Mobile-based online data mining : outdoor activity recognition
Bibliographic record
Abstract
One of the unique features of mobile applications is the context awareness. The mobility and power afforded by smartphones allow users to interact more directly and constantly with the external world more than ever before. The emerging capabilities of smartphones are fueling a rise in the use of mobile phones as input devices for a great range of application fields; one of these fields is the activity recognition. In pervasive computing, activity recognition has a significant weight because it can be applied to many real-life, human-centric problems. This important role allows providing services to various application domains ranging from real-time traffic monitoring to fitness monitoring, social networking, marketing and healthcare. However, one of the major problems that can shatter any mobile-based activity recognition model is the limited battery life. It represents a big hurdle for the quality and the continuity of the service. Indeed, excessive power consumption may become a major obstacle to broader acceptance context-aware mobile applications, no matter how useful the proposed service may be. We present during this thesis a novel unsupervised battery-aware approach to online recognize users’ outdoor activities without depleting the mobile resources. We succeed in associating the places visited by individuals during their movements to meaningful human activities. Our approach includes novel models that incrementally cluster users’ movements into different types of activities without any massive use of historical records. To optimize battery consumption, our approach behaves variably according to users’ behaviors and the remaining battery level. Moreover, we propose to learn users’ habits in order to reduce the activity recognition computation. Our innovative battery-friendly method combines activity recognition and prediction in order to recognize users’ activities accurately without draining the battery of their phones. We show that our approach reduces significantly the battery consumption while keeping the same high accuracy. \n \nUne des caractéristiques uniques des applications mobiles est la sensibilité au contexte. La mobilité et la puissance de calcul offertes par les smartphones permettent aux utilisateurs d’interagir plus directement et en permanence avec le monde extérieur. Ces capacités émergentes ont pu alimenter plusieurs champs d’applications comme le domaine de la reconnaissance d’activités. Dans le domaine de l'informatique omniprésente, la reconnaissance des activités humaines reçoit une attention particulière grâce à son implication profonde dans plusieurs problématiques de vie quotidienne. Ainsi, ce domaine est devenu une pièce majeure qui fournit des services à un large éventail de domaines comme la surveillance du trafic en temps réel, les réseaux sociaux, le marketing et la santé. Cependant, l'un des principaux problèmes qui peuvent compromettre un modèle de reconnaissance d’activité sur les smartphones est la durée de vie limitée de la batterie. Ce handicap représente un grand obstacle pour la qualité et la continuité du service. En effet, la consommation d'énergie excessive peut devenir un obstacle majeur aux applications sensibles au contexte, peu importe à quel point ce service est utile. Nous présentons dans de cette thèse une nouvelle approche non supervisée qui permet la détection incrémentale des activités externes sans épuiser les ressources du téléphone. Nous parvenons à associer efficacement les lieux visités par des individus lors de leurs déplacements à des activités humaines significatives. Notre approche comprend de nouveaux modèles de classification en ligne des activités humaines sans une utilisation massive des données historiques. Pour optimiser la consommation de la batterie, notre approche se comporte de façon variable selon les comportements des utilisateurs et le niveau de la batterie restant. De plus, nous proposons d'apprendre les habitudes des utilisateurs afin de réduire la complexité de l’algorithme de reconnaissance d'activités. Pour se faire, notre méthode combine la reconnaissance d’activités et la prédiction des prochaines activités afin d’atteindre une consommation raisonnable des ressources du téléphone. Nous montrons que notre proposition réduit remarquablement la consommation de la batterie tout en gardant un taux de précision élevé.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.008 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".