An exploratory application of machine learning methods to optimize prediction of responsiveness to digital interventions for eating disorder symptoms
- Linardon, Jake, Fuller-Tyszkiewicz, Matthew, Shatte, Adrian, Greenwood, Christopher
- Authors: Linardon, Jake , Fuller-Tyszkiewicz, Matthew , Shatte, Adrian , Greenwood, Christopher
- Date: 2022
- Type: Text , Journal article
- Relation: International Journal of Eating Disorders Vol. 55, no. 6 (2022), p. 845-850
- Full Text:
- Reviewed:
- Description: Objective: Digital interventions show promise to address eating disorder (ED) symptoms. However, response rates are variable, and the ability to predict responsiveness to digital interventions has been poor. We tested whether machine learning (ML) techniques can enhance outcome predictions from digital interventions for ED symptoms. Method: Data were aggregated from three RCTs (n = 826) of self-guided digital interventions for EDs. Predictive models were developed for four key outcomes: uptake, adherence, drop-out, and symptom-level change. Seven ML techniques for classification were tested and compared against the generalized linear model (GLM). Results: The seven ML methods used to predict outcomes from 36 baseline variables were poor for the three engagement outcomes (AUCs = 0.48–0.52), but adequate for symptom-level change (R2 =.15–.40). ML did not offer an added benefit to the GLM. Incorporating intervention usage pattern data improved ML prediction accuracy for drop-out (AUC = 0.75–0.93) and adherence (AUC = 0.92–0.99). Age, motivation, symptom severity, and anxiety emerged as influential outcome predictors. Conclusion: A limited set of routinely measured baseline variables was not sufficient to detect a performance benefit of ML over traditional approaches. The benefits of ML may emerge when numerous usage pattern variables are modeled, although this validation in larger datasets before stronger conclusions can be made. © 2022 The Authors. International Journal of Eating Disorders published by Wiley Periodicals LLC.
- Authors: Linardon, Jake , Fuller-Tyszkiewicz, Matthew , Shatte, Adrian , Greenwood, Christopher
- Date: 2022
- Type: Text , Journal article
- Relation: International Journal of Eating Disorders Vol. 55, no. 6 (2022), p. 845-850
- Full Text:
- Reviewed:
- Description: Objective: Digital interventions show promise to address eating disorder (ED) symptoms. However, response rates are variable, and the ability to predict responsiveness to digital interventions has been poor. We tested whether machine learning (ML) techniques can enhance outcome predictions from digital interventions for ED symptoms. Method: Data were aggregated from three RCTs (n = 826) of self-guided digital interventions for EDs. Predictive models were developed for four key outcomes: uptake, adherence, drop-out, and symptom-level change. Seven ML techniques for classification were tested and compared against the generalized linear model (GLM). Results: The seven ML methods used to predict outcomes from 36 baseline variables were poor for the three engagement outcomes (AUCs = 0.48–0.52), but adequate for symptom-level change (R2 =.15–.40). ML did not offer an added benefit to the GLM. Incorporating intervention usage pattern data improved ML prediction accuracy for drop-out (AUC = 0.75–0.93) and adherence (AUC = 0.92–0.99). Age, motivation, symptom severity, and anxiety emerged as influential outcome predictors. Conclusion: A limited set of routinely measured baseline variables was not sufficient to detect a performance benefit of ML over traditional approaches. The benefits of ML may emerge when numerous usage pattern variables are modeled, although this validation in larger datasets before stronger conclusions can be made. © 2022 The Authors. International Journal of Eating Disorders published by Wiley Periodicals LLC.
Filter feature selection based boolean modelling for genetic network inference
- Gamage, Hasini, Chetty, Madhu, Shatte, Adrian, Hallinan, Jennifer
- Authors: Gamage, Hasini , Chetty, Madhu , Shatte, Adrian , Hallinan, Jennifer
- Date: 2022
- Type: Text , Journal article
- Relation: BioSystems Vol. 221, no. (2022), p.
