Explore the vast range of titles available.

How do climate change deniers differ from believers? Is there any correlation between human sentiment and deviations from historic temperature? We answer nine such questions using 13 years of Twitter data and 15 million tweets. Seven aspects are explored, namely, user gender, climate change stance and sentiment, aggressiveness, deviations from historic temperature, topics discussed, and environmental disaster events. We found that: a) climate change deniers use the term global warming much often than believers and use aggressive language, while believers tweet more about taking actions to fight the phenomenon, b) deniers are more present in the American Region, South Africa, Japan, and Eastern China and less present in Europe, India, and Central Africa, c) people connect much more the warm temperatures with man-made climate change than cold temperatures, d) the same regions that had more climate change deniers also tweet with negative sentiment, e) a positive correlation is observed between human sentiment and deviations from historic temperature; when the deviation is between −1.143° C and +2.401° C , people tweet the most positive, f) there exist 90% correlation between sentiment and stance, and -94% correlation between sentiment and aggressiveness, g) no clear patterns are observed to correlate sentiment and stance with disaster events based on total deaths, number of affected, and damage costs, h) topics discussed on Twitter indicate that climate change is a politicized issue and people are expressing their concerns especially when witnessing extreme weather; the global stance could be considered optimistic, as there are many discussions that point out the importance of human intervention to fight climate change and actions are being taken through events to raise the awareness of this phenomenon.

[This corrects the article DOI: 10.1371/journal.pone.0274213.].[This corrects the article DOI: 10.1371/journal.pone.0274213.].

In the realm of disaster-related tweet classification, this study presents a comprehensive analysis of various machine learning algorithms, shedding light on crucial factors influencing algorithm performance. The exceptional efficacy of simpler models is attributed to the quality and size of the dataset, enabling them to discern meaningful patterns. While powerful, complex models are time-consuming and prone to overfitting, particularly with smaller or noisier datasets. Hyperparameter tuning, notably through Bayesian optimization, emerges as a pivotal tool for enhancing the performance of simpler models. A practical guideline for algorithm selection based on dataset size is proposed, consisting of Bernoulli Naive Bayes for datasets below 5000 tweets and Logistic Regression for larger datasets exceeding 5000 tweets. Notably, Logistic Regression shines with 20,000 tweets, delivering an impressive combination of performance, speed, and interpretability. A further improvement of 0.5% is achieved by applying ensemble and stacking methods.

In this work a fish species distribution model (SDM) was developed, by merging species occurrence data with environmental layers, with the scope to produce high resolution habitability maps for the whole Mediterranean Sea. The final model is capable to predict the probability of occurrence of each fish species at any location in the Mediterranean Sea. Eight pelagic, commercial fish species were selected for this study namely Engraulis encrasicolus, Sardina pilchardus, Sardinella aurita, Scomber colias, Scomber scombrus, Spicara smaris, Thunnus thynnus and Xiphias gladius. The SDM environmental predictors were obtained from the databases of Copernicus Marine Environmental Service (CMEMS) and the European Marine Observation and Data Network (EMODnet). The probabilities of fish occurrence data in low resolution and with several gaps were obtained from Aquamaps (FAO Fishbase). Data pre-processing involved feature engineering to construct 6830 features, representing the distribution of several mean-monthly environmental variables, covering a time-span of 10 years. Feature selection with the ensemble Reciprocal Ranking method was used to rank the features according to their relative importance. This technique increased model’s performance by 34%. Ten machine learning algorithms were then applied and tested based on their overall performance per species. The XGBoost algorithm performed better and was used as the final model. Feature categories were explored, with neighbor-based, extreme values, monthly and surface ones contributing most to the model. Environmental variables like salinity, temperature, distance to coast, dissolved oxygen and nitrate were found the strongest ones in predicting the probability of occurrence for the above eight species.

The classic example of machine learning is based on isolated learning—a single model for each task using a single dataset [...]

We focus on personal data generated by the sensors and through the everyday usage of smart devices and take advantage of these data to build a non-invasive contextual suggestion system for tourism. The system, which we call Pythia, exploits the computational capabilities of modern smart devices to offer high quality personalized POI (point of interest) recommendations. To protect user privacy, we apply a privacy by design approach within all of the steps of creating Pythia. The outcome is a system that comprises important architectural and operational innovations. The system is designed to process sensitive personal data, such as location traces, browsing history and web searches (query logs), to automatically infer user preferences and build corresponding POI-based user profiles. These profiles are then used by a contextual suggestion engine to anticipate user choices and make POI recommendations for tourists. Privacy leaks are minimized by implementing an important part of the system functionality at the user side, either as a mobile app or as a client-side web application, and by taking additional precautions, like data generalization, wherever necessary. As a proof of concept, we present a prototype that implements the aforementioned mechanisms on the Android platform accompanied with certain web applications. Even though the current prototype focuses only on location data, the results from the evaluation of the contextual suggestion algorithms and the user experience feedback from volunteers who used the prototype are very positive.

