scholarly journals Behavior Features for Automatic Detection of Depression from Facebook Users

2020 ◽  
Author(s):  
Siranuch Hemtanon ◽  
Saifon Aekwarangkoon ◽  
Nichnan Kittphattanabawon
Keyword(s):  
At Risk ◽  
Depressive Disorder ◽  
Learning Algorithm ◽  
Common Mental Disorders ◽  
Late Night ◽  
Deep Learning Algorithm ◽  
Standard Tool ◽  
Active Data ◽  
Detection Of Depression ◽  
F Measure

Major depressive disorder is one of common mental disorders globally. It is best to be early detected and cured. This work introduces a method to detect depressive disorder at risk via a behavior made on Facebook platform. A set of features related to Facebook main functions including amount of posting, sharing, commenting and replying is designed to represent users’ activities in a numerical value form. The collected data with periodic and consecutive aspects are gathered without interpreting content. Thus, the data are easier to be collected with less privacy issue. To distinct between positive and negative depression-at risk, PHQ-9 questionnaire, a standard tool commonly used to screen depression patient in Thailand, was used. These features hence are used in supervised learning classification algorithm for detecting a risk of being depressive disorder. From the experiment of 160 Thai Facebook users, the statistical result indicated that depression-positive users tend to do consecutive actions and rarely reply to other comments. Moreover, they often have activities in late night. The classification experiment shows that the designed features based on users’ activities from Facebook with deep learning algorithm yields about 87% in terms of F-measure. After analyzing the data, we thus split data regarding users’ gender and removed obviously low active data, and the F-measure from classification went up to 91.4 which improves for 4 points.

A Prediction Approach Based on Self-Training and Deep Learning for Biological Data

2020 ◽  
Vol 10 (4) ◽  
pp. 50-64
Author(s):  
Mohamed Nadjib Boufenara ◽  
Mahmoud Boufaida ◽  
Mohamed Lamine Berkane
Keyword(s):  
Deep Learning ◽  
Supervised Learning ◽  
Learning Algorithm ◽  
Classification Performance ◽  
Biological Data ◽  
Unlabeled Data ◽  
Learning Methods ◽  
Deep Learning Algorithm ◽  
Prediction Approach ◽  
F Measure

With the exponential growth of biological data, labeling this kind of data becomes difficult and costly. Although unlabeled data are comparatively more plentiful than labeled ones, most supervised learning methods are not designed to use unlabeled data. Semi-supervised learning methods are motivated by the availability of large unlabeled datasets rather than a small amount of labeled examples. However, incorporating unlabeled data into learning does not guarantee an improvement in classification performance. This paper introduces an approach based on a model of semi-supervised learning, which is the self-training with a deep learning algorithm to predict missing classes from labeled and unlabeled data. In order to assess the performance of the proposed approach, two datasets are used with four performance measures: precision, recall, F-measure, and area under the ROC curve (AUC).

scholarly journals Implant Term Extraction from Swedish Medical Records – Phase 1: Lessons Learned

2021 ◽  
Author(s):  
Oskar Jerdhaf ◽  
Marina Santini ◽  
Peter Lundberg ◽  
Anette Karlsson ◽  
Arne Jönsson
Keyword(s):  
At Risk ◽  
Medical Records ◽  
Learning Algorithm ◽  
Phase 1 ◽  
Lessons Learned ◽  
Automatic Identification ◽  
Deep Learning Algorithm ◽  
Term Extraction ◽  
Swedish Medical ◽  
Magnetic Resonance Imaging Mri

We present the case of automatic identification of “implant terms”. Implant terms are specialized terms that are important for domain experts (e.g. radiologists), but they are difficult to retrieve automatically because their presence is sparse. The need of an automatic identification of implant terms spurs from safety reasons because patients who have an implant may be at risk if they undergo Magnetic Resonance Imaging (MRI). At present, the workflow to verify whether a patient could be at risk of MRI side-effects is manual and laborious. We claim that this workflow can be sped up, streamlined and become safer by automatically sieving through patients’ medical records to ascertain if they have or have had an implant. To this aim we use BERT, a state-of-the-art deep learning algorithm based on pre-trained word embeddings and we create a model that outputs term clusters. We then assess the linguistic quality or term relatedness of individual term clusters using a simple intra-cluster metric that we call cleanliness. Results are promising.

Sign in / Sign up

Export Citation Format

Share Document