limitations of Existing Research**: Existing research mainly focuses on real-time detection of loneliness, lacking long-term prediction of loneliness. - **Capabilities of Modern Wearable Devices**: Modern wearable devices (e.g., Oura Ring and Samsung smartwatches) and smartphones can collect a large amount of physiological, behavioral, and contextual information, providing new possibilities for predicting loneliness. ### Research Methods: - **Data Collection**: The study collected self-reported loneliness, physiological data, and behavioral data from 29 full-time university students through the ZotCare platform. - **Feature Extraction**: Physiological, behavioral, social, and contextual features were extracted from wearable devices and smartphones. - **Feature Alignment**: Features of different resolutions were aligned to the corresponding loneliness labels. - **Missing Data Handling**: Single imputation methods were used to handle missing data. - **Classification Model**: A random forest model was used for binary classification tasks to predict future loneliness. - **Model Interpretation**: Shapley values were calculated to interpret the model's prediction results. ### Research Results: - **Prediction Performance**: The model achieved high accuracy (0.823) and F-1 score (0.820) in predicting loneliness for the next 7 days. - **Feature Importance**: Sleep disturbance index and activity balance score were the most important features affecting loneliness prediction. ### Conclusion: This study proposes a method for predicting loneliness based on multimodal time-series data, achieving high prediction accuracy through personalized machine learning models. The study also enhances the interpretability of the models through Shapley values, providing new possibilities for clinical applications. Future research can further explore the use of deep learning models to improve prediction performance.

Loneliness Forecasting Using Multi-modal Wearable and Mobile Sensing in Everyday Settings

Objective monitoring of loneliness levels using smart devices: A multi-device approach for mental health applications

Identifying Behavioral Phenotypes of Loneliness and Social Isolation with Passive Sensing: Statistical Analysis, Data Mining and Machine Learning of Smartphone and Fitbit Data

Evolving AI for Wellness: Dynamic and Personalized Real-time Loneliness Detection Using Passive Sensing

Improving prediction of real-time loneliness and companionship type using geosocial features of personal smartphone data

Managing Perceived Loneliness and Social-Isolation Levels for Older Adults: A Survey with Focus on Wearables-Based Solutions

Objective Prediction of Tomorrow's Affect Using Multi-Modal Physiological Data and Personal Chronicles: A Study of Monitoring College Student Well-being in 2020

Predicting Symptoms of Depression and Anxiety Using Smartphone and Wearable Data

Unmasking the Nuances of Loneliness: Using Digital Biomarkers to Understand Social and Emotional Loneliness in College Students

Objective Prediction of Next-Day’s Affect Using Multimodal Physiological and Behavioral Data: Algorithm Development and Validation Study

A Forecasting Framework for Predicting Perceived Fatigue: Using Time Series Methods to Forecast Ratings of Perceived Exertion with Features from Wearable Sensors

The Relationship between Loneliness and Depression among College Students: Mining data derived from Passive Sensing

Inferring Mood Instability Via Smartphone Sensing

Predicting states of elevated negative affect in adolescents from smartphone sensors: a novel personalized machine learning approach

Predicting the Risk of Loneliness in Children and Adolescents: A Machine Learning Study

Can Loneliness be Predicted? Development of a Risk Prediction Model for Loneliness among Elderly Chinese: A Study Based on CLHLS

Integrating Wearable Sensor Data and Self-reported Diaries for Personalized Affect Forecasting

Predicting Depression in Adolescents Using Mobile and Wearable Sensors: Multimodal Machine Learning–Based Exploratory Study

DEVELOPING A MOBILE PHONE APPLICATION TO DETECT SOCIAL ISOLATION OVER MULTIPLE TIME POINTS

Wearable Social Sensing: Content-Based Processing Methodology and Implementation

StresSense: Real-Time detection of stress-displaying behaviors