Text classification for automatic detection of e-cigarette use and use for smoking cessation… – europe pmc article – europe pmc
Survey results are necessary to understand usage trends, establish national and regional health goals and inform regulations and prevention campaigns. Smoking cessation counseling guidelines These surveys – while excellent in many ways – have several limitations. Benefits of quitting smoking cdc First, there is a time lag before new products of abuse are incorporated into the surveys.[ 8] For example, neither the BRFSS,[ 9] the National Health Interview Survey,[ 10] nor the National Survey on Drug Use and Health (NSDUH)[ 11] ask about e-cigarette use yet. Smoking cessation drugs chantix Second, the time lag in collection and analysis may delay timely policy interventions. Laser smoking cessation halifax Third, the surveys are sized to capture general trends across demographics and may lack focus for specific populations. Smoking cessation drugs zyban Fourth, surveys have limitations in detecting usage by minors as most are not allowed to take the surveys. Benefits of quitting smoking cigarettes Fifth, surveys may contain limited content for any specific question as every additional question competes against other questions for time and space in the survey. Smoking cessation tips Sixth, surveys capture high level geo-located information of use. Laser therapy for smoking cessation Continuing use of high-quality national surveys to inform prevention and treatment services is critical, yet new technologies may address some of these limitations.
An ideal surveillance solution could capture new drugs of abuse, collect data in real time, focus on populations of interest, include populations unable to take the survey, allow a breadth of questions to answer, and enable geo-location analysis.
Smoking cessation training for nurses We believe that social media streams may provide one solution. Laser smoking cessation barrie Social media, in this case, specifically Twitter, may include up to date vernacular for drugs of abuse, is inherently real time in how Tweets are broadcast, includes many potential populations of interest and their demographic characteristics, has populations such as minors who may not qualify for surveys, contains Tweets that indicate other potentially risky behaviors, and includes geo-locations. Benefits of quitting smoking using the patch To realize using social media for surveillance, a foundational question is whether we can detect drug use at all. Smoking cessation with laser therapy This work addresses this foundational concern and reports two pilot tasks for e-cigarettes. Smoking cessation clinical practice guidelines In the first, we identify automatically e-cigarette Tweets that indicate e-cigarette use. Smoking cessation best practice guidelines In the second we identify automatically Tweets that indicate e-cigarette use for smoking cessation.
Among social media platforms, Twitter offers unique potential to serve as a tool for tracking substance use. Smoking cessation guidelines 2012 Twitter is a micro-blogging service (with posts limited to 140 characters) through which users can send messages to a set of followers. Benefits of quitting smoking by day It has over 600 million users worldwide with 46% of users logging on daily. Smoking cessation guidelines 2016 In a recent Pew Research survey conducted August-September 2013, 18% of US adults use Twitter. Icd 9 smoking cessation counseling [ 12] A higher percentage of Blacks/African-Americans (29%) use Twitter compared with Whites (16%) and Hispanics (16%). Laser therapy smoking cessation Of Twitter subscribers, 31% are 18-29 and 19% are 30-49 years old.[ 12] Interestingly, there are relatively no differences in use by education level, gender, or income suggesting that use cuts across socioeconomic differences.
A few studies have specifically addressed e-cigarettes via Twitter. Benefits of smoking cessation for longevity Clark et al.[ 13] used 700,000 tweets collected from January 2012 to July 2014 to survey the general popularity and sentiment of consumer opinions regarding e-cigarettes.[ 14] In a follow up publication, they focused on approximately 20,000 geo-located tweets to characterize density and sentiment surrounding tobacco and e-cigarette tweets and link prevalence of word choices to tobacco and e-cigarette use at various localities.[ 14] In another publication, Huang et al.[ 15] labeled 73,672 tweets related to e-cigarettes to characterize how e-cigarettes are marketed, and Harris et al.[ 16] conducted a manual content analysis of tweets related to Chicago’s regulation of e-cigarettes. Smoking cessation and weight gain While these studies produced a one-time picture of e-cigarette sentiment, neither the methodology of identifying e-cigarettes with a simple term search nor using manual coding are useful for ongoing surveillance purposes. Laser acupuncture for smoking cessation reviews A system that harnesses social media posts could serve as a low-cost method of examining usage trends and attitudes toward particular products.
In these previous studies, a common theme for analysis is the manual labeling of tweets. How does laser smoking cessation work Manual labeling requires (1) time, (2) expertise, and (3) consistency. Smoking cessation guidelines canada In addition, the samples must be small enough to allow feasible manual labeling which inherently is limited to the snapshot in time when the tweets are collected. Smoking cessation prescription drugs Our aim in this paper is to use text classification machine learning techniques to address these limitations and convert these manual classifications to automated classifications. Smoking cessation classes With tweet volumes of nearly 500 million per day, automation is the only realistic and feasible solution.