One of the causes of several problems on the internet, such as financial fraud, cyber-bullying, and seduction of minors, is the complete anonymity that a malicious user can maintain. Most methods that have been proposed to remove this anonymity are either intrusive, or violate privacy, or expensive. This paper proposes the recognition of certain characteristics of an unknown user through keystroke dynamics, which is the way a person is typing. The evaluation of the method consists of three stages: the acquisition of keystroke dynamics data from 118 volunteers during the daily use of their devices, the extraction and selection of keystroke dynamics features based on their information gain, and the testing of user characteristics recognition by training five well-known machine learning models. Experimental results show that it is possible to identify the gender, the age group, the handedness, and the educational level of an unknown user with high accuracy.

We propose and experimentally investigate the usefulness of several features for selecting image content (objects) suitable for image captioning. The approach taken explores three broad categories of features, namely geometric, conceptual, and visual. Experiments suggest that widely known geometric ‘rules’ in art–aesthetics or photography (such as the golden ratio or the rule-of-thirds) and facts about the human visual system (such as its wider horizontal angle than its vertical) provide no useful information for the task. Human captioners seem to prefer large, elongated (but not in the golden ratio) objects, positioned near the image center, irrespective of orientation. Conceptually, the preferred objects are either too specific or too general, and animate things are almost always mentioned; furthermore, some evidence is found for selecting diverse objects in order to achieve maximal image coverage in captions. Visual object features such as saliency, depth, edges, entropy, and contrast, are all found to provide useful information. Beyond evaluating features in isolation, we investigate how well these are combined by performing feature and feature-category ablation studies, leading to an effective set of features which can be proven useful for operational systems. Moreover, we propose alternative ways for feature engineering and evaluation, dealing with the drawbacks of the evaluation methodology proposed in past literature.

The present study aims to create a framework that analyses user posts related to a product of interest on social networking platforms. More precisely, by applying information mining techniques, posts are categorised according to the intention they express, the sentiment polarisation, and the type of opinion. The model operates based on linguistic rules, machine learning, and combinations. Six different methodologies are implemented to extract intent, sentiment, and type of opinion from a tweet. The final model automatically detects intention to buy or not to buy the product, intention to compare the product with other competitors, and finally, intention to search for information about the product. It then categorises the text according to the sentiment and depending on their expressed opinion. The dataset comprises tweets for each day of the iPhone 5’s life cycle, corresponding to 365 days. Additionally, it demonstrated that the business’s external or internal decisions affect the public purchasing audience’s opinions, sentiments, and intentions expressed on social media. Lastly, as a Business Intelligence tool, the framework recognises and analyses these points, which contribute substantially to the company’s decision-making through the findings.

How do climate change deniers differ from believers? Is there any correlation between human sentiment and deviations from historic temperature? We answer nine such questions using 13 years of Twitter data and 15 million tweets. Seven aspects are explored, namely, user gender, climate change stance and sentiment, aggressiveness, deviations from historic temperature, topics discussed, and environmental disaster events. We found that: a) climate change deniers use the term global warming much often than believers and use aggressive language, while believers tweet more about taking actions to fight the phenomenon, b) deniers are more present in the American Region, South Africa, Japan, and Eastern China and less present in Europe, India, and Central Africa, c) people connect much more the warm temperatures with man-made climate change than cold temperatures, d) the same regions that had more climate change deniers also tweet with negative sentiment, e) a positive correlation is observed between human sentiment and deviations from historic temperature; when the deviation is between −1.143°C and +2.401°C, people tweet the most positive, f) there exist 90% correlation between sentiment and stance, and -94% correlation between sentiment and aggressiveness, g) no clear patterns are observed to correlate sentiment and stance with disaster events based on total deaths, number of affected, and damage costs, h) topics discussed on Twitter indicate that climate change is a politicized issue and people are expressing their concerns especially when witnessing extreme weather; the global stance could be considered optimistic, as there are many discussions that point out the importance of human intervention to fight climate change and actions are being taken through events to raise the awareness of this phenomenon.