- Full Text:
- Reviewed:
- Description: The reconstruction of Gene Regulatory Networks (GRNs) from time series gene expression data is highly relevant for the discovery of complex biological interactions and dynamics. Various computational strategies have been developed for this task, but most approaches have low computational efficiency and are not able to cope with high-dimensional, low sample-number, gene expression data. In this paper, we introduce a novel combined filter feature selection approach for efficient and accurate inference of GRNs. A Boolean framework for network modelling is used to demonstrate the efficacy of the proposed approach. Using discretized microarray expression data, the genes most relevant to each target gene are first filtered using ReliefF, an instance-based feature ranking method that is here applied for the first time to GRN inference. Then, further gene selection from the filtered-gene list is done using a mutual information-based min-redundancy max-relevance criterion by eliminating irrelevant genes. This combined method is executed on resampled datasets to finalize the optimal set of regulatory genes. Building upon our previous research, a Pearson correlation coefficient-based Boolean modelling approach is utilized for the efficient identification of the optimal regulatory rules associated with selected regulatory genes. The proposed approach was evaluated using gene expression datasets from small-scale and medium-scale real gene networks, and was observed to be more effective than Linear Discriminant Analysis, performed better than the individual feature selection methods, and obtained improved Structural Accuracy with a higher number of true positives than other state-of-the-art methods, while outperforming these methods with respect to Dynamic Accuracy and efficiency. © 2022 Elsevier B.V.
- Authors: Gamage, Hasini , Chetty, Madhu , Shatte, Adrian , Hallinan, Jennifer
- Date: 2022
- Type: Text , Journal article
- Relation: BioSystems Vol. 221, no. (2022), p.
- Full Text:
- Reviewed:
- Description: The reconstruction of Gene Regulatory Networks (GRNs) from time series gene expression data is highly relevant for the discovery of complex biological interactions and dynamics. Various computational strategies have been developed for this task, but most approaches have low computational efficiency and are not able to cope with high-dimensional, low sample-number, gene expression data. In this paper, we introduce a novel combined filter feature selection approach for efficient and accurate inference of GRNs. A Boolean framework for network modelling is used to demonstrate the efficacy of the proposed approach. Using discretized microarray expression data, the genes most relevant to each target gene are first filtered using ReliefF, an instance-based feature ranking method that is here applied for the first time to GRN inference. Then, further gene selection from the filtered-gene list is done using a mutual information-based min-redundancy max-relevance criterion by eliminating irrelevant genes. This combined method is executed on resampled datasets to finalize the optimal set of regulatory genes. Building upon our previous research, a Pearson correlation coefficient-based Boolean modelling approach is utilized for the efficient identification of the optimal regulatory rules associated with selected regulatory genes. The proposed approach was evaluated using gene expression datasets from small-scale and medium-scale real gene networks, and was observed to be more effective than Linear Discriminant Analysis, performed better than the individual feature selection methods, and obtained improved Structural Accuracy with a higher number of true positives than other state-of-the-art methods, while outperforming these methods with respect to Dynamic Accuracy and efficiency. © 2022 Elsevier B.V.
From general language understanding to noisy text comprehension
- Kasthuriarachchy, Buddhika, Chetty, Madhu, Shatte, Adrian, Walls, Darren
- Authors: Kasthuriarachchy, Buddhika , Chetty, Madhu , Shatte, Adrian , Walls, Darren
- Date: 2021
- Type: Text , Journal article
- Relation: Applied Sciences (Switzerland) Vol. 11, no. 17 (2021), p.
