This study investigates the performance of Logistic Regression, k-Nearest Neighbors (KNN), and Random Forest algorithms in predicting customer churn within an e-commerce platform. The choice of the mentioned algorithms was due to the unique characteristics of the dataset and the unique perception and value provided by each algorithm. Iterative models ‘examinations, encompassing preprocessing techniques, feature engineering, and rigorous evaluations, were conducted. Logistic Regression showcased moderate predictive capabilities but lagged in accurately identifying potential churners due to its assumptions of linearity between log odds and predictors. KNN emerged as the most accurate classifier, achieving superior sensitivity and specificity (98.22% and 96.35%, respectively), outperforming other models. Random Forest, with sensitivity and specificity (91.75% and 95.83% respectively) excelled in specificity but slightly lagged in sensitivity. Feature importance analysis highlighted "Tenure" as the most impactful variable for churn prediction. Preprocessing techniques differed in performance across models, emphasizing the importance of tailored preprocessing. The study's findings underscore the significance of continuous model refinement and optimization in addressing complex business challenges like customer churn. The insights serve as a foundation for businesses to implement targeted retention strategies, mitigating customer attrition, and promote growth in e-commerce platforms.

In collaboration with Westinghouse Electric AB this project aims to study the possibilities with using machine learning methods to predict fuel failure in a Boiling Water Reactors (BWRs). The main objective has been to create a dataset consisting of both empirical measurements and simulated samples from a physics model and evaluate different machine learning algorithms, that use these datasets to predict fuel defects. The simulated data is created using a physics model derived from the ANS-5.4 standard which allows for good control over specific parameter values. Three machine learning algorithms were deemed fit for this type of problem and used throughout the project: Random Forest (RF), K-Nearest Neighbor (KNN) and Neural Network (NN). Both classification and regression type problems have been assessed. All three methods showed good results for the classification problems, where the goal was to predict if there was a fuel failure or not. All models reached an accuracy above 97% and performed well, the RF model had the highest overall, with an accuracy of 98.2 %. However, the NN method made the fewest false negative predictions and can therefore be seen as the best model for this purpose. For the regression, problems with the aim of predicting escape rates, both the RF and KNN had similar promising results with very small errors overall. Yet, there is a slight increase in errors when predicting higher escape rates for both models. This is most likely due to the available data being of mostly low escape rates. The NN did not perform well with this problem, the predictions having large error for both low and high escape rates, a possible explanation is the lack of data. To improve the results, and create even better models, the empirical measurements need to contain more information such as defect location and fuel failure size, also an increase in the number of samples taken at fuel failure operation would be valuable.

This thesis proposes a deep learning anomaly detection pipeline to detect possible anomalies during the operation of a fleet of batteries and presents its development and evaluation. The pipeline employs sensors that connect to each battery in the fleet to remotely collect real-time measurements of their operating characteristics, such as voltage, current, and temperature. The deep learning based time-series anomaly detection model was developed using Variational Autoencoder (VAE) architecture that utilizes either Long Short-Term Memory (LSTM) or, its cousin, Gated Recurrent Unit (GRU) as the encoder and the decoder networks (LSTMVAE and GRUVAE). Both variants were evaluated against three well-known conventional anomaly detection algorithms Isolation Nearest Neighbour (iNNE), Isolation Forest (iForest), and kth Nearest Neighbour (k-NN) algorithms. All five models were trained using two variations in the training dataset (full-year dataset and partial recent dataset), producing a total of 10 different model variants. The models were trained using the unsupervised method and the results were evaluated using a test dataset consisting of a few known anomaly days in the past operation of the customer’s battery fleet. The results demonstrated that k-NN and GRUVAE performed close to each other, outperforming the rest of the models with a notable margin. LSTMVAE and iForest performed moderately, while the iNNE and iForest variant trained with the full dataset, performed the worst in the evaluation. A general observation also reveals that limiting the training dataset to only a recent period produces better results nearly consistently across all models. / Detta examensarbete föreslår en pipeline för djupinlärning av avvikelser för att upptäcka möjliga anomalier under driften av en flotta av batterier och presenterar dess utveckling och utvärdering. Rörledningen använder sensorer som ansluter till varje batteri i flottan för att på distans samla in realtidsmätningar av deras driftsegenskaper, såsom spänning, ström och temperatur. Den djupinlärningsbaserade tidsserieanomalidetekteringsmodellen utvecklades med VAE-arkitektur som använder antingen LSTM eller, dess kusin, GRU som kodare och avkodarnätverk (LSTMVAE och GRU) VAE). Båda varianterna utvärderades mot tre välkända konventionella anomalidetekteringsalgoritmer -iNNE, iForest och k-NN algoritmer. Alla fem modellerna tränades med hjälp av två varianter av träningsdatauppsättningen (helårsdatauppsättning och delvis färsk datauppsättning), vilket producerade totalt 10 olika modellvarianter. Modellerna tränades med den oövervakade metoden och resultaten utvärderades med hjälp av en testdatauppsättning bestående av några kända anomalidagar under tidigare drift av kundens batteriflotta. Resultaten visade att k-NN och GRUVAE presterade nära varandra och överträffade resten av modellerna med en anmärkningsvärd marginal. LSTMVAE och iForest presterade måttligt, medan varianten iNNE och iForest tränade med hela datasetet presterade sämst i utvärderingen. En allmän observation avslöjar också att en begränsning av träningsdatauppsättningen till endast en ny period ger bättre resultat nästan konsekvent över alla modeller.

