Search

Community detection is an important area of research in social media mining. Numerous algorithms have been developed to detect disjoint, overlapping, and dynamic communities in a network. However, most of the existing algorithms do not consider the influence of each node in a group, which may be different and can affect the result of community detection. In this paper, we have proposed a novel Louvain-based algorithm named “NI-Louvain,” which considers the influence of each node in a group that not only detects community but also most influential nodes in the community. The proposed algorithm works in three steps. First, the input graph is processed to reduce its density. This is done by calculating cliques from the graph. In second step, Louvain's multilevel algorithm is applied to this clique graph. The resultant community obtained has a better modularity value than the communities obtained from existing algorithms like edge betweenness, label propagation, leading eigen, Louvain, fast greedy, walktrap, and infomap. In the third step, the nodes of resultant communities are transformed back to the nodes of the original graph. This step is beneficial in detecting overlapping communities, fuzzy membership of each node, most influential node in each community formed, and outlier nodes.

Alter ego is a condition of someone who creates a new character with a conscious state. Original character role play is a game to create new imaginary characters that is used as research material for identification alter ego accounts. The negative effects of playing alter ego are stress, depression, and multiple personalities. Current research only focuses on the phenomenon and impacts of a role-playing game. We propose a new method to detect accounts of alter ego players in social media, especially Twitter. We develop an application to analyze the characteristics of alter ego accounts.Psychologists can use this application to discover the characteristics of alter ego accounts that are useful for analyzing personality so that the results can be used to appropriately handle alter ego players. Most user profiles, tweets, and platforms are used to detect account Twitter. This research proposes a new method using bio features as input data. We crawled and collected 565 bios from Twitter for one month. We observe the data to search for unique words and collect them into a classification dictionary. In this research, we use the cosine similarity method because this method is popular for detecting text and has a good performance in many cases. This research could identify alter ego accounts and other types of Twitter accounts.From the detection results of alter ego accounts, it is possible to analyze the characteristics of Twitter accounts. We use a sampling technique that takes 30% of the data as testing data. According to the results of the experiment cosine similarity obtained an accuracy of 0.95. &nbsp, Alter ego is a condition of someone who creates a new character with a conscious state. Original character role play is a game to create new imaginary characters that is used as research material for identification alter ego accounts. The negative effects of playing alter ego are stress, depression, and multiple personalities. Current research only focuses on the phenomenon and impacts of a role-playing game. We propose a new method to detect accounts of alter ego players in social media, especially Twitter. We develop an application to analyze the characteristics of alter ego accounts.Psychologists can use this application to discover the characteristics of alter ego accounts that are useful for analyzing personality so that the results can be used to appropriately handle alter ego players. Most user profiles, tweets, and platforms are used to detect account Twitter. This research proposes a new method using bio features as input data. We crawled and collected 565 bios from Twitter for one month. We observe the data to search for unique words and collect them into a classification dictionary. In this research, we use the cosine similarity method because this method is popular for detecting text and has a good performance in many cases. This research could identify alter ego accounts and other types of Twitter accounts.From the detection results of alter ego accounts, it is possible to analyze the characteristics of Twitter accounts. We use a sampling technique that takes 30% of the data as testing data. According to the results of the experiment cosine similarity obtained an accuracy of 0.95.

Purpose Treatment with the tyrosine kinase inhibitor (TKI) imatinib in patients with gastrointestinal stromal tumours (GIST) causes symptoms that could negatively impact health-related quality of life (HRQoL). Treatment-related symptoms are usually clinician-reported and little is known about patient reports. We used survey and online patient forum data to investigate (1) prevalence of patient-reported symptoms; (2) coverage of symptoms mentioned on the forum by existing HRQoL questionnaires; and (3) priorities of prevalent symptoms in HRQoL assessment. Methods In the cross-sectional population-based survey study, Dutch GIST patients completed items from the EORTC QLQ-C30 and Symptom-Based Questionnaire (SBQ). In the forum study, machine learning algorithms were used to extract TKI side-effects from English messages on an international online forum for GIST patients. Prevalence of symptoms related to imatinib treatment in both sources was calculated and exploratively compared. Results Fatigue and muscle pain or cramps were reported most frequently. Seven out of 10 most reported symptoms (i.e. fatigue, muscle pain or cramps, facial swelling, joint pain, skin problems, diarrhoea, and oedema) overlapped between the two sources. Alopecia was frequently mentioned on the forum, but not in the survey. Four out of 10 most reported symptoms on the online forum are covered by the EORTC QLQ-C30. The EORTC-SBQ and EORTC Item Library cover 9 and 10 symptoms, respectively. Conclusion This first overview of patient-reported imatinib-related symptoms from two data sources helps to determine coverage of items in existing questionnaires, and prioritize HRQoL issues. Combining cancer-generic instruments with treatment-specific item lists will improve future HRQoL assessment in care and research in GIST patients using TKI.

Location-based social media data can offer useful insights on the spatial and temporal dynamics of public attitudes. In this study, we aim to investigate the gendered attitudes toward transit services in China, utilizing the case of Shenzhen. We collected 44,257 Weibo microblogs, a major source of social media data in China, and applied a series of text mining and visualization techniques to examine the gender differences among our focused themes. The microblogs reveal a distinct gender gap in terms of quantity, as nearly 74% are posted by women. While women tend to be more concerned about the comfort of transit environment (e.g., temperature, crowdedness, and safety, especially at night), men tend to be more interested in transit systems’ e-payment services and reporting traffic incidents. Overall, this study presents a methodological framework and empirical case study about how we can utilize certain social media mining techniques to investigate gendered, subjective travel experiences, providing researchers and practitioners with an innovative way to gather customer service feedback and build more inclusive service systems.