- Full Text:
- Reviewed:
- Description: Obtaining meaning-rich representations of social media inputs, such as Tweets (unstructured and noisy text), from general-purpose pre-trained language models has become challenging, as these inputs typically deviate from mainstream English usage. The proposed research establishes effective methods for improving the comprehension of noisy texts. For this, we propose a new generic methodology to derive a diverse set of sentence vectors combining and extracting various linguistic characteristics from latent representations of multi-layer, pre-trained language models. Further, we clearly establish how BERT, a state-of-the-art pre-trained language model, comprehends the linguistic attributes of Tweets to identify appropriate sentence representations. Five new probing tasks are developed for Tweets, which can serve as benchmark probing tasks to study noisy text comprehension. Experiments are carried out for classification accuracy by deriving the sentence vectors from GloVe-based pre-trained models and Sentence-BERT, and by using different hidden layers from the BERT model. We show that the initial and middle layers of BERT have better capability for capturing the key linguistic characteristics of noisy texts than its latter layers. With complex predictive models, we further show that the sentence vector length has lesser importance to capture linguistic information, and the proposed sentence vectors for noisy texts perform better than the existing state-of-the-art sentence vectors. © 2021 by the authors. Licensee MDPI, Basel, Switzerland.
- Authors: Kasthuriarachchy, Buddhika , Chetty, Madhu , Shatte, Adrian , Walls, Darren
- Date: 2021
- Type: Text , Journal article
- Relation: Applied Sciences (Switzerland) Vol. 11, no. 17 (2021), p.
- Full Text:
- Reviewed:
- Description: Obtaining meaning-rich representations of social media inputs, such as Tweets (unstructured and noisy text), from general-purpose pre-trained language models has become challenging, as these inputs typically deviate from mainstream English usage. The proposed research establishes effective methods for improving the comprehension of noisy texts. For this, we propose a new generic methodology to derive a diverse set of sentence vectors combining and extracting various linguistic characteristics from latent representations of multi-layer, pre-trained language models. Further, we clearly establish how BERT, a state-of-the-art pre-trained language model, comprehends the linguistic attributes of Tweets to identify appropriate sentence representations. Five new probing tasks are developed for Tweets, which can serve as benchmark probing tasks to study noisy text comprehension. Experiments are carried out for classification accuracy by deriving the sentence vectors from GloVe-based pre-trained models and Sentence-BERT, and by using different hidden layers from the BERT model. We show that the initial and middle layers of BERT have better capability for capturing the key linguistic characteristics of noisy texts than its latter layers. With complex predictive models, we further show that the sentence vector length has lesser importance to capture linguistic information, and the proposed sentence vectors for noisy texts perform better than the existing state-of-the-art sentence vectors. © 2021 by the authors. Licensee MDPI, Basel, Switzerland.
Methods and applications of social media monitoring of mental health during disasters : scoping review
- Teague, Samantha, Shatte, Adrian, Weller, Emmelyn, Fuller-Tyszkiewicz, Matthew, Hutchinson, Delyse
- Authors: Teague, Samantha , Shatte, Adrian , Weller, Emmelyn , Fuller-Tyszkiewicz, Matthew , Hutchinson, Delyse
- Date: 2022
- Type: Text , Journal article , Review
- Relation: JMIR Mental Health Vol. 9, no. 2 (2022), p.
- Full Text:
- Reviewed:
- Description: Background: With the increasing frequency and magnitude of disasters internationally, there is growing research and clinical interest in the application of social media sites for disaster mental health surveillance. However, important questions remain regarding the extent to which unstructured social media data can be harnessed for clinically meaningful decision-making. Objective: This comprehensive scoping review synthesizes interdisciplinary literature with a particular focus on research methods and applications. Methods: A total of 6 health and computer science databases were searched for studies published before April 20, 2021, resulting in the identification of 47 studies. Included studies were published in peer-reviewed outlets and examined mental health during disasters or crises by using social media data. Results: Applications across 31 mental health issues were identified, which were grouped into the following three broader themes: estimating mental health burden, planning or evaluating interventions and policies, and knowledge discovery. Mental health assessments were completed by primarily using lexical dictionaries and human annotations. The analyses included a range of supervised and unsupervised machine learning, statistical modeling, and qualitative techniques. The overall reporting quality was poor, with key details such as the total number of users and data features often not being reported. Further, biases in sample selection and related limitations in generalizability were often overlooked. Conclusions: The application of social media monitoring has considerable potential for measuring mental health impacts on populations during disasters. Studies have primarily conceptualized mental health in broad terms, such as distress or negative affect, but greater focus is required on validating mental health assessments. There was little evidence for the clinical integration of social media-based disaster mental health monitoring, such as combining surveillance with social media-based interventions or developing and testing real-world disaster management tools. To address issues with study quality, a structured set of reporting guidelines is recommended to improve the methodological quality, replicability, and clinical relevance of future research on the social media monitoring of mental health during disasters. © 2022 Samantha J Teague, Adrian B R Shatte, Emmelyn Weller, Matthew Fuller-Tyszkiewicz, Delyse M Hutchinson.
