Akane Sano

Khadija Zanna, Kusha Sridhar, Han Yu et al.

There has been an increase in research in developing machine learning models for mental health detection or prediction in recent years due to increased mental health issues in society. Effective use of mental health prediction or detection models can help mental health practitioners re-define mental illnesses more objectively than currently done, and identify illnesses at an earlier stage when interventions may be more effective. However, there is still a lack of standard in evaluating bias in such machine learning models in the field, which leads to challenges in providing reliable predictions and in addressing disparities. This lack of standards persists due to factors such as technical difficulties, complexities of high dimensional clinical health data, etc., which are especially true for physiological signals. This along with prior evidence of relations between some physiological signals with certain demographic identities restates the importance of exploring bias in mental health prediction models that utilize physiological signals. In this work, we aim to perform a fairness analysis and implement a multi-task learning based bias mitigation method on anxiety prediction models using ECG data. Our method is based on the idea of epistemic uncertainty and its relationship with model weights and feature space representation. Our analysis showed that our anxiety prediction base model introduced some bias with regards to age, income, ethnicity, and whether a participant is born in the U.S. or not, and our bias mitigation method performed better at reducing the bias in the model, when compared to the reweighting mitigation technique. Our analysis on feature importance also helped identify relationships between heart rate variability and multiple demographic groupings.

Peikun Guo, Huiyuan Yang, Akane Sano

Data augmentation is a common practice to help generalization in the procedure of deep model training. In the context of physiological time series classification, previous research has primarily focused on label-invariant data augmentation methods. However, another class of augmentation techniques (\textit{i.e., Mixup}) that emerged in the computer vision field has yet to be fully explored in the time series domain. In this study, we systematically review the mix-based augmentations, including mixup, cutmix, and manifold mixup, on six physiological datasets, evaluating their performance across different sensory data and classification tasks. Our results demonstrate that the three mix-based augmentations can consistently improve the performance on the six datasets. More importantly, the improvement does not rely on expert knowledge or extensive parameter tuning. Lastly, we provide an overview of the unique properties of the mix-based augmentation methods and highlight the potential benefits of using the mix-based augmentation in physiological time series data.

Huiyuan Yang, Han Yu, Akane Sano

Deep learning has performed remarkably well on many tasks recently. However, the superior performance of deep models relies heavily on the availability of a large number of training data, which limits the wide adaptation of deep models on various clinical and affective computing tasks, as the labeled data are usually very limited. As an effective technique to increase the data variability and thus train deep models with better generalization, data augmentation (DA) is a critical step for the success of deep learning models on biobehavioral time series data. However, the effectiveness of various DAs for different datasets with different tasks and deep models is understudied for biobehavioral time series data. In this paper, we first systematically review eight basic DA methods for biobehavioral time series data, and evaluate the effects on seven datasets with three backbones. Next, we explore adapting more recent DA techniques (i.e., automatic augmentation, random augmentation) to biobehavioral time series data by designing a new policy architecture applicable to time series data. Last, we try to answer the question of why a DA is effective (or not) by first summarizing two desired attributes for augmentations (challenging and faithful), and then utilizing two metrics to quantitatively measure the corresponding attributes, which can guide us in the search for more effective DA for biobehavioral time series data by designing more challenging but still faithful transformations. Our code and results are available at Link.

Maryam Khalid, Akane Sano

Emotion prediction plays an essential role in mental health and emotion-aware computing. The complex nature of emotion resulting from its dependency on a person's physiological health, mental state, and his surroundings makes its prediction a challenging task. In this work, we utilize mobile sensing data to predict happiness and stress. In addition to a person's physiological features, we also incorporate the environment's impact through weather and social network. To this end, we leverage phone data to construct social networks and develop a machine learning architecture that aggregates information from multiple users of the graph network and integrates it with the temporal dynamics of data to predict emotion for all the users. The construction of social networks does not incur additional cost in terms of EMAs or data collection from users and doesn't raise privacy concerns. We propose an architecture that automates the integration of a user's social network affect prediction, is capable of dealing with the dynamic distribution of real-life social networks, making it scalable to large-scale networks. Our extensive evaluation highlights the improvement provided by the integration of social networks. We further investigate the impact of graph topology on model's performance.