[Resumen] Actualmente, las capacidades de interacción de los robots sociales son limitadas y, después de un tiempo de convivencia, los diálogos que pueden mantener son percibidos como predecibles, repetitivos y poco naturales. Esto puede llevar a una pérdida de interés de la persona en el robot. Si se quiere apostar por una convivencia exitosa y larga, es necesario dotar los robots de discursos más variados y que se adapten fácilmente a los cambios de necesidades que pueda haber en los usuarios. En esta contribución se propone una metodología que combina técnicas de minería de datos y de aprendizaje automático para, mediante el contenido publicado en las redes sociales, definir la comunicación verbal del robot de una forma dinámica. Se propone extraer información útil de las redes sociales para construir modelos de conocimiento basados en temas que son de interés para el usuario y en el contexto de la interacción con el robot. Además, se plantea mantener el modelo actualizado de acuerdo con la nueva información que se publique o cambios que pudieran ocurrir desde un punto de vista del usuario. [Abstract] Currently, the interaction capabilities of social robots are limited and, especially in long-term interactions, the human-robot dialogues can be perceived as predictable, repetitive, and unnatural. This situation can lead the user to lose interest in the robot. If we want to bet on a successful and long-lasting coexistence, it is necessary to provide robots with more varied discourses that can adapt to the changing needs of the users. This contribution proposes a methodology that combines data mining and machine learning techniques to dynamically define the robot’s verbal communication through the content published on social networks. We propose to extract useful information from social networks to build knowledge models based on the context of the interaction so that it changes according to the published information. Acción financiada por la Comunidad de Madrid a través de la línea ”Estímulo a la Investigación de Jóvenes Doctores/as” del Convenio Plurianual con la UC3M (SMM4HRI-CM-UC3M), en el marco del V PRICIT (V Plan Regional de Investigación Científica e Innovación Tecnológica. Comunidad de Madrid; SMM4HRI-CM-UC3M

Since its emergence, neuromarketing has generated multiple debates about its uses and misuses, the ethics of this highly sensitive data technique, and the lack of regulation for consumer protection. To explore how consumers perceive neuromarketing, a content analysis of user-generated content (UGC) on blogs and social media was conducted. The main purpose of this research was to identify consumers’ opinions about neuromarketing, its perceived benefits and risks, and the ethical concerns and implications for those involved in the use of this method. Taking a social media mining approach, the study uncovers the most common neuromarketing concerns discussed by consumers, describes the main participants (or actors) of the neuromarketing narrative and their interrelationships, and evaluates the overall consumer sentiment of neuromarketing. Six main themes and eight actors emerged from the exploratory analysis. Furthermore, the sentiment analysis demonstrated a positive perception of neuromarketing among consumers, but it also revealed some concerns about the lack of regulation and consumer privacy. The study concludes with a discussion of implications for future research.

This study suggests a social media mining approach for monitoring critical complaints in the early stage of product launch via word embedding, clustering and sentiment analysis. This approach selectively uses social media product reviews created in the beginning stage of product launch. Word embedding and clustering are used to identify customer complaint factors from reviews. Next, sentence-level sentiment analysis is conducted based on the understanding of review writing pattern, thereby calculating urgency and severity in a quantitative manner. Finally, a customer-stated complaint portfolio map is developed, showing critical complaint factors. In the case study of smartphone ‘Galaxy S10’, the critical complaint factors related to Samsung pay connection error, and fingerprint recognition are identified, which have been rapidly improved since the product launch, indeed. We expect that this approach contributes to identifying critical customer complaints in the early stage, in addition to enabling the prompt response to customer complaints.

La crisis sanitaria de la covid-19 vino acompañada de medidas de aislamiento, entre las que se encontraba el confinamiento domiciliario, que provocaron múltiples reacciones en las redes sociales. El objetivo de este artículo consiste en analizar la conversación digital observada en Twitter®, en torno al hashtag #MeQuedoEnCasa, en el periodo comprendido entre el 20 y el 27 de marzo de 2020. El estudio parte de una metodología mixta, en la que se combinan técnicas de análisis del social media mining con estrategias cualitativas propias del análisis lingüístico. Desde el punto de vista teórico, nos apoyamos en conceptos de las teorías del encuadre y de la valoración. Los resultados apuntan al enmarque positivo del confinamiento, a través de rasgos semióticos de distinto nivel: léxico, semántico y pragmático. Las cuentas más influyentes inclinaron su producción discursiva hacia la polaridad positiva. El análisis empírico permite concluir que el encuadre discursivo de esta conversación digital combina dos ejes semánticos (colectividad y salud), dos ejes enunciativos (aquí y ahora) y un eje emocional, lo que implica que hashtags como #MeQuedoEnCasa funcionan como señas de identidad social, como marcas de anclaje enunciativo y como instrumentos para fomentar la responsabilidad del individuo desde valores positivos.

Purpose—Using a sample of universities from Europe and North America the research herein seeks to understand the content trends of university brand pages through an exploration of the social community and the measurement of user participation and behavior. The analysis relies on an artificial intelligence approach. Through the verification of interactions between users and content on the university brand pages, recommendations are made, which aim to ensure the pages meet the needs of users in the future. Design/methodology/approach—The study sample was drawn from six well-known universities in Europe and North America. The content of 23,158 posts made over the course of nine years between 1 January 2011 to 31 December 2019 was obtained by a web crawler. Concepts in the fields of computer science, data mining, big data and ensemble learning (Random Decision Forests, eXtreme Gradient Boosting and AdaBoost) were combined to analyze the results obtained from social media exploration. Findings—By exploring community content and using artificial intelligence analysis, the research identified key information on the university brand pages that significantly affected the cognition and behavior of users. The results suggest that distinct levels of user participation arise from the use of different key messages on the university fan page. The interactive characteristics identified within the study sample were classified as one of the following module-types: (a) information and entertainment satisfaction module, (b) compound identity verification module or (c) compound interactive satisfaction module. Research limitations/implications—The study makes a contribution to the literature by developing a university community information interaction model, which explains different user behaviors, and by examining the impact of common key (image) clues contained within community information. This work also confirms that the behavioral involvement of users on the university’s brand page is closely related to the information present within the university community. A limitation of the study was the restriction of the sample to only European and North American cultural and economic backgrounds and the use of Facebook as the sole source of information about the university community. Practical implications—Practically, the research contributes to our understanding of how, in official community interactions, user interactions can be directed by features such as information stimuli and brand meanings. In addition, the work clarifies the relationship between information and user needs, explaining how the information characteristics and interaction rules particular to a given school can be strengthened in order to better manage the university brand page and increase both the attention and interaction of page users. Originality/value—This research provides an important explanation of the role of key information on the university fan pages and verifies the importance of establishing key (image) clues in the brand community, which in turn affect user cognition and interaction. Although related research exists on information manipulation and the importance of online communities, few studies have directly discussed the influence of key information on the fan pages of university brands. Therefore, this research will help to fill gaps in the literature and practice by examining a specific context, while at the same time providing a valuable and specific reference for the community operation and management of other related university brands.