- Authors: Teague, Samantha , Shatte, Adrian , Weller, Emmelyn , Fuller-Tyszkiewicz, Matthew , Hutchinson, Delyse
- Date: 2022
- Type: Text , Journal article , Review
- Relation: JMIR Mental Health Vol. 9, no. 2 (2022), p.
- Full Text:
- Reviewed:
- Description: Background: With the increasing frequency and magnitude of disasters internationally, there is growing research and clinical interest in the application of social media sites for disaster mental health surveillance. However, important questions remain regarding the extent to which unstructured social media data can be harnessed for clinically meaningful decision-making. Objective: This comprehensive scoping review synthesizes interdisciplinary literature with a particular focus on research methods and applications. Methods: A total of 6 health and computer science databases were searched for studies published before April 20, 2021, resulting in the identification of 47 studies. Included studies were published in peer-reviewed outlets and examined mental health during disasters or crises by using social media data. Results: Applications across 31 mental health issues were identified, which were grouped into the following three broader themes: estimating mental health burden, planning or evaluating interventions and policies, and knowledge discovery. Mental health assessments were completed by primarily using lexical dictionaries and human annotations. The analyses included a range of supervised and unsupervised machine learning, statistical modeling, and qualitative techniques. The overall reporting quality was poor, with key details such as the total number of users and data features often not being reported. Further, biases in sample selection and related limitations in generalizability were often overlooked. Conclusions: The application of social media monitoring has considerable potential for measuring mental health impacts on populations during disasters. Studies have primarily conceptualized mental health in broad terms, such as distress or negative affect, but greater focus is required on validating mental health assessments. There was little evidence for the clinical integration of social media-based disaster mental health monitoring, such as combining surveillance with social media-based interventions or developing and testing real-world disaster management tools. To address issues with study quality, a structured set of reporting guidelines is recommended to improve the methodological quality, replicability, and clinical relevance of future research on the social media monitoring of mental health during disasters. © 2022 Samantha J Teague, Adrian B R Shatte, Emmelyn Weller, Matthew Fuller-Tyszkiewicz, Delyse M Hutchinson.