Han Yu, Huiyuan Yang, Akane Sano

Electrocardiogram (ECG) is an essential signal in monitoring human heart activities. Researchers have achieved promising results in leveraging ECGs in clinical applications with deep learning models. However, the mainstream deep learning approaches usually neglect the periodic and formative attribute of the ECG heartbeat waveform. In this work, we propose a novel ECG-Segment based Learning (ECG-SL) framework to explicitly model the periodic nature of ECG signals. More specifically, ECG signals are first split into heartbeat segments, and then structural features are extracted from each of the segments. Based on the structural features, a temporal model is designed to learn the temporal information for various clinical tasks. Further, due to the fact that massive ECG signals are available but the labeled data are very limited, we also explore self-supervised learning strategy to pre-train the models, resulting significant improvement for downstream tasks. The proposed method outperforms the baseline model and shows competitive performances compared with task-specific methods in three clinical applications: cardiac condition diagnosis, sleep apnea detection, and arrhythmia classification. Further, we find that the ECG-SL tends to focus more on each heartbeat's peak and ST range than ResNet by visualizing the saliency maps.

Han Yu, Huiyuan Yang, Akane Sano

Contrastive learning has been utilized as a promising self-supervised learning approach to extract meaningful representations from unlabeled data. The majority of these methods take advantage of data-augmentation techniques to create diverse views from the original input. However, optimizing augmentations and their parameters for generating more effective views in contrastive learning frameworks is often resource-intensive and time-consuming. While several strategies have been proposed for automatically generating new views in computer vision, research in other domains, such as time-series biobehavioral data, remains limited. In this paper, we introduce a simple yet powerful module for automatic view generation in contrastive learning frameworks applied to time-series biobehavioral data, which is essential for modern health care, termed learning views for time-series data (LEAVES). This proposed module employs adversarial training to learn augmentation hyperparameters within contrastive learning frameworks. We assess the efficacy of our method on multiple time-series datasets using two well-known contrastive learning frameworks, namely SimCLR and BYOL. Across four diverse biobehavioral datasets, LEAVES requires only approximately 20 learnable parameters -- dramatically fewer than the about 580k parameters demanded by frameworks like ViewMaker, a previously proposed adversarially trained convolutional module in contrastive learning, while achieving competitive and often superior performance to existing baseline methods. Crucially, these efficiency gains are obtained without extensive manual hyperparameter tuning, which makes LEAVES particularly suitable for large-scale or real-time healthcare applications that demand both accuracy and practicality.

Zhaoyang Cao, Han Yu, Huiyuan Yang et al.

Due to individual heterogeneity, performance gaps are observed between generic (one-size-fits-all) models and person-specific models in data-driven health applications. However, in real-world applications, generic models are usually more favorable due to new-user-adaptation issues and system complexities, etc. To improve the performance of the generic model, we propose a representation learning framework that learns participant-invariant representations, named PiRL. The proposed framework utilizes maximum mean discrepancy (MMD) loss and domain-adversarial training to encourage the model to learn participant-invariant representations. Further, a triplet loss, which constrains the model for inter-class alignment of the representations, is utilized to optimize the learned representations for downstream health applications. We evaluated our frameworks on two public datasets related to physical and mental health, for detecting sleep apnea and stress, respectively. As preliminary results, we found the proposed approach shows around a 5% increase in accuracy compared to the baseline.

Bishal Lamichhane, Joanne Zhou, Akane Sano

Mobile sensing-based modeling of behavioral changes could predict an oncoming psychotic relapse in schizophrenia patients for timely interventions. Deep learning models could complement existing non-deep learning models for relapse prediction by modeling latent behavioral features relevant to the prediction. However, given the inter-individual behavioral differences, model personalization might be required for a predictive model. In this work, we propose RelapsePredNet, a Long Short-Term Memory (LSTM) neural network-based model for relapse prediction. The model is personalized for a particular patient by training using data from patients most similar to the given patient. Several demographics and baseline mental health scores were considered as personalization metrics to define patient similarity. We investigated the effect of personalization on training dataset characteristics, learned embeddings, and relapse prediction performance. We compared RelapsePredNet with a deep learning-based anomaly detection model for relapse prediction. Further, we investigated if RelapsePredNet could complement ClusterRFModel (a random forest model leveraging clustering and template features proposed in prior work) in a fusion model, by identifying latent behavioral features relevant for relapse prediction. The CrossCheck dataset consisting of continuous mobile sensing data obtained from 63 schizophrenia patients, each monitored for up to a year, was used for our evaluations. The proposed RelapsePredNet outperformed the deep learning-based anomaly detection model for relapse prediction. The F2 score for prediction were 0.21 and 0.52 in the full test set and the Relapse Test Set (consisting of data from patients who have had relapse only), respectively. These corresponded to a 29.4% and 38.8% improvement compared to the existing deep learning-based model for relapse prediction.