Social networks have taken an irreplaceable role in our lives. They are used daily by millions of people to communicate and inform themselves. This success has also led to a lot of irrelevant content and even misinformation on social media. In this paper, we propose a user-centred framework to reduce the amount of irrelevant content in social networks to support further stages of data mining processes. The system also helps in the reduction of misinformation in social networks, since it selects credible and reputable users. The system is based on the belief that if a user is credible then their content will be credible. Our proposal uses word embeddings in a first stage, to create a set of interesting users according to their expertise. After that, in a later stage, it employs social network metrics to further narrow down the relevant users according to their credibility in the network. To validate the framework, it has been tested with two real Big Data problems on Twitter. One related to COVID-19 tweets and the other to last United States elections on 3rd November. Both are problems in which finding relevant content may be difficult due to the large amount of data published during the last years. The proposed framework, called NOFACE, reduces the number of irrelevant users posting about the topic, taking only those that have a higher credibility, and thus giving interesting information about the selected topic. This entails a reduction of irrelevant information, mitigating therefore the presence of misinformation on a posterior data mining method application, improving the obtained results, as it is illustrated in the mentioned two topics using clustering, association rules and LDA techniques., European Commission 786687, Andalusian government FEDER operative program P18-RT-2947 B-TIC-145-UGR18, University of Granada's internal plan PPJIB2021-04, Spanish Government FPU18/00150

Social Network Mental Disorder Detection” or “SNMD” is an approach to analyse data and retrieve sentiment that it embodies. Twitter SNMD analysis is an application of sentiment analysis on data from Twitter (tweets), in order to extract sentiments conveyed by the user. In this paper, we aim to review some papers regarding research in sentiment analysis on Twitter, describing the methodologies adopted and models applied, along with describing a generalized Python based approach. A prototype system is developed and tested.

Akıllı kentler hem doğal afetlere karşı dayanıklı olmak hem de turist çekme açısından avantajlar sunmaktadır. Bu çalışmada, ülkemizde akıllı kentler hakkındaki farkındalığın belirlemek için sosyal medya madenciliği yöntemi önerilmiştir. Akıllı kent hakkında yazılan tweetler düzenli olarak üç ay boyunca toplanmıştır. Daha sonra tweetlerdeki veriler kümeleme algoritması k-means ile üç (3) gruba ayrılmıştır. Kümelenmiş tweetler analiz edilerek en çok geçen kelime sayısı belirlenmiştir. Rapidminer yazılımı tweetlerin toplanmasında ve metin verilerinin düzenlenmesinde kullanılmıştır. Çalışma sonucunda, tweetlerde sırası ile birinci grupta: güç, altyapı, Gaziantep ikinci grupta: dünya, Ankara, dijital üçüncü grupta: memleket, mutluluk, sürekli en çok geçen kelimeler olarak tespit edilmiştir. En sık kullanılan kelimeler twitter kullanıcılarının akıllı şehirler konusunda bilgi sahibi olduğunu göstermektedir. Makalenin sosyal medya verilerinden akıllı kentler hakkındaki farkındalığın belirlenmesi açısından katkıda bulunacağı düşünülmektedir.

Covid-19 is an unprecedented disaster that is still difficult to contain During the pandemic, there were a lot of cases that were reported to increase exponentially In this situation, the dissemination of messages and information was very important The social media platform has contributed as a channel of communication with unprecedented speed However, the uncontrolled and irresponsible dissemination of information will result in new problems that can be detrimental to many parties A lot of information may trigger panic, fear and result in lose hope and even paranoia The provision of correct and timely information as well as any curative and preventive effort to stop the disease are very important This study aims to present a method in finding out public opinion through Twitter social media mining in the Indonesian context We are particularly interested in finding out what people’s stance with the pandemic Some people may fully aware of this threat, but the remaining could be careless about what is going on It is assumed that this stance could lead to people’s obedience to the government’s policy on COVID 19 Protocol It is believed that the opinion is hidden behind the comments in the media By scrapping the tweets on Twitter during March 2020 using Corona and COVID Keywords, we obtained as many as 31,003 tweets We manually classified the tweets into 3 classes, positive, negative and neutral stances Predictive models are derived using Support Vector Machine, Random Forest and Naive Bayes algorithms Random Forest-based model gives the highest accuracy level as high as 89%, followed by Support Vector Machine as high as 87% and Naive Bayes as high as 68% The model can further be used to classify opinions in the future giving valuable information for the government in making policies and steps in overcoming the pandemic © 2020, Faisal Binsar and Tuga Mauritsius

The IT service industry values the experience of social familiarity, which is based on routine interactions with suppliers and customers and is at the frontier of social media marketing. To further understand how IT service companies use social media to engage their customers or potential leads, the objectives of this research are to delineate the differences between IT service companies’ use of social media and users’ expectations based on knowledge extracted from user-generated content on Twitter. We applied a text mining approach called two-tier concept-linking analysis to extract patterns in Twitter posts from top IT service companies as well as the related tweets from the Twitter users. We further compare the yearly similarities and differences of the key concepts from the company’s official account and from the users. Besides, the key concepts between users’ expectations and IT service companies’ social media use were compared on the basis of first-tier concepts and further elaborated by their corresponding second-tier concepts. Our approach contributes to further understand the socio-cognitive insights underlying the dynamic social media data, from which organizations and individuals in charge can note the objectives they wish to maintain and the marketing gaps they wish to improve on.

Community detection is a dominant research field in social media mining. Numerous algorithms have been developed to detect community structure in network. Each of them focuses on different characte...

Coronavirus 2019, or COVID-19, is a contagious disease triggered by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). With origins in Wuhan, China, this disease has since spread globally, resulting in the ongoing 2019–2020 coronavirus pandemic. As of May 3, 2020, the Ministry of Health and Family Welfare confirmed a total of 39,980 positive COVID-19 cases and 1,301 deaths in India (more than 3.42 million positive COVID-19 cases resulting in more than 243,000 deaths worldwide). To flatten the curve, India has been locking down its country from March 24 to May 17, 2020. This study collected “COVID-19 in India” related tweets (totaling 410,643 tweets in English) from March 22 to April 21, 2020 to gauge the unknowns and contexts associated with public sentiments during the lockdown. This work contributes to the growing body of studies on COVID-19 social media mining by extracting emotions and sentiments over time, which could potentially shed some lights on the contexts of expressions during pandemic.