The Protein Data Bank (PDB) is the definitive electronic repository for experimentally-derived protein structures, composed mainly of those determined by X-ray crystallography. Approximately 200 new structures are added weekly to the PDB, and at the time of writing, it contains approximately 97,000 structures. This represents an expanding wealth of high-quality information but there seem to be few bioinformatics tools that consider and analyse these data as an ensemble. This thesis explores the development of three efficient, fast algorithms and software implementations to study protein structure using the entire PDB. The first project is a crystal-form matching tool that takes a unit cell and quickly (< 1 second) retrieves the most related matches from the PDB. The unit cell matches are combined with sequence alignments using a novel Family Clustering Algorithm to display the results in a user-friendly way. The software tool, Nearest-cell, has been incorporated into the X-ray data collection pipeline at the Diamond Light Source, and is also available as a public web service. The bulk of the thesis is devoted to the study and prediction of protein disorder. Initially, trying to update and extend an existing predictor, RONN, the limitations of the method were exposed and a novel predictor (called MoreRONN) was developed that incorporates a novel sequence-based clustering approach to disorder data inferred from the PDB and DisProt. MoreRONN is now clearly the best-in-class disorder predictor and will soon be offered as a public web service. The third project explores the development of a clustering algorithm for protein structural fragments that can work on the scale of the whole PDB. While protein structures have long been clustered into loose families, there has to date been no comprehensive analytical clustering of short (~6 residue) fragments. A novel fragment clustering tool was built that is now leading to a public database of fragment families and representative structural fragments that should prove extremely helpful for both basic understanding and experimentation. Together, these three projects exemplify how cutting-edge computational approaches applied to extensive protein structure libraries can provide user-friendly tools that address critical everyday issues for structural biologists.

Ο στόχος αυτής της εργασίας ήταν η αναγνώριση έξι βασικών κινήσεων του χεριού με χρήση δύο συστημάτων. Όντας θέμα διεπιστημονικού επιπέδου έγινε μελέτη της ανατομίας των μυών του πήχη, των βιοσημάτων, της μεθόδου της ηλεκτρομυογραφίας (ΗΜΓ) και μεθόδων αναγνώρισης προτύπων. Παράλληλα, το σήμα περιείχε αρκετό θόρυβο και έπρεπε να αναλυθεί, με χρήση του EMD, να εξαχθούν χαρακτηριστικά αλλά και να μειωθεί η διαστασιμότητά τους, με χρήση των RELIEF και PCA, για βελτίωση του ποσοστού επιτυχίας ταξινόμησης. Στο πρώτο μέρος γίνεται χρήση συστήματος ΗΜΓ της Delsys αρχικά σε ένα άτομο και στη συνέχεια σε έξι άτομα με το κατά μέσο όρο επιτυχημένης ταξινόμησης, για τις έξι αυτές κινήσεις, να αγγίζει ποσοστά άνω του 80%. Το δεύτερο μέρος περιλαμβάνει την κατασκευή αυτόνομου συστήματος ΗΜΓ με χρήση του Arduino μικροελεγκτή, αισθητήρων ΗΜΓ και ηλεκτροδίων, τα οποία είναι τοποθετημένα σε ένα ελαστικό γάντι. Τα αποτελέσματα ταξινόμησης σε αυτή την περίπτωση αγγίζουν το 75%. / The aim of this work was to identify six basic movements of the hand using two systems. Being an interdisciplinary topic, there has been conducted studying in the anatomy of forearm muscles, biosignals, the method of electromyography (EMG) and methods of pattern recognition. Moreover, the signal contained enough noise and had to be analyzed, using EMD, to extract features and to reduce its dimensionality, using RELIEF and PCA, to improve the success rate of classification. The first part uses an EMG system of Delsys initially for an individual and then for six people with the average successful classification, for these six movements at rates of over 80%. The second part involves the construction of an autonomous system EMG using an Arduino microcontroller, EMG sensors and electrodes, which are arranged in an elastic glove. Classification results in this case reached 75% of success.