Peer support of fathers on reddit : quantifying the stressors, behaviors, and drivers
- Teague, Samantha, Shatte, Adrian
- Authors: Teague, Samantha , Shatte, Adrian
- Date: 2021
- Type: Text , Journal article
- Relation: Psychology of Men and Masculinity Vol. 22, no. 4 (2021), p. 757-766
- Full Text:
- Reviewed:
- Description: This article aimed to delineate the behavioral patterns of fathers in seeking and providing peer support on the popular social media site Reddit using a sample of 2,393 users. First, fathers’ support-seeking posts were characterized, finding that fathers self-disclosed a range of individual, familial, and societal stressors, including topics sensitive to traditional male gender roles. Second, peers’ comments were differentiated by support type, with differences observed in the behaviors, emotions, and language that peers use when providing advice, confirmation, and encouragement. Third, the relationship between types of fatherhood stressors and their associated peer comments was mapped. While fathers seeking support for individual stressors received fewer comments, the support provided utilized more actionoriented language. Finally, a statistical model was developed to examine the factors that drive peer support on the fatherhood forums, which are observed to influence the quality of peers’ comments and peers’ commenting behaviors. Combined, the findings provide a comprehensive understanding of the peer support environment for fathers on social media like Reddit, strengthening the research literature that is limited to qualitative evidence to date. The results have important implications for formal support services targeting fathers, both online and offline © 2021 American Psychological Association
- Authors: Teague, Samantha , Shatte, Adrian
- Date: 2021
- Type: Text , Journal article
- Relation: Psychology of Men and Masculinity Vol. 22, no. 4 (2021), p. 757-766
- Full Text:
- Reviewed:
- Description: This article aimed to delineate the behavioral patterns of fathers in seeking and providing peer support on the popular social media site Reddit using a sample of 2,393 users. First, fathers’ support-seeking posts were characterized, finding that fathers self-disclosed a range of individual, familial, and societal stressors, including topics sensitive to traditional male gender roles. Second, peers’ comments were differentiated by support type, with differences observed in the behaviors, emotions, and language that peers use when providing advice, confirmation, and encouragement. Third, the relationship between types of fatherhood stressors and their associated peer comments was mapped. While fathers seeking support for individual stressors received fewer comments, the support provided utilized more actionoriented language. Finally, a statistical model was developed to examine the factors that drive peer support on the fatherhood forums, which are observed to influence the quality of peers’ comments and peers’ commenting behaviors. Combined, the findings provide a comprehensive understanding of the peer support environment for fathers on social media like Reddit, strengthening the research literature that is limited to qualitative evidence to date. The results have important implications for formal support services targeting fathers, both online and offline © 2021 American Psychological Association
Schema : an open-source, distributed mobile platform for deploying mHealth research tools and interventions
- Shatte, Adrian, Teague, Samantha
- Authors: Shatte, Adrian , Teague, Samantha
- Date: 2020
- Type: Text , Journal article
- Relation: BMC Medical Research Methodology Vol. 20, no. 1 (2020), p.
- Full Text:
- Reviewed:
- Description: Background: Mobile applications for health, also known as 'mHealth apps', have experienced increasing popularity over the past ten years. However, most publicly available mHealth apps are not clinically validated, and many do not utilise evidence-based strategies. Health researchers wishing to develop and evaluate mHealth apps may be impeded by cost and technical skillset barriers. As traditionally lab-based methods are translated onto mobile platforms, robust and accessible tools are needed to enable the development of quality, evidence-based programs by clinical experts. Results: This paper introduces schema, an open-source, distributed, app-based platform for researchers to deploy behavior monitoring and health interventions onto mobile devices. The architecture and design features of the platform are discussed, including flexible scheduling, randomisation, a wide variety of survey and media elements, and distributed storage of data. The platform supports a range of research designs, including cross-sectional surveys, ecological momentary assessment, randomised controlled trials, and micro-randomised just-in-time adaptive interventions. Use cases for both researchers and participants are considered to demonstrate the flexibility and usefulness of the platform for mHealth research. Conclusions: The paper concludes by considering the strengths and limitations of the platform, and a call for support from the research community in areas of technical development and evaluation. To get started with schema, please visit the GitHub repository: Https://github.com/schema-app/schema. © 2020 The Author(s).
- Authors: Shatte, Adrian , Teague, Samantha
- Date: 2020
- Type: Text , Journal article
- Relation: BMC Medical Research Methodology Vol. 20, no. 1 (2020), p.