Recommendation of a relevant and suitable news article is an essential but a challenging task due to changes in the user interest categories over time. Moreover, the Internet technology provides abundant news articles from a huge amount of resources. Meanwhile, nowadays, many people are confronted with viral news articles through social media cost-free without considering the news sites. Therefore, mining of social media for addressing such viral news articles has become another key challenge. To overcome the above challenges, this paper proposes fuzzy logic approach for predicting users’ diversified interest and its categories by analysing their implicit user profile. Depending on users’ interest categories, the viral news articles and their categories were determined and analysed through mining social media feeds-Facebook and Twitter. Furthermore, fresh news articles are retrieved from news feeds incorporated with retrieved viral news articles provided as recommendation with respect to users’ diversified interest. The performance of the proposed approach for predicting overall users’ interest for all categories attained 84.238%, and recommendation accuracy from News feed, Facebook, and Twitter attained 100%, 90%, and 100% with respect to users’ interest categories.

With the exponential growth of users and user-generated content present on online social networks, fake news and its detection have become a major problem. Through these, smear campaigns can be generated, aimed for example at trying to change the political orientation of some people. Twitter has become one of the main spreaders of fake news in the network. Therefore, in this paper, we present a solution based on Text Mining that tries to find which text patterns are related to tweets that refer to fake news and which patterns in the tweets are related to true news. To test and validate the results, the system faces a pre-labelled dataset of fake and real tweets during the U.S. presidential election in 2016. In terms of results interesting patterns are obtained that relate the size and subtle changes of the real news to create fake news. Finally, different ways to visualize the results are provided.

The research reported in this paper was partially supported by the COPKIT project under the European Union's Horizon 2020 research and innovation program (grant agreement No 786687), the Andalusian government and the FEDER operative program under the project BigDataMed (P18-RT-2947 and B-TIC-145-UGR18). Finally the project is also partially supported by the Spanish Ministry of Education, Culture and Sport (FPU18/00150)., The incursion of social media in our lives has been much accentuated in the last decade. This has led to a multiplication of data mining tools aimed at obtaining knowledge from these data sources. One of the greatest challenges in this area is to be able to obtain this knowledge without the need for training processes, which requires structured information and pre-labelled datasets. This is where unsupervised data mining techniques come in. These techniques can obtain value from these unstructured and unlabelled data, providing very interesting solutions to enhance the decision-making process. In this paper, we first address the problem of social media mining, as well as the need for unsupervised techniques, in particular association rules, for its treatment. We follow with a broad overview of the applications of association rules in the domain of social media mining, specifically, their application to the problems of mining textual entities, such as tweets. We also focus on the strengths and weaknesses of using association rules for solving different tasks in textual social media. Finally, the paper provides a perspective overview of the challenges that association rules must face in the next decade within the field of social media mining., COPKIT project under the European Union's Horizon 2020 research and innovation program 786687, Andalusian government, FEDER operative program under the project BigDataMed P18-RT-2947 B-TIC-145-UGR18, Spanish Government FPU18/00150

This study explores the interactive characteristics of the public, referencing existing data mining methods. This research attempts to develop a community data mining and integration technology to investigate the trends of global retail chain brands. Using social media mining and ensemble learning, it examines key image cues to highlight the various reasons motivating participation by fans. Further, it expands the discussion on image and marketing cues to explore how various social brands induce public participation and the evaluation of information efficiency. This study integrates random decision forests, extreme gradient boost, and adaboost for statistical verification. From 1 January 2011 to 31 December 2019, the studied brands published a total of 25,538 posts. The study combines community information and participation in its research framework. The samples are divided into three categories: retail food brand, retail home improvement brand, and retail warehouse club brand. This research draws on brand image and information cue theory to design the theoretical framework, and then uses behavior response factors for the theoretical integration. This study contributes a model that classifies brand community posts and mines related data to analyze public needs and preferences. More specifically, it proposes a framework with supervised and ensemble learning to classify information users′ behavioral characteristics.

Technological development of the last several decades has driven open innovation towards organizational, business, social, and economic change. Open innovation has emerged as the main driver of change in a business sector that needs to be flexible and resilient, rapidly adapting to change through innovation. In this context, the present study aimed to explore the limits of open innovation by extracting evidence from user-generated content (UGC) on Twitter using social media mining. To this end, in terms of the methodology, we first applied machine learning Sentiment Analysis algorithm texted using Support Vector Classifier, Multinomial Naïve Bayes, Logistic Regression, and Random Forest Classifier to divide the sample of n = 586.348 tweets into three groups expressing the following three sentiments: positive, negative, and neutral. Then, we used a mathematical topic modeling algorithm known as Latent Dirichlet allocation to analyze the tweet databases. Finally, Python was used to develop textual analysis techniques under the theoretical framework of Computer-Aided Text Analysis and Natural Language Processing. The results revealed that, in the tweets dataset, there were eight topics. Of these topics, two contained tweets expressing negative sentiments (Culture and Business Models/Management), three topics contained tweets expressing positive sentiments (Communities, Creative projects and Ideas), and three topics contained tweets expressing neutral sentiments (Entrepreneurship, Teams and Technology). These topics are discussed in the context of limitations, risks, and characteristics of open innovation according to the UGC on Twitter. The paper concludes with the formulation of 20 limits of open innovation and 27 research questions for further research on open innovation, as well as a discussion of theoretical and practical implications of the study. UGCUser-Generated ContentUGDUser-Generated DataCATAComputer-Aided Text AnalysisNLPNatural Language ProcessingLDALatent Dirichlet AllocationSVCSupport Vector ClassifierMNBMultinomial Naïve BayesLRLogistic RegressionRFCRandom Forest Classifier

The outbreak of the novel coronavirus, COVID-19, has become one of the most severe pandemics in human history. In this paper, we propose to leverage social media users as social sensors to simultaneously predict the pandemic trends and suggest potential risk factors for public health experts to understand spread situations and recommend proper interventions. More precisely, we develop novel deep learning models to recognize important entities and their relations over time, thereby establishing dynamic heterogeneous graphs to describe the observations of social media users. A dynamic graph neural network model can then forecast the trends (e.g. newly diagnosed cases and death rates) and identify high-risk events from social media. Based on the proposed computational method, we also develop a web-based system for domain experts without any computer science background to easily interact with. We conduct extensive experiments on large-scale datasets of COVID-19 related tweets provided by Twitter, which show that our method can precisely predict the new cases and death rates. We also demonstrate the robustness of our web-based pandemic surveillance system and its ability to retrieve essential knowledge and derive accurate predictions across a variety of circumstances. Our system is also available athttp://scaiweb.cs.ucla.edu/covidsurveiller/.This article is part of the theme issue ‘Data science approachs to infectious disease surveillance’.