Zeyu Yang, Han Yu, Peikun Guo et al.

Diffusion models have emerged as a robust framework for various generative tasks, including tabular data synthesis. However, current tabular diffusion models tend to inherit bias in the training dataset and generate biased synthetic data, which may influence discriminatory actions. In this research, we introduce a novel tabular diffusion model that incorporates sensitive guidance to generate fair synthetic data with balanced joint distributions of the target label and sensitive attributes, such as sex and race. The empirical results demonstrate that our method effectively mitigates bias in training data while maintaining the quality of the generated samples. Furthermore, we provide evidence that our approach outperforms existing methods for synthesizing tabular data on fairness metrics such as demographic parity ratio and equalized odds ratio, achieving improvements of over $10\%$. Our implementation is available at https://github.com/comp-well-org/fair-tab-diffusion.

Akane Sano, Judith Amores, Mary Czerwinski

We explore the application of large language models (LLMs), pre-trained models with massive textual data for detecting and improving these altered states. We investigate the use of LLMs to estimate attention states, sleep stages, and sleep quality and generate sleep improvement suggestions and adaptive guided imagery scripts based on electroencephalogram (EEG) and physical activity data (e.g. waveforms, power spectrogram images, numerical features). Our results show that LLMs can estimate sleep quality based on human textual behavioral features and provide personalized sleep improvement suggestions and guided imagery scripts; however detecting attention, sleep stages, and sleep quality based on EEG and activity data requires further training data and domain-specific knowledge.

Huiyuan Yang, Han Yu, Kusha Sridhar et al.

Accurately recognizing health-related conditions from wearable data is crucial for improved healthcare outcomes. To improve the recognition accuracy, various approaches have focused on how to effectively fuse information from multiple sensors. Fusing multiple sensors is a common scenario in many applications, but may not always be feasible in real-world scenarios. For example, although combining bio-signals from multiple sensors (i.e., a chest pad sensor and a wrist wearable sensor) has been proved effective for improved performance, wearing multiple devices might be impractical in the free-living context. To solve the challenges, we propose an effective more to less (M2L) learning framework to improve testing performance with reduced sensors through leveraging the complementary information of multiple modalities during training. More specifically, different sensors may carry different but complementary information, and our model is designed to enforce collaborations among different modalities, where positive knowledge transfer is encouraged and negative knowledge transfer is suppressed, so that better representation is learned for individual modalities. Our experimental results show that our framework achieves comparable performance when compared with the full modalities. Our code and results will be available at https://github.com/compwell-org/More2Less.git.

Han Yu, Peikun Guo, Akane Sano

Time series data analysis is a critical component in various domains such as finance, healthcare, and meteorology. Despite the progress in deep learning for time series analysis, there remains a challenge in addressing the non-stationary nature of time series data. Traditional models, which are built on the assumption of constant statistical properties over time, often struggle to capture the temporal dynamics in realistic time series, resulting in bias and error in time series analysis. This paper introduces the Adaptive Wavelet Network (AdaWaveNet), a novel approach that employs Adaptive Wavelet Transformation for multi-scale analysis of non-stationary time series data. AdaWaveNet designed a lifting scheme-based wavelet decomposition and construction mechanism for adaptive and learnable wavelet transforms, which offers enhanced flexibility and robustness in analysis. We conduct extensive experiments on 10 datasets across 3 different tasks, including forecasting, imputation, and a newly established super-resolution task. The evaluations demonstrate the effectiveness of AdaWaveNet over existing methods in all three tasks, which illustrates its potential in various real-world applications.

