# What are some examples of mobile sensing data that can be used to predict mood changes? Related: - [[Paper - Correlations between objective behavioural features collected from mobile and wearable devices and depressive mood symptoms in patients with affective disorders systematic review]] - [[Article - ‪Passive Sensing of Prediction of Moment-to-Moment Depressed Mood among Undergraduates with Clinical Levels of Depression Sample using Smartphones‬]] - [[Digital Phenotype Signals associated with Depression]] - [[Using Machine Learning and Smartphone and Smartwatch Data to Detect Emotional States and Transitions An Exploratory Study]] - [[Digital Phenotype Signals associated with Psychosis]] - [[EMA combined with passive digital phenotyping to give more holistic picture of patient functioning]] - [[Paper - A systematic review of location data for depression prediction]] - [[Article -Digital biomarkers of mood disorders and symptom change]] --- # **Physical Activity and Movement**: - Based on the provided sources, here are some examples of mobile sensing data that can be used to predict mood changes: ### Types of Mobile Sensing Data 1. **Physical Activity and Movement**: - **Accelerometer**: Measures the intensity and frequency of physical movements. Changes in activity levels can indicate mood variations, such as reduced activity during depressive episodes[2][9].[9](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8976254/) - **Gyroscope**: Tracks orientation and rotation, providing additional context to movement patterns[7]. 2. **Location and Mobility**: - **GPS**: Tracks location data to monitor mobility patterns. Reduced mobility or changes in routine can be associated with mood disorders like depression or anxiety[3][10]. - **Bluetooth**: Detects proximity to other devices, which can infer social interactions and isolation levels[2]. 3. **Phone Usage Patterns**: - **Call and Text Logs**: Frequency and duration of phone calls and text messages can indicate social engagement or withdrawal[7]. - **App Usage**: Time spent on different apps, especially social media and messaging apps, can reflect mood states. Increased usage might correlate with anxiety or stress[2]. 4. **Sleep Patterns**: - **Sleep Tracking**: Monitors sleep duration and quality using sensors in wearables or smartphone apps. Poor sleep is often linked to mood disorders[4][9]. 5. **Environmental Context**: - **Microphone**: Captures ambient noise levels, which can provide context about the user's environment and social interactions[2][9]. - **Ambient Light Sensor**: Measures light exposure, which affects circadian rhythms and mood[9]. 6. **Linguistic and Behavioral Data**: - **Text and Speech Analysis**: Analyzes the content and tone of text messages and voice recordings to detect emotional states and mood changes[1][18][20]. ### Examples from Studies 1. **FedTherapist**: Utilizes continuous speech and keyboard input to predict self-reported depression, stress, anxiety, and mood with higher accuracy compared to non-language features[1]. 2. **Smartphone-tracked Digital Biomarkers**: Uses smartphone usage logs to predict momentary subjective stress based on messenger app use frequency, social network site usage, and sleep proxies[2]. 3. **Machine Learning for Passive Mental Health Symptom Prediction**: Employs mobile sensing data like movement (accelerometer) and noise levels (microphone) to predict mood deviations using passive sensing and psychological traits[2]. 4. **Tracking Subjective Sleep Quality and Mood**: Investigates how various automatically gathered mobile sensing features, such as accelerometer, charging status, light sensor, physical activity, screen activity, and Wi-Fi status, can predict subjective sleep quality, negative affect, and depression[9][12]. 5. **Personalized Mood Prediction**: Uses continuous digital assessments of behavior from smartphone sensors, including GPS, accelerometer, and screen state, to predict future mood and depression severity[8]. ### Summary Mobile sensing data provides a rich and continuous stream of information that can be used to predict mood changes. By leveraging various sensors and data types, such as physical activity, location, phone usage patterns, sleep quality, environmental context, and linguistic data, researchers can develop predictive models that offer insights into an individual's mental health. These models can be further enhanced using machine learning techniques to improve accuracy and personalization. Citations: [1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9812759/ [2] https://www.semanticscholar.org/paper/17b2539b1ad0a9856a3d175dcb7a575250fd3926 [3] https://www.semanticscholar.org/paper/1a7fa79a7bfae4a126614e537552a41a443b4956 [4] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9185362/ [5] https://www.semanticscholar.org/paper/c7a1b204ce6cc10cbdf488d02cba8703f5af7220 [6] https://www.semanticscholar.org/paper/76318baa04dc08e4b96381554f1925883fc48422 [7] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10599285/ [8] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10902386/ [9] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8976254/ [10] https://www.semanticscholar.org/paper/f535642b6fa9b0be7b56359ee065c47949ef4e86 [11] https://pubmed.ncbi.nlm.nih.gov/33019316/ [12] https://www.semanticscholar.org/paper/4f9fd9a517d877f1101490db4d6684ae90d3f734 [13] https://pubmed.ncbi.nlm.nih.gov/38200262/ [14] https://www.semanticscholar.org/paper/e918ff676dd9d88211f9754a708bd1af95186b28 [15] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10787337/ [16] https://arxiv.org/abs/1711.06350 [17] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8784598/ [18] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8881775/ [19] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10523016/ [20] https://www.semanticscholar.org/paper/7e9f154980ce3912d9d6b713c8d79da35aba8663