A challenge for building a labeled training corpus from Twitter is the low prevalence of Tweets in a target category. Laser therapy for smoking cessation reviews To enrich the e-cigarette use target category, we filtered Tweets by e-cigarette brand followers and hashtags. Smoking cessation counselor training Specifically, we downloaded a 28.6 million tweet collection in January 2015. Smoking cessation plan The tweet collection represents the tweets of 29,410 followers of the largest e-cigarette brands @v2cigs, @VaporFl, @HaloCigs, @bluecigs, @NJOYVape, @KRAVEeCig, and @LogicECig. Smoking cessation guidelines 2013 To further increase the probability of encountering tweets about e-cigarette use and e-cigarette use for smoking cessation, we filtered the 28.6 million tweets with the Boolean OR of #vape #mod_ #vapeing #vaping #flavhub #eliquid #ejuice #pureclass or #ecigs. Benefits of quitting smoking after 2 months This corpus had 5,435 Tweeters covering a time span from Jan 2010 to Jan 2015 representing 228,145 Tweets. Smoking cessation guidelines for australian general practice From these Tweets, we build a final corpus consisting of 13,146 randomly selected Tweets to label for our classifiers as outlined in section 3.2. Cdc smoking cessation The remaining 214,999 Tweets were not used for this pilot work and limitations with labeler time constrained the number of labeled Tweets.
We employed a linear Support Vector Machine (SVM) classification algorithm as implemented in the liblinear package. Benefits of quitting smoking after 2 weeks The linear SVM’s calculate maximal margin hyperplane(s) separating the two classes of the data. National smoking cessation programs For text data, the linear SVMs demonstrated superior text classification performance compared to other methods [ 24], and this motivated our use of them. Smoking cessation counselling techniques The liblinear implementation is an optimized version of the support vector machine optimized for quickly finding a linear separating hyperplane. American heart association smoking cessation training We used liblinear as implemented in libSVM v1.96. Non nicotine smoking cessation drugs [ 25] We used the default solver of L2-regularized L2-loss and the default penalty parameter of 1.
We compared the machine learning models to a simple keyword based approach for identifying tweets. Benefits of quitting smoking sexually Based on our definition, we asked PK to look at our protocol and the portion of the dataset that he reviewed and generate a Boolean keyword set that would provide a relative non machine learning baseline for this classification task. Symptoms of smoking cessation We added this analysis to address whether this task is difficult and to counter the claim that a human could craft a wordset that performs as well as the machine learning models. Smoking cessation side effects To simplify the comparison, we compare the keywords to the “unigram” encoding in one 10% split of the data. What is a smoking cessation program We used the keyword “OR” searches shown in Table 5.
Keyword based searches are inferior to the machine learning methods. 5 benefits of quitting smoking The keyword search returns a sensitivity and specificity while the machine learning methods return a ranked result. Smoking cessation training for health professionals To make the comparison, we take one split from the 10 fold cross validation and obtain the sensitivity and specificity for the keyword search. Level 1 smoking cessation training The keyword search has a sensitivity of 0.75 and a specificity of 0.36. Laser smoking cessation reviews At 0.75 sensitivity, the best learning algorithm performs at 0.87 specificity (compared to 0.36). Laser acupuncture for smoking cessation At 0.36 specificity, the best algorithm performs at 0.99 sensitivity (compared to 0.75). Smoking cessation drugs side effects Task 1 of defining e-cigarette use through a keyword search benefits from machine learned models.
E-cigarette Use for Smoking Cessation – 10 Fold Cross Validation Area Under the Receiver Operating Curve (AUC) Performance (Range of Performances Across 10 folds)
These results highlight performance differences in encoding the tweets. Laser smoking cessation calgary Using unigram or bigram representations demonstrate little performance differences within each classifier. Smoking cessation guidelines evidence based recommendations Including stopwords increases performances as shown by comparison between the top 4 and bottom 4 rows that reflect keeping and removing stopwords respectively. Benefits of quitting smoking 1 year Stemming seems to have a marginal effect in this classification task.
In contrast to the previous task, the ranges across the 10 folds are wide. Smoking cessation effects These results likely reflect the small positive sample size of 73 in this dataset (even less in each fold) and the suspected heterogeneity (e.g. Smoking cessation definition the many ways of communicating e-cigarette use for smoking cessation) in this labeled task.
For both tasks, retaining stopwords improves performance. What quitting smoking does to your skin This observation runs contrary to most other text classification tasks where removing stopwords typically does not affect performance. Smoking cessation nice Stopwords can make a difference and prior researchers have shown that these words can affect performance depending on the task. Smoking cessation support [ 31] Further study is needed to examine which stopwords are important for classification in these tasks.
Keyword based searches are inferior to the machine learning methods. Maudsley smoking cessation training The keyword search returns a sensitivity and specificity while the machine learning methods return a ranked result. Smoking cessation meaning To make the comparison, we take one split from the 10 fold cross validation and obtain the sensitivity and specificity for the keyword search. Smoking cessation treatment guidelines The keyword search has a sensitivity of 0.29 and a specificity of 0.99. Smoking cessation counseling cpt At 0.29 sensitivity, the best learning algorithm performs at 0.99 specificity (compared to 0.99). Define smoking cessation At 0.99 specificity, the best algorithm performs at 0.37 sensitivity (compared to 0.29). Benefits of quitting smoking first 24 hours Task 2 of defining e-cigarette use for smoking cessation through a keyword search is not trivial and benefits from machine learned models.