Khadija Zanna, Akane Sano

Causal discovery (CD) plays a pivotal role in numerous scientific fields by clarifying the causal relationships that underlie phenomena observed in diverse disciplines. Despite significant advancements in CD algorithms that enhance bias and fairness analyses in machine learning, their application faces challenges due to the high computational demands and complexities of large-scale data. This paper introduces a framework that leverages Large Language Models (LLMs) for CD, utilizing a metadata-based approach akin to the reasoning processes of human experts. By shifting from pairwise queries to a more scalable breadth-first search (BFS) strategy, the number of required queries is reduced from quadratic to linear in terms of variable count, thereby addressing scalability concerns inherent in previous approaches. This method utilizes an Active Learning (AL) and a Dynamic Scoring Mechanism that prioritizes queries based on their potential information gain, combining mutual information, partial correlation, and LLM confidence scores to refine the causal graph more efficiently and accurately. This BFS query strategy reduces the required number of queries significantly, thereby addressing scalability concerns inherent in previous approaches. This study provides a more scalable and efficient solution for leveraging LLMs in fairness-driven CD, highlighting the effects of the different parameters on performance. We perform fairness analyses on the inferred causal graphs, identifying direct and indirect effects of sensitive attributes on outcomes. A comparison of these analyses against those from graphs produced by baseline methods highlights the importance of accurate causal graph construction in understanding bias and ensuring fairness in machine learning systems.

Khadija Zanna, Akane Sano

Bias originates from both data and algorithmic design, often exacerbated by traditional fairness methods that fail to address the subtle impacts of protected attributes. This study introduces an approach to mitigate bias in machine learning by leveraging model uncertainty. Our approach utilizes a multi-task learning (MTL) framework combined with Monte Carlo (MC) Dropout to assess and mitigate uncertainty in predictions related to protected labels. By incorporating MC Dropout, our framework quantifies prediction uncertainty, which is crucial in areas with vague decision boundaries, thereby enhancing model fairness. Our methodology integrates multi-objective learning through pareto-optimality to balance fairness and performance across various applications. We demonstrate the effectiveness and transferability of our approach across multiple datasets and enhance model explainability through saliency maps to interpret how input features influence predictions, thereby enhancing the interpretability of machine learning models in practical applications.

Emily Alsentzer, Marie-Laure Charpignon, Bill Chen et al.

The 6th Annual Conference on Health, Inference, and Learning (CHIL 2025), hosted by the Association for Health Learning and Inference (AHLI), was held in person on June 25-27, 2025, at the University of California, Berkeley, in Berkeley, California, USA. As part of this year's program, we hosted Research Roundtables to catalyze collaborative, small-group dialogue around critical, timely topics at the intersection of machine learning and healthcare. Each roundtable was moderated by a team of senior and junior chairs who fostered open exchange, intellectual curiosity, and inclusive engagement. The sessions emphasized rigorous discussion of key challenges, exploration of emerging opportunities, and collective ideation toward actionable directions in the field. In total, eight roundtables were held by 19 roundtable chairs on topics of "Explainability, Interpretability, and Transparency," "Uncertainty, Bias, and Fairness," "Causality," "Domain Adaptation," "Foundation Models," "Learning from Small Medical Data," "Multimodal Methods," and "Scalable, Translational Healthcare Solutions."

Khadija Zanna, Akane Sano

Causal discovery (CD) plays a pivotal role in understanding the mechanisms underlying complex systems. While recent algorithms can detect spurious associations and latent confounding, many struggle to recover fairness-relevant pathways in realistic, noisy settings. Large Language Models (LLMs), with their access to broad semantic knowledge, offer a promising complement to statistical CD approaches, particularly in domains where metadata provides meaningful relational cues. Ensuring fairness in machine learning requires understanding how sensitive attributes causally influence outcomes, yet CD methods often introduce spurious or biased pathways. We propose a hybrid LLM-based framework for CD that extends a breadth-first search (BFS) strategy with active learning and dynamic scoring. Variable pairs are prioritized for LLM-based querying using a composite score based on mutual information, partial correlation, and LLM confidence, improving discovery efficiency and robustness. To evaluate fairness sensitivity, we construct a semi-synthetic benchmark from the UCI Adult dataset, embedding a domain-informed causal graph with injected noise, label corruption, and latent confounding. We assess how well CD methods recover both global structure and fairness-critical paths. Our results show that LLM-guided methods, including the proposed method, demonstrate competitive or superior performance in recovering such pathways under noisy conditions. We highlight when dynamic scoring and active querying are most beneficial and discuss implications for bias auditing in real-world datasets.