Hateful and offensive content detection has been extensively explored in a single modality such as text. However, such toxic information could also be communicated via multimodal content such as online memes. Therefore, detecting multimodal hateful content has recently garnered much attention in academic and industry research communities. This paper aims to contribute to this emerging research topic by proposing DisMultiHate, which is a novel framework that performed the classification of multimodal hateful content. Specifically, DisMultiHate is designed to disentangle target entities in multimodal memes to improve hateful content classification and explainability. We conduct extensive experiments on two publicly available hateful and offensive memes datasets. Our experiment results show that DisMultiHate is able to outperform state-of-the-art unimodal and multimodal baselines in the hateful meme classification task. Empirical case studies were also conducted to demonstrate DisMultiHate's ability to disentangle target entities in memes and ultimately showcase DisMultiHate's explainability of the multimodal hateful content classification task., Comment: Paper accepted in ACM Multimedia 2021

This paper examines the potential role of social media in enhancing the understanding and perception of victims of police killings and the data collection surrounding these incidents. Through a series of content analysis and social media mining exercises, the authors observe the emergence of three distinct types of social media content offered on victims of police killings: persistence of the deceased’s activity across social media, sensational commentary on videos and blog postings, and memorials on Facebook, Twitter, and Tumblr. As part of a larger investigation of the availability and accessibility of official police homicide data, this paper aims to present social media data as a potentially powerful source of information to supplement quantitative reports. This process may be especially useful for the most affected communities, particularly BIPOC communities.

Point-of-Interest recommendation is an efficient way to explore interesting unknown locations in social media mining of social networks. In order to solve the problem of sparse data and inaccuracy of single user model, we propose a User-City-Sequence Probabilistic Generation Model (UCSPGM) integrating a collective individual self-adaptive Markov model and the topic model. The collective individual self-adaptive Markov model consists of three parts such as the collective Markov model, the individual self-adaptive Markov model and the self-adaptive rank method. The former determines the topic sequence for all users in system and mines the behavioral patterns of users in a large environment. The later mines behavioral patterns for each user in a small environment. The last determines a self-adaptive-rank for each user in niche. We conduct a large amount of experiments to verify the effectiveness and efficiency of our method.

The present article, placed within the epistemological framework of Political Communication, analyses citizens’ reaction to politicians’ messages on the social network Twitter during the presidential elections in Argentina, Peru, Ecuador, Honduras, and Chile, held between 2015 and 2017. Through a script developed for the present research, almost 200,000 tweets have been studied according to the following questions: What are citizens’ emotional reactions to the messages of presidential candidates? Does digital public opinion analysis have a predictive nature from the electoral point of view? As a result, we note the existence of “sympathy currents” and “antipathy currents” on social media, where positive emotions prevail, especially towards candidates on the right side of the ideological spectrum, with progressive politicians generating a higher anger and sadness index than conservative ones. Similarly, emotions on social media largely correlate to the subsequent electoral result.

Social media mining has proven useful in multiple research fields as a tool for public opinion extraction and analysis. Such mining can discover knowledge from unstructured data in booming social media sources that provide instant public responses and also capture long-term data. Environmental scientists have realized its potential and conducted various studies where public opinion matters. We focus our discussion in this article on mining social media text on environmental issues, with particular emphasis on sentiment analysis, fitting the theme of Data Science and Sustainability. The data science community today is interested in topics that overlap with environmental issues and their broader impacts on sustainability. Such work appeals to scientists focusing on areas such as smart cities, climate change and geo-informatics. Future issues emerging from this research include domain-specific multilingual mining, and advanced geo-location tagging with demographically focused sentiment analysis.

11th International Conference on City Logistics, DUBROVNIK, CROATIE, 12-/06/2019 - 14/06/2019; City Logistics is characterized by multiple stakeholders that often have different views of such a complex system. From a public policy perspective, identifying stakeholders, issues and trends is a daunting challenge, only partially addressed by traditional observation systems. Nowadays social media is one of the biggest channels of public expression and it is often used to communicate opinions and content related to City Logistics. The idea of this research is that analysing social media content could help in understanding the public perception of City logistics. This paper proposes a methodology for collecting content from Twitter and implementing Machine Learning techniques (unsupervised learning and Natural Language Processing), to perform content and sentiment analysis. The proposed methodology is applied to more than 110 000 tweets containing City Logistics key-terms. Results allowed building an Interest Map of concepts and a Sentiment Analysis to determine if City Logistics entries are positive, negative or neutral.

Online behavior recommendation is an attractive research topic related to social media mining. This topic focuses on suggesting suitable behaviors for users in online platforms, including music listening, video watching, e-commerce, to name but a few to improve the user experience, an essential factor for the success of online services. A successful online behavior recommendation system should have the ability to predict behaviors that users used to performs and also suggest behaviors that users never performed before. In this paper, we develop a mixture model that contains two components to address this problem. The first component is the user-specific preference component that represents the habits of users based on their behavior history. The second component is the latent group preference component based on variational autoencoder, a deep generative neural network. This component corresponds to the hidden interests of users and allows us to discover the unseen behavior of users. We conduct experiments on various real-world datasets with different characteristics to show the performance of our model in different situations. The result indicates that our proposed model outperforms the previous mixture models for recommendation problem.

Signed graphs are often used as models for social media mining, social networks analysis and nature language processing. In this paper, we study clustering algorithms for signed graphs that can be scaled for use in large-scale signed networks. A proposed mechanism, called a random walk gap (RWG), is used to extract more cluster structure information. RWG uses two types of random walks on signed graphs. The first considers positive edges only. The second takes negative edges into account. Three types of edges in signed graphs are identified and their different characteristics in terms of transition probability changes are determined for the two types of random walk graphs. Different characteristics contain different information about cluster structure and a RWG matrix is used to reflect the differences. The RWG matrix is used to adjust the weights of edges in signed graphs. This is the first study to perform a random walk on negative edges. A fast clustering for signed graphs (FCSG) algorithm is then proposed. This FCSG is compared with existing methods. The computational times for different algorithms are measured. The experiments show that the proposed FCSG algorithm gives better results than existing algorithms based on the performance criteria of imbalance and modularity.

