Identifying health-related discussions of cannabis use on Twitter: a proof-of-concept study (Preprint)

BACKGROUND The cannabis product and regulatory landscape is changing in the United States. Against the backdrop of these changes, there have been increasing reports on health-related motives for cannabis use and of adverse events from its use. The use of social media data in monitoring cannabis-related health conversations may be useful to state and federal-level regulatory agencies as they grapple with identifying cannabis safety signals in a comprehensive and scalable fashion. OBJECTIVE This study attempted to determine the extent to which a medical dictionary, the Unified Medical Language System (UMLS) Consumer Health Vocabulary (CHV), could identify cannabis-related motivations of use and health consequences of its use as discussed on Twitter in 2020. METHODS Twitter posts containing cannabis-related terms were obtained from January 1 to August 31, 2020. Each post from the sample (n = 353,353) was classified into at least one of 17 a priori categories of commonly health-related topics, using a rule-based classifier with each category defined by the terms in the medical dictionary. A subsample of posts (n=1094) was then manually annotated to help validate the rule-based classifier and determine if each post pertained to health-related motivations for cannabis use or perceived adverse health effects from its use or neither. RESULTS The validation process suggested that the medical dictionary could identify health-related conversations in 31.2% of posts. Specifically, 20.4% of posts were accurately identified as relating to a health-related motivation for cannabis use, while 10.8% of posts were accurately identified as relating to a health-related consequence from cannabis use. Potential health-related conversations around cannabis use ranged from issues with the respiratory system and stress to the immune system and gastrointestinal problems, among other health topics. CONCLUSIONS The mining of social media data may prove helpful in improving surveillance of cannabis products and their adverse health effects. However, future research needs to develop and validate a dictionary and codebook that captures cannabis use-specific health conversations on Twitter.

A Co-optimization PSO for Fuzzy Rule-Based Classifier Design Problem Based on Enlarged Hedge Algebras

Fuzzy Rule-Based Classifier (FRBC) design problem has been widely studied due to many practical applications. Hedge Algebras based Classifier Design Methods (HACDMs) are the outstanding and effective approaches because these approaches based on a mathematical formal formalism allowing the fuzzy sets based computational semantics generated from their inherent qualitative semantics of linguistic terms. HACDMs include two phase optimization process. The first phase is to optimize the semantic parameter values by applying an optimization algorithm. Then, in the second phase, the optimal fuzzy rule based system for FRBC is extracted based on the optimal semantic parameter values provided by the first phase. The performance of FRBC design methods depends on the quality of the applied optimization algorithms. This paper presents our proposed co-optimization Particle Swarm Optimization (PSO) algorithm for designing FRBC with trapezoidal fuzzy sets based computational semantics generated by Enlarged Hedge Algebras (EHAs). The results of experiments executed over 23 real world datasets have shown that Enlarged Hedge Algebras based classifier with our proposed co-optimization PSO algorithm outperforms the existing classifiers which are designed based on Enlarged Hedge Algebras methodology with two phase optimization process and the existing fuzzy set theory based classifiers.

Classification of arrhythmia’s ECG signal using cascade transparent classifier

Arrhythmia is a disease often encountered in patients with cardiac problems. The presence of arrhythmia can be detected by an electrocardiogram (ECG) test. Automatic observation based on machine learning has been developed for long time. Unfortunately, only few of them have capability of explaining the knowledge inside themselves. Thus, transparency is important to improve human understanding of knowledge. To achieve this goal, a method based on cascaded transparent classifier is proposed, a method was prepared. Firstly, ECG signals were separated and every single signal was extracted using feature extraction method. Several of extracted feature’s attributes were selected, and the final step was classifying data using cascade classifier which consists of decision tree and the rule based classifier. Classification performance was evaluated with publicly available dataset, the MIT-BIH Physionet Dataset. The methods were tested using 10-fold cross validation. The average of both accuracy and number of rules generated was considered. The best result using rule-based classifier achieves the accuracy and the number of rules 92.40% and 40, respectively. And the best result using cascade classifier achieves the accuracy and the number of rules 92.84% and 80, respectively. As a conclusion, transparent classifier shows a competitive performance with reasonable accuracy compared with previous research and promising in addressing the need for interpretability model.