- Full Text:
- Reviewed:
- Description: Background: Mobile applications for health, also known as 'mHealth apps', have experienced increasing popularity over the past ten years. However, most publicly available mHealth apps are not clinically validated, and many do not utilise evidence-based strategies. Health researchers wishing to develop and evaluate mHealth apps may be impeded by cost and technical skillset barriers. As traditionally lab-based methods are translated onto mobile platforms, robust and accessible tools are needed to enable the development of quality, evidence-based programs by clinical experts. Results: This paper introduces schema, an open-source, distributed, app-based platform for researchers to deploy behavior monitoring and health interventions onto mobile devices. The architecture and design features of the platform are discussed, including flexible scheduling, randomisation, a wide variety of survey and media elements, and distributed storage of data. The platform supports a range of research designs, including cross-sectional surveys, ecological momentary assessment, randomised controlled trials, and micro-randomised just-in-time adaptive interventions. Use cases for both researchers and participants are considered to demonstrate the flexibility and usefulness of the platform for mHealth research. Conclusions: The paper concludes by considering the strengths and limitations of the platform, and a call for support from the research community in areas of technical development and evaluation. To get started with schema, please visit the GitHub repository: Https://github.com/schema-app/schema. © 2020 The Author(s).
Social media markers to identify fathers at risk of postpartum depression : a machine learning approach
- Shatte, Adrian, Hutchinson, Delyse, Fuller-Tyszkiewicz, Matthew, Teague, Samantha
- Authors: Shatte, Adrian , Hutchinson, Delyse , Fuller-Tyszkiewicz, Matthew , Teague, Samantha
- Date: 2020
- Type: Text , Journal article
- Relation: Cyberpsychology, Behavior, and Social Networking Vol. 23, no. 9 (2020), p. 611-618
- Full Text:
- Reviewed:
- Description: Postpartum depression (PPD) is a significant mental health issue in mothers and fathers alike; yet at-risk fathers often come to the attention of health care professionals late due to low awareness of symptoms and reluctance to seek help. This study aimed to examine whether passive social media markers are effective for identifying fathers at risk of PPD. We collected 67,796 Reddit posts from 365 fathers, spanning a 6-month period around the birth of their child. A list of "at-risk"words was developed in collaboration with a perinatal mental health expert. PPD was assessed by evaluating the change in fathers' use of words indicating depressive symptomatology after childbirth. Predictive models were developed as a series of support vector machine classifiers using behavior, emotion, linguistic style, and discussion topics as features. The performance of these classifiers indicates that fathers at risk of PPD can be predicted from their prepartum data alone. Overall, the best performing model used discussion topic features only with a recall score of 0.82. These findings could assist in the development of support and intervention tools for fathers during the prepartum period, with specific applicability to personalized and preventative support tools for at-risk fathers. © Copyright 2020, Mary Ann Liebert, Inc., publishers 2020.
- Authors: Shatte, Adrian , Hutchinson, Delyse , Fuller-Tyszkiewicz, Matthew , Teague, Samantha
- Date: 2020
- Type: Text , Journal article
- Relation: Cyberpsychology, Behavior, and Social Networking Vol. 23, no. 9 (2020), p. 611-618
- Full Text:
- Reviewed:
- Description: Postpartum depression (PPD) is a significant mental health issue in mothers and fathers alike; yet at-risk fathers often come to the attention of health care professionals late due to low awareness of symptoms and reluctance to seek help. This study aimed to examine whether passive social media markers are effective for identifying fathers at risk of PPD. We collected 67,796 Reddit posts from 365 fathers, spanning a 6-month period around the birth of their child. A list of "at-risk"words was developed in collaboration with a perinatal mental health expert. PPD was assessed by evaluating the change in fathers' use of words indicating depressive symptomatology after childbirth. Predictive models were developed as a series of support vector machine classifiers using behavior, emotion, linguistic style, and discussion topics as features. The performance of these classifiers indicates that fathers at risk of PPD can be predicted from their prepartum data alone. Overall, the best performing model used discussion topic features only with a recall score of 0.82. These findings could assist in the development of support and intervention tools for fathers during the prepartum period, with specific applicability to personalized and preventative support tools for at-risk fathers. © Copyright 2020, Mary Ann Liebert, Inc., publishers 2020.
- «
- ‹
- 1
- ›
- »