On Feature Weighting and Selection for Medical Document Classification

Parlak, Bekir; Uysal, Alper Kürşat

doi:10.1007/978-3-319-58965-7_19

Cited by 19 publications

(17 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It has long been recognized that clinical reports are beneficial for secondary use. A number of researchers has deployed clinical data mining to mine useful information (such as medical concepts or medical entity) from clinical reports [27]. Various applications of clinical data mining include clinical information extraction [28], clinical relation extraction [29], clinical document clustering [30] and classification of clinical documents [7].…”

Section: Resultsmentioning

confidence: 99%

A Predictive Machine Learning Model to Determine Alcohol Use Disorder

Ebrahimi

Wiil

Andersen

et al. 2020

2020 IEEE Symposium on Computers and Communications (ISCC)

View full text Add to dashboard Cite

Prediction of alcohol use disorder (AUD) may reduce the number of deaths caused by alcohol-related diseases. However, prediction of AUD based on patients' historical clinical data is still an open research objective. This study proposes a method to predict AUD from electronic health record (EHR) data through supervised machine learning. The study creates a dataset based on the combination of EHR data with patient reported data from 2,571 patients in the Region of Southern Denmark. After that, the dataset is labeled into two categories, AUD positive (457) and AUD negative (2,114). This unique dataset is used to validate the proposed method for prediction of AUD using machine learning methods based on historical clinical data from EHRs.

show abstract

Section: Resultsmentioning

confidence: 99%

A Predictive Machine Learning Model to Determine Alcohol Use Disorder

Ebrahimi

Wiil

Andersen

et al. 2020

2020 IEEE Symposium on Computers and Communications (ISCC)

View full text Add to dashboard Cite

show abstract

“…Another challenge is the dataset imbalance especially when trying to encode all 4-character codes as finding sufficient representatives for all classes is very difficult. Applying extreme text classification approaches to this task like splitting feature spaces or compressing label dimension helps reduce the imbalance effects [18].…”

Section: Related Workmentioning

confidence: 99%

“…The text classification corpus consists of 345,591 diagnoses and their corresponding ICD-10 codes. We apply data pre-processing on the diagnosis text by stemming using bulstempy 18 and removing the stop words using the BTB list 19 . We split the dataset using stratification and assign 80% for training, 10% for hyper-parameter tuning and 10% for testing using scikit-learn library 20 .…”

Section: Fine-tuning Bert For Icd-10 Classification Taskmentioning

confidence: 99%

“…As input for the fine-tuning procedure we tokenize the input diagnosis and prepend and append the special tokens [CLS] and 17 https://towardsdatascience.com/perplexity-intuition-and-derivation-105dd481c8f3 18 https://github.com/mhardalov/bulstem-py 19 http://bultreebank.org/wp-content/uploads/2017/04/BTB-StopWordList.zip 20 https://scikit-learn.org/stable/ [SEP] around it, as described in the original BERT paper [13].…”

Section: Fine-tuning Bert For Icd-10 Classification Taskmentioning

confidence: 99%

See 1 more Smart Citation

Automatic ICD-10 codes association to diagnosis: Bulgarian case

Velichkov

Gerginov

Panayotov

et al. 2020

CSBio '20: Proceedings of the Eleventh International Conference on Computational Systems-Biology and Bioinformatics

View full text Add to dashboard Cite

This paper presents an approach for the automatic association of diagnoses in Bulgarian language to ICD-10 codes. Since this task is currently performed manually by medical professionals, the ability to automate it would save time and allow doctors to focus more on patient care. The presented approach employs a fine-tuned language model (i.e. BERT) as a multi-class classification model. As there are several different types of BERT models, we conduct experiments to assess the applicability of domain and language specific model adaptation. To train our models we use a big corpora of about 350,000 textual descriptions of diagnosis in Bulgarian language annotated with ICD-10 codes. We conduct experiments comparing the accuracy of ICD-10 code prediction using different types of BERT language models. The results show that the MultilingualBERT model (Accuracy Top 1-81%; Macro F1-86%, MRR Top 5-88%) outperforms other models. However, all models seem to suffer from the class imbalance in the training dataset. The achieved accuracy of prediction in the experiments can be evaluated as very high, given the huge amount of classes and noisiness of the data. The result also provides evidence that the collected dataset and the proposed approach can be useful in building an application to help medical practitioners with this task and encourages further research to improve the prediction

show abstract

“…Generally speaking the effectiveness of the classifier fully depends on the characteristics of the data to be categorized. Application of classification varies across different dimensions in the arena of engineering and science such as: text categorization [1], biological classification [2], natural language processing [3], document classification [4], internet search engine [5], pattern recognition [6], medical imaging [7], handwritten character recognition [8], micro-array classification [9], voice classification [10], gene expression classification [11]. Major classifiers available in the literature are Bayesian classifier [12], Support vector machine [13], K-nearest neighbor [14], Decision tree [15] and Artificial Neural Network [16].…”

Section: Introductionmentioning

confidence: 99%

How Effective is Spotted Hyena Optimizer for Training Multilayer Perceptrons

Panda¹,

Majhi²

2019

IJRTE

View full text Add to dashboard Cite

This paper focuses on training multilayer perceptron (MLP) using a recently proposed meta-heuristic algorithm termed as Spotted Hyena Optimizer (SHO). To test the efficacy of the said algorithm fifteen standard datasets are used. At the same time the result of the proposed method is examined by some popular heuristic training algorithms such as Differential Evolution (DE), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Salp Swarm Algorithm (SSA) and Grey Wolf Optimization algorithm (GWO). Final result shows that SHO successfully avoids the local minima trap problem, simultaneously showing higher accuracy in classification as compared to other meta-heuristic methods. The statistical significance of the proposed SHO-MLP has been verified by deploying the Friedman & Holm’s test. It has been observed that the SHO-MLP is giving promisingly better result than other compared method for training MLP.

show abstract

On Feature Weighting and Selection for Medical Document Classification

Cited by 19 publications

References 28 publications

A Predictive Machine Learning Model to Determine Alcohol Use Disorder

A Predictive Machine Learning Model to Determine Alcohol Use Disorder

Automatic ICD-10 codes association to diagnosis: Bulgarian case

How Effective is Spotted Hyena Optimizer for Training Multilayer Perceptrons

Contact Info

Product

Resources

About