Effect of Training Data Order for Machine Learning

Mange, Jeremy

doi:10.1109/csci49370.2019.00078

Cited by 11 publications

(6 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The type of data used to train a prediction model can significantly affect the performance of the model, and impact on the model's reliability and prediction outcomes [53]. Most of the research studies reviewed in this work included clinical data (n = 30; 96.8%), followed by patient demographic information (n = 21; 67.7%), molecular data (n = 15; 48.4%), and pathological image data (n = 9; 29.0%).…”

Section: Rq2: What Type Of Features Have Been Used?mentioning

confidence: 99%

Application of Artificial Intelligence Techniques to Predict Risk of Recurrence of Breast Cancer: A Systematic Review

Mazo

Aura

Rahman

et al. 2022

JPM

View full text Add to dashboard Cite

Breast cancer is the most common disease among women, with over 2.1 million new diagnoses each year worldwide. About 30% of patients initially presenting with early stage disease have a recurrence of cancer within 10 years. Predicting who will have a recurrence and who will not remains challenging, with consequent implications for associated treatment. Artificial intelligence strategies that can predict the risk of recurrence of breast cancer could help breast cancer clinicians avoid ineffective overtreatment. Despite its significance, most breast cancer recurrence datasets are insufficiently large, not publicly available, or imbalanced, making these studies more difficult. This systematic review investigates the role of artificial intelligence in the prediction of breast cancer recurrence. We summarise common techniques, features, training and testing methodologies, metrics, and discuss current challenges relating to implementation in clinical practice. We systematically reviewed works published between 1 January 2011 and 1 November 2021 using the methodology of Kitchenham and Charter. We leveraged Springer, Google Scholar, PubMed, and IEEE search engines. This review found three areas that require further work. First, there is no agreement on artificial intelligence methodologies, feature predictors, or assessment metrics. Second, issues such as sampling strategies, missing data, and class imbalance problems are rarely addressed or discussed. Third, representative datasets for breast cancer recurrence are scarce, which hinders model validation and deployment. We conclude that predicting breast cancer recurrence remains an open problem despite the use of artificial intelligence.

show abstract

Section: Rq2: What Type Of Features Have Been Used?mentioning

confidence: 99%

Application of Artificial Intelligence Techniques to Predict Risk of Recurrence of Breast Cancer: A Systematic Review

Mazo

Aura

Rahman

et al. 2022

JPM

View full text Add to dashboard Cite

show abstract

“…The “Infant-Trained-ASHS” model was trained over both CE and JF’s manual infant segmentations. On each leave-one-scan-out training iteration for this latter model, the order of CE and JF’s segmentations was shuffled randomly to prevent order effects (Mange, 2019).…”

Section: Methodsmentioning

confidence: 99%

“…The "Infant-Trained-ASHS" model was trained over both CE and JF's manual infant segmentations. On each leave-one-scan-out training iteration for this latter model, the order of CE and JF's segmentations was shuffled randomly to prevent order effects (Mange, 2019). In addition to this leave-one-scan-out approach, we performed an even more stringent test of generalization by omitting all scans from one participant from the training (some participants had more than one scan).…”

Section: Ashsmentioning

confidence: 99%

Automated and manual segmentation of the hippocampus in human infants

Fel

Ellis

Turk‐Browne

2022

Preprint

View full text Add to dashboard Cite

The hippocampus, critical for learning and memory, undergoes substantial changes early in life. Investigating the developmental trajectory of hippocampal structure and function requires an accurate method for segmenting this region from anatomical MRI scans. Although manual segmentation is regarded as the “gold standard” approach, it is laborious and subjective. This has fueled the pursuit of automated segmentation methods in adults. However, little is known about the reliability of these protocols in human infants, particularly when anatomical scan quality is low from increased head motion or shorter sequences that minimize head motion. During a task-based fMRI protocol, we collected quiet T1-weighted anatomical scans from 42 sessions with awake infants aged 4–23 months. We first had two expert tracers manually segment the hippocampus bilaterally and assess inter-rater reliability. We then attempted to predict these manual segmentations using four protocols: average adult template, average infant template, FreeSurfer software, and Automated Segmentation of Hippocampal Subfields (ASHS) software. ASHS generated the most reliable hippocampal segmentations in infants, exceeding manual inter-rater reliability of the experts. Automated methods can thus provide robust hippocampal segmentations of noisy T1-weighted infant scans, opening new possibilities for interrogating early hippocampal development.HighlightsInter-rater reliability of manual segmentation of infant hippocampus is moderate.Template-based methods and FreeSurfer provide reasonably accurate segmentations.ASHS produces highly accurate segmentations, exceeding manual inter-rater reliability.

show abstract

“…Several researchers studied the effect of training data to improve the performance of predictive models [19]- [21]. Research conducted by [19] shows that if there is harmonization in the dataset, classifier performance can be improved.…”

Section: Introductionmentioning

confidence: 99%

“…So, it is important to choose the appropriate training and testing data for harmonization. The training data order also greatly affected the performance of the classifier [21]. On exactly the same training dataset, classifier performance can vary from 10% to 100%.…”

Section: Introductionmentioning

confidence: 99%

Effects of Training Data on Prediction Model for Students' Academic Progress

Limanto¹,

Buliali²,

Saikhu³

2023

IJACSA

View full text Add to dashboard Cite

The ability to predict students' academic performance before the start of the class with credible accuracy could significantly aid the preparation of effective teaching and learning strategies. Several studies have been conducted to enhance the performance of prediction models by emphasizing three key factors: developing effective prediction algorithms, identifying significant predictor variables, and developing preprocessing techniques. Importantly, none of these studies focused on the effect of using different types of training data on the performance of prediction models. Therefore, this study was conducted to evaluate the effects of differences in training data on the performance of a prediction model designed to monitor students' academic progress. The findings showed that the performance of the prediction model was strongly influenced by the heterogeneity of the values of the predictor variables, which should accommodate all the existing possibilities. It was also discovered that the application of training data with different characteristics and sizes did not improve the performance of the prediction model when its heterogeneity was not representative.

show abstract

Effect of Training Data Order for Machine Learning

Cited by 11 publications

References 3 publications

Application of Artificial Intelligence Techniques to Predict Risk of Recurrence of Breast Cancer: A Systematic Review

Application of Artificial Intelligence Techniques to Predict Risk of Recurrence of Breast Cancer: A Systematic Review

Automated and manual segmentation of the hippocampus in human infants

Effects of Training Data on Prediction Model for Students' Academic Progress

Contact Info

Product

Resources

About