Maryam Khalid, Akane Sano

Graph-based Active Learning (AL) leverages the structure of graphs to efficiently prioritize label queries, reducing labeling costs and user burden in applications like health monitoring, human behavior analysis, and sensor networks. By identifying strategically positioned nodes, graph AL minimizes data collection demands while maintaining model performance, making it a valuable tool for dynamic environments. Despite its potential, existing graph AL methods are often evaluated on static graph datasets and primarily focus on prediction accuracy, neglecting user-centric considerations such as sampling diversity, query fairness, and adaptability to dynamic settings. To bridge this gap, we introduce GRAIL, a novel benchmarking framework designed to evaluate graph AL strategies in dynamic, real-world environments. GRAIL introduces novel metrics to assess sustained effectiveness, diversity, and user burden, enabling a comprehensive evaluation of AL methods under varying conditions. Extensive experiments on datasets featuring dynamic, real-life human sensor data reveal trade-offs between prediction performance and user burden, highlighting limitations in existing AL strategies. GRAIL demonstrates the importance of balancing node importance, query diversity, and network topology, providing an evaluation mechanism for graph AL solutions in dynamic environments.

Maryam Khalid, Elizabeth B. Klerman, Andrew W. Mchill et al.

Sleep behavior significantly impacts health and acts as an indicator of physical and mental well-being. Monitoring and predicting sleep behavior with ubiquitous sensors may therefore assist in both sleep management and tracking of related health conditions. While sleep behavior depends on, and is reflected in the physiology of a person, it is also impacted by external factors such as digital media usage, social network contagion, and the surrounding weather. In this work, we propose SleepNet, a system that exploits social contagion in sleep behavior through graph networks and integrates it with physiological and phone data extracted from ubiquitous mobile and wearable devices for predicting next-day sleep labels about sleep duration. Our architecture overcomes the limitations of large-scale graphs containing connections irrelevant to sleep behavior by devising an attention mechanism. The extensive experimental evaluation highlights the improvement provided by incorporating social networks in the model. Additionally, we conduct robustness analysis to demonstrate the system's performance in real-life conditions. The outcomes affirm the stability of SleepNet against perturbations in input data. Further analyses emphasize the significance of network topology in prediction performance revealing that users with higher eigenvalue centrality are more vulnerable to data perturbations.

Han Yu, Akane Sano

Physiological and behavioral data collected from wearable or mobile sensors have been used to estimate self-reported stress levels. Since the stress annotation usually relies on self-reports during the study, a limited amount of labeled data can be an obstacle in developing accurate and generalized stress predicting models. On the other hand, the sensors can continuously capture signals without annotations. This work investigates leveraging unlabeled wearable sensor data for stress detection in the wild. We first applied data augmentation techniques on the physiological and behavioral data to improve the robustness of supervised stress detection models. Using an auto-encoder with actively selected unlabeled sequences, we pre-trained the supervised model structure to leverage the information learned from unlabeled samples. Then, we developed a semi-supervised learning framework to leverage the unlabeled data sequences. We combined data augmentation techniques with consistency regularization, which enforces the consistency of prediction output based on augmented and original unlabeled data. We validated these methods using three wearable/mobile sensor datasets collected in the wild. Our results showed that combining the proposed methods improved stress classification performance by 7.7% to 13.8% on the evaluated datasets, compared to the baseline supervised learning models.

Han Yu, Thomas Vaessen, Inez Myin-Germeys et al.

Multimodal wearable physiological data in daily life have been used to estimate self-reported stress labels. However, missing data modalities in data collection makes it challenging to leverage all the collected samples. Besides, heterogeneous sensor data and labels among individuals add challenges in building robust stress detection models. In this paper, we proposed a modality fusion network (MFN) to train models and infer self-reported binary stress labels under both complete and incomplete modality conditions. In addition, we applied personalized attention (PA) strategy to leverage personalized representation along with the generalized one-size-fits-all model. We evaluated our methods on a multimodal wearable sensor dataset (N=41) including galvanic skin response (GSR) and electrocardiogram (ECG). Compared to the baseline method using the samples with complete modalities, the performance of the MFN improved by 1.6% in f1-scores. On the other hand, the proposed PA strategy showed a 2.3% higher stress detection f1-score and approximately up to 70% reduction in personalized model parameter size (9.1 MB) compared to the previous state-of-the-art transfer learning strategy (29.3 MB).

Bishal Lamichhane, Dror Ben-Zeev, Andrew Campbell et al.