Fuzzy Rule-Based Layered Classifier and Entropy-Based Feature Selection for Intrusion Detection System

Intrusion detection systems must detect the vulnerability consistently in a network and also perform efficiently with the huge amount of traffic. Intrusion detection systems must be capable of detecting emerging and proactive threats in the networks. Various classifiers are used to classify the threats as normal or intrusive by supervising the system activity. In this chapter, layered fuzzy rule-based classifier is proposed to detect the various intrusions, and fuzzy entropy-based feature selection is proposed to identify the relevant features. Layered fuzzy rule-based classifier is proposed to improve the performance of the intrusion detection system. KDD dataset contains various attacks; these attacks are grouped into four classes, namely Denial-of-Service (DoS), Probe, Remote-to-Local (R2L), and User-to-Root (U2R). Real-time dataset is also considered in this research. Experimental result shows that the proposed method provides good detection rate, minimizes the false positive rate, and less computational time.

Privacy Protection in Enterprise Social Networks Using a Hybrid De-Identification System

Enterprise social networks (ESN) have been widely used within organizations as a communication infrastructure that allows employees to collaborate with each other and share files and documents. The shared documents may contain a large amount of sensitive information that affect the privacy of persons such as phone numbers, which must be protected against any kind of disclosure or unauthorized access. In this study, authors propose a hybrid de-identification system that extract sensitive information from textual documents shared in ESNs. The system is based on both machine learning and rule-based classifiers. Gradient boosted trees (GBTs) algorithm is used as machine learning classifier. Experiments ran on a modified CoNLL 2003 dataset show that GBTs algorithm achieve a very high F1-score (95%). Additionally, the rule-based classifier is consisted of regular expression and gazetteers in order to complement the machine learning classifier. Thereafter, the sensitive information extracted by the two classifiers are merged and encrypted using Format Preserving Encryption method.

Gaussian Mixture Models for Detecting Sleep Apnea Events Using Single Oronasal Airflow Record

Sleep apnea is a common sleep-related disorder that significantly affects the population. It is characterized by repeated breathing interruption during sleep. Such events can induce hypoxia, which is a risk factor for multiple cardiovascular and cerebrovascular diseases. Polysomnography, the gold standard, is expensive, inaccessible, uncomfortable and an expert technician is needed to score sleep-related events. To address these limitations, many previous studies have proposed and implemented automatic scoring processes based on fewer sensors and machine learning classification algorithms. However, alternative device technologies developed for both home and hospital still have limited diagnostic accuracy for detecting apnea events even though many of the previous investigational algorithms are based on multiple physiological channel inputs. In this paper, we propose a new probabilistic algorithm based on (only) oronasal respiration signal for automated detection of apnea events during sleep. The proposed model leverages AASM recommendations for characterizing apnea events with respect to dynamic changes in the local respiratory airflow baseline. Unlike classical threshold-based classification models, we use a Gaussian mixture probability model for detecting sleep apnea based on the posterior probabilities of the respective events. Our results show significant improvement in the ability to detect sleep apnea events compared to a rule-based classifier that uses the same classification features and also compared to two previously published studies for automated apnea detection using the same respiratory flow signal. We use 96 sleep patients with different apnea severity levels as reflected by their Apnea-Hypopnea Index (AHI) levels. The performance was not only analyzed over obstructive sleep apnea (OSA) but also over other types of sleep apnea events including central and mixed sleep apnea (CSA, MSA). Also the performance was comprehensively analyzed and evaluated over patients with varying disease severity conditions, where it achieved an overall performance of TPR=88.5%, TNR=82.5%, and AUC=86.7%. The proposed approach contributes a new probabilistic framework for detecting sleep apnea events using a single airflow record with an improved capability to generalize over different apnea severity conditions