Health authorities worldwide strive to detect Influenza prevalence as early as possible in order to prepare for it and minimize its impacts. To this end, we address the Influenza prevalence surveillance and prediction problem. In this paper, we develop a new Influenza prevalence prediction model, called Tweetluenza, to predict the spread of the Influenza in real time using cross-lingual data harvested from Twitter data streams with emphases on the United Arab Emirates (UAE). Based on the features of tweets, Tweetluenza filters the Influenza tweets and classifies them into two classes, reporting and non-reporting. To monitor the growth of Influenza, the reporting tweets were employed. Furthermore, a linear regression model leverages the reporting tweets to predict the Influenza-related hospital visits in the future. We evaluated Tweetluenza empirically to study its feasibility and compared the results with the actual hospital visits recorded by the UAE Ministry of Health. The results of our experiments demonstrate the practicality of Tweetluenza, which was verified by the high correlation between the Influenza-related Twitter data and hospital visits due to Influenza. Furthermore, the evaluation of the analysis and prediction of Influenza shows that combining English and Arabic tweets improves the correlation results.

Purpose Food plays an important role in every culture around the world. Recently, cuisine preference analysis has become a popular research topic. However, most of these studies are conducted through questionnaires and interviews, which are highly limited by the time, cost and scope of data collection, especially when facing large-scale survey studies. Some researchers have, therefore, attempted to mine cuisine preferences based on online recipes, while this approach cannot reveal food preference from people’s perspective. Today, people are sharing what they eat on social media platforms by posting reviews about the meal, reciting the names of appetizers or entrees, and photographing as well. Such large amount of user-generated contents (UGC) has potential to indicate people’s preferences over different cuisines. Accordingly, the purpose of this paper is to explore Chinese cuisine preferences among online users of social media. Design/methodology/approach Based on both UGC and online recipes, the authors first investigated the cuisine preference distribution in different regions. Then, dish preference similarity between regions was calculated and few geographic factors were identified, which might lead to such regional similarity appeared in our study. By applying hierarchical clustering, the authors clustered regions based on dish preference and ingredient usage separately. Findings Experimental results show that, among 20 types of traditional Chinese cuisines, Sichuan cuisine is most favored across all regions in China. Geographical proximity is the more closely related to differences of regional dish preference than climate proximity. Originality/value Different from traditional definitions of regions to which cuisine belong, the authors found new association between region and cuisine based on dish preference from social media and ingredient usage of dishes. Using social media may overcome problems with using traditional questionnaires, such as high costs and long cycle for questionnaire design and answering.

Social media data have recently attracted considerable attention as an emerging voice of the customer as it has rapidly become a channel for exchanging and storing customer-generated, large-scale, and unregulated voices about products. Although product planning studies using social media data have used systematic methods for product planning, their methods have limitations, such as the difficulty of identifying latent product features due to the use of only term-level analysis and insufficient consideration of opportunity potential analysis of the identified features. Therefore, an opportunity mining approach is proposed in this study to identify product opportunities based on topic modeling and sentiment analysis of social media data. For a multifunctional product, this approach can identify latent product topics discussed by product customers in social media using topic modeling, thereby quantifying the importance of each product topic. Next, the satisfaction level of each product topic is evaluated using sentiment analysis. Finally, the opportunity value and improvement direction of each product topic from a customer-centered view are identified by an opportunity algorithm based on product topics’ importance and satisfaction. We expect that our approach for product planning will contribute to the systematic identification of product opportunities from large-scale customer-generated social media data and will be used as a real-time monitoring tool for changing customer needs analysis in rapidly evolving product environments.

In the ever-connected social networking era, terrorists exploit social media platforms via sophisticated approaches. To curb these activities, a rich collection of computational methods was developed. This article surveys the use of social media by terrorists, followed by a temporal classification framework that overviews computational countermeasures at four major stages, including inception of an attack, immediately before an attack, onset of an attack, and after an attack. The literature surveyed was organized around the four temporal stages. The resulting survey is summarized in a table with the main technology used in each stage based on the time of the attack.

Data mining on social media platforms (Instagram™, Flickr™, and Twitter™) is rapidly increasing and application of data mining techniques has contributed to significant findings in various fields such as tourism, ecology, and politics etc. In the face of globalization and nature-based tourism is thriving in many countries, social media activity on tourism is increasing despite the socio-economical barriers. In this context, this paper attempts to understand the metadata of photographs related to whale watching in Sri Lanka in Flickr social media platform. Photographs related to whale watching was extracted and analyzed for i) photographic content ii) Geo-tags iii) Social-tags and iv) Photographers’ nationalities by using Flickr API (Application Programming Interface) and self-written python program script. Content analysis of the photographs has identified five major categories (human activity, accommodation, natural phenomena, animals and other) of photographs based on the major element present in each photograph. Mapping of geo-tagged photographs indicated that Mirissa was the hotspot for whale watching in Sri Lanka. Moreover, the present study suggests that mapping of geo-tagged photographs can be used as proxy data for whale distribution in Sri Lanka. Analysis of social tags indicated that tags indicating whale (156), Sri Lanka (144) and Mirissa (133) were popular among the photographers. The demographic profile of the photographers indicated that the highest number of photographers (25%) from the United Kingdom followed by Sri Lanka (18.69%) and China (12.94%) interested in whale watching. Despite some of the weaknesses, this study has demonstrated that metadata of Flickr photographs can effectively be used for understanding the basic information related to whale-watching tourism in Sri Lanka.

The surveillance and preventions of infectious disease epidemics such as influenza and Ebola are important and challenging issues. It is therefore crucial to characterize the disease progress and epidemics process efficiently and accurately. Computational epidemiology can model the progression of the disease and its underlying contact network, but as yet lacks the ability to process of real-time and fine-grained surveillance data. Social media, on the other hand, provides timely and detailed disease surveillance but is insensible to the underlying contact network and disease model. To address these challenges simultaneously, this paper proposes a novel semi-supervised neural network framework that integrates the strengths of computational epidemiology and social media mining techniques for influenza epidemiological modeling. Specifically, this framework learns social media users’ health states and intervention actions in real time, regularized by the underlying disease model and contact network. The learned knowledge from social media can then be fed into the computational epidemic model to improve the efficiency and accuracy of disease diffusion modeling. We propose an online optimization algorithm that iteratively processes the above interactive learning process. The extensive experimental results provided demonstrated that our approach can not only outperform competing methods by a substantial margin in forecasting disease outbreaks, but also characterize the individual-level disease progress and diffusion effectively and efficiently.