A schizophrenia relapse has severe consequences for a patient's health, work, and sometimes even life safety. If an oncoming relapse can be predicted on time, for example by detecting early behavioral changes in patients, then interventions could be provided to prevent the relapse. In this work, we investigated a machine learning based schizophrenia relapse prediction model using mobile sensing data to characterize behavioral features. A patient-independent model providing sequential predictions, closely representing the clinical deployment scenario for relapse prediction, was evaluated. The model uses the mobile sensing data from the recent four weeks to predict an oncoming relapse in the next week. We used the behavioral rhythm features extracted from daily templates of mobile sensing data, self-reported symptoms collected via EMA (Ecological Momentary Assessment), and demographics to compare different classifiers for the relapse prediction. Naive Bayes based model gave the best results with an F2 score of 0.083 when evaluated in a dataset consisting of 63 schizophrenia patients, each monitored for up to a year. The obtained F2 score, though low, is better than the baseline performance of random classification (F2 score of 0.02 $\pm$ 0.024). Thus, mobile sensing has predictive value for detecting an oncoming relapse and needs further investigation to improve the current performance. Towards that end, further feature engineering and model personalization based on the behavioral idiosyncrasies of a patient could be helpful.

Han Yu, Asami Itoh, Ryota Sakamoto et al.

Shift workers who are essential contributors to our society, face high risks of poor health and wellbeing. To help with their problems, we collected and analyzed physiological and behavioral wearable sensor data from shift working nurses and doctors, as well as their behavioral questionnaire data and their self-reported daily health and wellbeing labels, including alertness, happiness, energy, health, and stress. We found the similarities and differences between the responses of nurses and doctors. According to the differences in self-reported health and wellbeing labels between nurses and doctors, and the correlations among their labels, we proposed a job-role based multitask and multilabel deep learning model, where we modeled physiological and behavioral data for nurses and doctors simultaneously to predict participants' next day's multidimensional self-reported health and wellbeing status. Our model showed significantly better performances than baseline models and previous state-of-the-art models in the evaluations of binary/3-class classification and regression prediction tasks. We also found features related to heart rate, sleep, and work shift contributed to shift workers' health and wellbeing.

Joanne Zhou, Bishal Lamichhane, Dror Ben-Zeev et al.

We aim to develop clustering models to obtain behavioral representations from continuous multimodal mobile sensing data towards relapse prediction tasks. The identified clusters could represent different routine behavioral trends related to daily living of patients as well as atypical behavioral trends associated with impending relapse. We used the mobile sensing data obtained in the CrossCheck project for our analysis. Continuous data from six different mobile sensing-based modalities (e.g. ambient light, sound/conversation, acceleration etc.) obtained from a total of 63 schizophrenia patients, each monitored for up to a year, were used for the clustering models and relapse prediction evaluation. Two clustering models, Gaussian Mixture Model (GMM) and Partition Around Medoids (PAM), were used to obtain behavioral representations from the mobile sensing data. The features obtained from the clustering models were used to train and evaluate a personalized relapse prediction model using Balanced Random Forest. The personalization was done by identifying optimal features for a given patient based on a personalization subset consisting of other patients who are of similar age. The clusters identified using the GMM and PAM models were found to represent different behavioral patterns (such as clusters representing sedentary days, active but with low communications days, etc.). Significant changes near the relapse periods were seen in the obtained behavioral representation features from the clustering models. The clustering model based features, together with other features characterizing the mobile sensing data, resulted in an F2 score of 0.24 for the relapse prediction task in a leave-one-patient-out evaluation setting. This obtained F2 score is significantly higher than a random classification baseline with an average F2 score of 0.042.

Cheng Wan, Andrew W. McHill, Elizabeth Klerman et al.

Circadian rhythms influence multiple essential biological activities including sleep, performance, and mood. The dim light melatonin onset (DLMO) is the gold standard for measuring human circadian phase (i.e., timing). The collection of DLMO is expensive and time-consuming since multiple saliva or blood samples are required overnight in special conditions, and the samples must then be assayed for melatonin. Recently, several computational approaches have been designed for estimating DLMO. These methods collect daily sampled data (e.g., sleep onset/offset times) or frequently sampled data (e.g., light exposure/skin temperature/physical activity collected every minute) to train learning models for estimating DLMO. One limitation of these studies is that they only leverage one time-scale data. We propose a two-step framework for estimating DLMO using data from both time scales. The first step summarizes data from before the current day, while the second step combines this summary with frequently sampled data of the current day. We evaluate three moving average models that input sleep timing data as the first step and use recurrent neural network models as the second step. The results using data from 207 undergraduates show that our two-step model with two time-scale features has statistically significantly lower root-mean-square errors than models that use either daily sampled data or frequently sampled data.