Systematic human learning by literature and data mining for feature selection in machine learning

Sufriyana, Herdiantri; Wu, Yu Wei; Su, Emily Chia-Yu

doi:10.21203/rs.3.pex-1634/v1

“…Please kindly follow the protocol. 9 After verifying causal factors, we only included those in a prediction model that applied a logistic regression with a shrinkage method, as recommended by PROBAST, instead of using a stepwise selection method. 17 We chose an RR, which applies L 2 -norm or beta regularization, because this method retains all causal factors within the model after weights are updated by training.…”

Section: De Ne and Verify Causal Factors As Parts Of Candidate Predic...mentioning

confidence: 99%

“…We proposed an analysis pipeline using several protocols previously described elsewhere. [9][10][11][12] The rst building block of this pipeline is a systematic human learning algorithm. 9 Based on hypotheticodeductive reasoning, human learning involves collection of prior knowledge to construct a causal diagram as a central assumption for hypothesis testing.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Human and machine learning pipelines for responsible clinical prediction using high-dimensional data

Sufriyana

¹

,

Wu

²

,

Su

³

2023

Preprint

0

View full text Add to dashboard Cite

This protocol aims to develop, validate, and deploy a prediction model using high dimensional data by both human and machine learning. The applicability is intended for clinical prediction in healthcare providers, including but not limited to those using medical histories from electronic health records. This protocol applies diverse approaches to improve both predictive performance and interpretability while maintaining the generalizability of model evaluation. However, some steps require expensive computational capacity; otherwise, these will take longer time. The key stages consist of designs of data collection and analysis, feature discovery and quality control, and model development, validation, and deployment.

show abstract

“…We proposed an analysis pipeline using several protocols previously described elsewhere. [9][10][11][12] The rst building block of this pipeline is a systematic human learning algorithm. 9 Based on hypotheticodeductive reasoning, human learning involves collection of prior knowledge to construct a causal diagram as a central assumption for hypothesis testing.…”

Section: Introductionmentioning

confidence: 99%

“…[9][10][11][12] The rst building block of this pipeline is a systematic human learning algorithm. 9 Based on hypotheticodeductive reasoning, human learning involves collection of prior knowledge to construct a causal diagram as a central assumption for hypothesis testing. Subsequently, statistical methods are used to verify the assumption.…”

Section: Introductionmentioning

confidence: 99%

Human and machine learning pipelines for responsible clinical prediction using high-dimensional data

Sufriyana¹,

Wu²,

Su³

2021

Preprint

View full text Add to dashboard Cite

This protocol aims to develop, validate, and deploy a prediction model using high dimensional data by both human and machine learning. The applicability is intended for clinical prediction in healthcare providers, including but not limited to those using medical histories from electronic health records. This protocol applies diverse approaches to improve both predictive performance and interpretability while maintaining the generalizability of model evaluation. However, some steps require expensive computational capacity; otherwise, these will take longer time. The key stages consist of designs of data collection and analysis, feature discovery and quality control, and model development, validation, and deployment.

show abstract

Prognostication for prelabor rupture of membranes and the time of delivery in nationwide insured women: development, validation, and deployment

Sufriyana

¹

,

Wu²,

Ec³

2021

Preprint

2

1

0

View full text Add to dashboard Cite

Prognostic prediction of prelabor rupture of membrane (PROM) lacks of sample size and external validation. We compared a statistical model, machine learning algorithms, and a deep-insight visible neural network (DI-VNN) for PROM and estimating the time of delivery. We selected visits, including PROM (n=23,791/170,730), retrospectively from a nationwide health insurance dataset. DI-VNN achieved the best prediction (area under receiver operating characteristics curve [AUROC] 0.73, 95% CI 0.72 to 0.75). Meanwhile, random forest using principal components achieved the best estimation with root mean squared errors ± 2.2 and 2.6 weeks respectively for the predicted event and nonevent. DI-VNN outperformed previous models by an external validation set, including one using a biomarker (AUROC 0.641; n=1,177). We deployed our models as a web application requiring diagnosis/procedure codes and dates. In conclusion, our models may be used solely in low-resource settings or as a preliminary model to reduce a specific test requiring high-resource setting.

show abstract

Systematic human learning by literature and data mining for feature selection in machine learning

Cited by 3 publications

References 17 publications

Human and machine learning pipelines for responsible clinical prediction using high-dimensional data

Human and machine learning pipelines for responsible clinical prediction using high-dimensional data

Human and machine learning pipelines for responsible clinical prediction using high-dimensional data

Prognostication for prelabor rupture of membranes and the time of delivery in nationwide insured women: development, validation, and deployment

Contact Info

Product

Resources

About