Being able to automatically extract as much relevant posts as possible from social media in a timely manner is key in many activities, for example to provide useful information in order to rapidly create crisis maps during emergency events. While most social media support keyword-based searches, the amount and the accuracy of retrieved posts depend largely on the keywords employed. The goal of the proposed methodology is to dynamically extract relevant keywords for searching social media during an emergency event, following the event’s evolution. Starting from a set of keywords designed for the type of event being considered (floods and earthquakes, in particular), the set of keywords is automatically adjusted taking into account the spatio-temporal features of the monitored event. The goal is to retrieve posts following the event’s evolution and to benefit from cross-social crawling in order to exploit the specific characteristics of a social media over others. In the case considered in this paper, we exploit the precision of the geolocation of images posted in Flickr to extract keywords to search YouTube posts for the same event, since YouTube does not allow spatial crawling yet provides a richer source of information. The methodology was evaluated on three recent major emergency events, demonstrating a large increase in the number of retrieved posts compared with the use of generic seed keywords. This is a relevant improvement of relevance for providing information on emergency events, and the ability to follow the event’s development.

Emphasis on non-motorized travel modes (for example, biking) reduces motorized trips and provides positive effects on the environment and the quality of human life. Understanding factors that influence people to biking or bike commuting can help decision makers, transportation planners, and bike commuting networks. Historically, conventional methods like surveys and crash data analyses were conducted to understand relevant factors. Survey and crash data analysis are difficult to perform in broad scale due to data availability and efforts. An innovative approach to determining these factors is to conduct social media mining to understand sentiments or motivations of bike commuters. People use terms (with hashtag at the beginning of the term) in Twitter, a popular social media network, to express their thoughts, activities or information. This study developed a framework for using Twitter data in understating the sentiments of the bikers with minimal effort. In this study, Twitter data associated with bike commuting hashtags were obtained for eight years (2009–2016). This study provided a framework of data collection and application of various natural language processing (NLP) tools (for example, text mining, sentiment analysis) to extract knowledge from the unstructured text data. Findings show that biking is associated with weather and seasonal patterns. The general sentiment towards biking is positive. However, negative sentiments are associated with bad weather, crime, and other challenges. The polarity scores indicate somewhat positiveness in the recent few years. The developed framework and the findings of this study will help planners and decision makers to promote biking on a broader scale. Keywords: Bike commuting, Text mining, Sentiment analysis, Polarity, Twitter, Social media

Social media content is one of the most visible sources of big data and is often used in health studies to draw inferences about various behaviors. Though much can be gleaned from social media data and mining, the approaches used to collect and analyze data are generally strengthened when examined through established theoretical frameworks. Health behavior, a theory driven field, encourages interdisciplinary collaboration across fields and theories to help us draw robust conclusions about phenomena. This pilot study uses a combined computer informatics and SNA approach to analyze information spread about mask-wearing as a personal mitigation effort during the COVID-19 pandemic. We analyzed one week’s worth of Twitter data (n= 10,107 tweets across 4,289 users) by using at least one of four popular mask-support hashtags (e.g., #maskup). We calculated network-measures to assess structures and patterns present within the Twitter network, and used exponential random graph modeling (ERGM) to test factors related to the presence of retweets between users. The pro-mask Twitter network was largely fragmented, with a select few nodes occupying the most influential positions in the network. Verified accounts, accounts with more followers, and those who generated more tweets were more likely to be retweeted. Contrarily, verified accounts and those with more followers were less likely to retweet others. SNA revealed patterns and structures theoretically important to how information spreads across Twitter. We demonstrated the utility of an interdisciplinary collaboration between computer informatics and SNA to draw conclusions from social media data.

The huge volume of user-generated data on social media is the result of the aggregation of users’ personal backgrounds, past experiences, and daily activities. This huge size of the generated data, the so-called “big data,” has been studied and investigated intensively during the past few years. In spite of the impression one may get from the media, a great deal of data processing has not been uncovered by existing techniques of data engineering and processing. However, very few scholars have tried to do so, especially from the perspective of multiple-criteria decision-making (MCDM). These MCDM methods can derive influence relationships and weights associated with aspects and criteria, which can hardly be achieved by traditional data analytics and statistical approaches. Therefore, in this paper, we aim to propose an analytic framework to mine social networks, feed the meaningful information via MCDM methods based on a theoretical framework, derive causal relationships among the aspects of the theoretical framework, and finally compare the causal relationships with a social theory. Latent Dirichlet allocation (LDA) will be adopted to derive topic models based on the data retrieved from social media. By clustering the topics into aspects of the social theory, the probability associated with each aspect will be normalized and then transformed to a Likert-type 5-point scale. Afterwards, for every topic, the feature importance of all other topics will be derived using the random forest (RF) algorithm. The feature importance matrix will be transformed to the initial influence matrix of the decision-making trial and evaluation laboratory (DEMATEL). The influence relationships among the aspects and criteria and influence weights can then be derived by using the DEMATEL-based analytic network process (DANP). The influence weight versus each criterion can be derived by using DANP. To verify the feasibility of the proposed framework, Taiwanese users’ attitudes toward air pollution will be analyzed based on the value–belief–norm (VBN) theory by using social media data retrieved from Dcard (dcard.tw). Based on the analytic results, the causal relationships are fully consistent with the VBN framework. Further, the mutual influences derived in this work that were seldom discussed by earlier works, i.e., the mutual influences between altruistic concerns and egoistic concerns, as well as those between altruistic concerns and biosphere concerns, are worth further investigation in future.

Objective Research on pharmacovigilance from social media data has focused on mining adverse drug events (ADEs) using annotated datasets, with publications generally focusing on 1 of 3 tasks: ADE classification, named entity recognition for identifying the span of ADE mentions, and ADE mention normalization to standardized terminologies. While the common goal of such systems is to detect ADE signals that can be used to inform public policy, it has been impeded largely by limited end-to-end solutions for large-scale analysis of social media reports for different drugs. Materials and Methods We present a dataset for training and evaluation of ADE pipelines where the ADE distribution is closer to the average ‘natural balance’ with ADEs present in about 7% of the tweets. The deep learning architecture involves an ADE extraction pipeline with individual components for all 3 tasks. Results The system presented achieved state-of-the-art performance on comparable datasets and scored a classification performance of F1 = 0.63, span extraction performance of F1 = 0.44 and an end-to-end entity resolution performance of F1 = 0.34 on the presented dataset. Discussion The performance of the models continues to highlight multiple challenges when deploying pharmacovigilance systems that use social media data. We discuss the implications of such models in the downstream tasks of signal detection and suggest future enhancements. Conclusion Mining ADEs from Twitter posts using a pipeline architecture requires the different components to be trained and tuned based on input data imbalance in order to ensure optimal performance on the end-to-end resolution task.

In this study, we applied data mining tools to build a business intelligence platform combining different information, obtained through social media mining, and integrating it with information about the characteristics of movies and their box offices. We used data from the social media following of both the film and its main actors. In addition, data was collected from the number of movie publications on Facebook, Instagram, and Twitter, as well as the corresponding engagement. We built different prediction models of box office revenue by empirically incorporating the effect of the movie’s community on social media (movie and actors’ followers), its activity on social media, and users’ reactions to the movie’s activity. Our analyses are based on Fama-Macbeth two-step regressions applied to panel data. The prediction models confirm the value of social media posts and their engagement, especially on Facebook and on Instagram, to increase the film’s performance. The results show that the volume of the movie’s and actors’ followers, the actions, and the users’ reactions on social media positively affect the box office. Then, the study supports the importance of building a business intelligence system, so companies can take advantage of their social media accounts. Film studios must be proactive with the access and the treatment of different data sources available to the movie’s online community through data mining techniques to attract more viewers and to increase their box office.

Background Individuals frequently turning to social media to discuss medical conditions and medication, sharing their experiences and information and asking questions among themselves. These online discussions can provide valuable insights into individual perceptions of medical treatment, and increasingly, studies are focusing on the potential use of this information to improve health care management. Objective The objective of this infodemiology study was to identify social media posts mentioning paracetamol-containing products to develop a better understanding of patients’ opinions and perceptions of the drug. Methods Posts between January 2003 and March 2019 containing at least one mention of paracetamol were extracted from 18 French forums in May 2019 with the use of the Detec’t (Kap Code) web crawler. Posts were then analyzed using the automated Detec’t tool, which uses machine learning and text mining methods to inspect social media posts and extract relevant content. Posts were classified into groups: Paracetamol Only, Paracetamol and Opioids, Paracetamol and Others, and the Aggregate group. Results Overall, 44,283 posts were analyzed from 20,883 different users. Post volume over the study period showed a peak in activity between 2009 and 2012, as well as a spike in 2017 in the Aggregate group. The number of posts tended to be higher during winter each year. Posts were made predominantly by women (14,897/20,883, 71.34%), with 12.00% (2507/20,883) made by men and 16.67% (3479/20,883) by individuals of unknown gender. The mean age of web users was 39 (SD 19) years. In the Aggregate group, pain was the most common medical concept discussed (22,257/37,863, 58.78%), and paracetamol risk was the most common discussion topic, addressed in 20.36% (8902/43,725) of posts. Doliprane was the most common medication mentioned (14,058/44,283, 31.74%) within the Aggregate group, and tramadol was the most commonly mentioned drug in combination with paracetamol in the Aggregate group (1038/19,587, 5.30%). The most common unapproved indication mentioned within the Paracetamol Only group was fatigue (190/616, with 16.32% positive for an unapproved indication), with reference to dependence made by 1.61% (136/8470) of the web users, accounting for 1.33% (171/12,843) of the posts in the Paracetamol Only group. Dependence mentions in the Paracetamol and Opioids group were provided by 6.94% (248/3576) of web users, accounting for 5.44% (342/6281) of total posts. Reference to overdose was made by 245 web users across 291 posts within the Paracetamol Only group. The most common potential adverse event detected was nausea (306/12843, 2.38%) within the Paracetamol Only group. Conclusions The use of social media mining with the Detec’t tool provided valuable information on the perceptions and understanding of the web users, highlighting areas where providing more information for the general public on paracetamol, as well as other medications, may be of benefit.

Magistrate’ use of social networks has been scarcely studied. It is assumed that the discretion of the jurisdictional function limits their interaction; but judges and prosecutors can be active Internet users in Latin America. This article analyzes the affiliation networks of the supreme prosecutors of Ecuador and Peru: based on the data generated by users (alteri) in their conversations, it assesses the role played by these magistrates (egos) in matters related to justice in their countries. Through the Social Network Analysis and Social Media Mining, the study poses the following questions: What topics led the conversations between egos and alteri? And what kinds of relationships did they build in their Twitter communities? It is concluded that, in the online world, egos become “political actors”, putting their delicate function at risk, to the detriment of transparency required by the administration of justice and respect among powers, a pillar of democracy.

In this work we describe our submissions to the Social Media Mining for Health (SMM4H) 2021 Shared Task. We investigated the effectiveness of a joint training approach to Task 1, specifically classification, extraction and normalization of Adverse Drug Effect (ADE) mentions in English tweets. Our approach performed well on the normalization task, achieving an above average f1 score of 24%, but less so on classification and extraction, with f1 scores of 22% and 37% respectively. Our experiments also showed that a larger dataset with more negative results led to stronger results than a smaller more balanced dataset, even when both datasets have the same positive examples. Finally we also submitted a tuned BERT model for Task 6: Classification of Covid-19 tweets containing symptoms, which achieved an above average f1 score of 96%.

Dining is an essential part of human life. In order to pursue a healthier self, more and more people enjoy homemade cuisines. Consequently, the amount of recipe websites has increased significantly. These online recipes represent different cultures and cooking methods from various regions, and provide important indications on nutritional content. In recent years, the development of data science made data mining a popular research area. However, only a few researches in Taiwan have applied data mining in the studies of recipes and nutrients. Therefore, this work aims at utilizing machine learning models to discover health-related insights from recipes on social media. First, we collected over 15,000 Chinese recipes from the largest recipe website in Taiwan to build a recipe database. We then extracted information from this dataset through natural language processing methodologies so as to better understand the characteristics of various cuisines and ingredients. Thus, we can establish a classification model for the automatic categorization of recipes. We further performed cluster analysis for grouping nutrients to recognize the nutritional differences for each cluster and each cuisine type. The results showed that using the support vector machine (SVM) model can successfully classify recipes with an average F-score of 82%. We also analyzed the nutritional value of different cuisine categories and the possible health effects they may bring to the consumers. Our methods and findings can assist future work on extracting essential nutritional information from recipes and promoting healthier diets.

Fashion trend prediction is a complex problem due to a vast number of latent drivers and fashion products' highly temporal and dynamic characteristics. It requires a reliable and time-sensitive means to detect fashion trend. Analyzing social media data is well-suited for the task of trend detection from the consumer perspective. In this project, I leverage text analytics on time-series social media data. I first collected data across multiple categories of fashion over a span of ten years via mining, API and third-party sources. Then I extract fashion related keywords and conduct yearly and quarterly trend detection via keyword clustering and topic modeling. The ultimate goal is to discover hidden semantic structures in a text body at scale and detect types and drivers of different fashion life cycles, such as classic, cyclical, and fad.