La présente thèse s'intéresse à l'utilisation de méta-données textuelles pour l'analyse d'image. Nous cherchons à utiliser ces informations additionelles comme supervision faible pour l'apprentissage de modèles de reconnaissance visuelle. Nous avons observé un récent et grandissant intérêt pour les méthodes capables d'exploiter ce type de données car celles-ci peuvent potentiellement supprimer le besoin d'annotations manuelles, qui sont coûteuses en temps et en ressources. Nous concentrons nos efforts sur deux types de données visuelles associées à des informations textuelles. Tout d'abord, nous utilisons des images de dépêches qui sont accompagnées de légendes descriptives pour s'attaquer à plusieurs problèmes liés à la reconnaissance de visages. Parmi ces problèmes, la vérification de visages est la tâche consistant à décider si deux images représentent la même personne, et le nommage de visages cherche à associer les visages d'une base de données à leur noms corrects. Ensuite, nous explorons des modèles pour prédire automatiquement les labels pertinents pour des images, un problème connu sous le nom d'annotation automatique d'image. Ces modèles peuvent aussi être utilisés pour effectuer des recherches d'images à partir de mots-clés. Nous étudions enfin un scénario d'apprentissage multimodal semi-supervisé pour la catégorisation d'image. Dans ce cadre de travail, les labels sont supposés présents pour les données d'apprentissage, qu'elles soient manuellement annotées ou non, et absentes des données de test. Nos travaux se basent sur l'observation que la plupart de ces problèmes peuvent être résolus si des mesures de similarité parfaitement adaptées sont utilisées. Nous proposons donc de nouvelles approches qui combinent apprentissage de distance, modèles par plus proches voisins et méthodes par graphes pour apprendre, à partir de données visuelles et textuelles, des similarités visuelles spécifiques à chaque problème. Dans le cas des visages, nos similarités se concentrent sur l'identité des individus tandis que, pour les images, elles concernent des concepts sémantiques plus généraux. Expérimentalement, nos approches obtiennent des performances à l'état de l'art sur plusieurs bases de données complexes. Pour les deux types de données considérés, nous montrons clairement que l'apprentissage bénéficie de l'information textuelle supplémentaire résultant en l'amélioration de la performance des systèmes de reconnaissance visuelle.

Mit dem Begriff Wellenleiterquantenelektrodynamik (WQED) wird gemeinhin die Physik des quantisierten und in eindimensionalen Wellenleitern geführten Lichtes in Wechselwirkung mit einzelnen Emittern bezeichnet. In dieser Arbeit untersuche ich Effekte der WQED für einzelne Dreiniveausysteme (3NS) bzw. Paare von Zweiniveausystemen (2NS), die in den Wellenleiter eingebettet sind. Hierzu bediene ich mich hauptsächlich numerischer Methoden und betrachte die Modellsysteme im Rahmen der Drehwellennäherung. Ich untersuche die Dynamik der Streuung einzelner Photonen an einzelnen, in den Wellenleiter eingebetteten 3NS. Dabei analysiere ich den Einfluss dunkler bzw. nahezu dunkler Zustände der 3NS auf die Streuung und zeige, wie sich mit Hilfe stationärer elektrischer Treibfelder gezielt auf die Streuung einwirken lässt. Ich quantifiziere Verschränkung zwischen dem Lichtfeld im Wellenleiter und den Emittern mit Hilfe der Schmidt-Zerlegung und untersuche den Einfluss der Form der Einhüllenden eines Einzelphotonpulses auf die Ausbeute der Verschränkungserzeugung bei der Streuung des Photons an einem einzelnen Lambda-System im Wellenleiter. Hier zeigt sich, dass die Breite der Einhüllenden im k-Raum und die Emissionszeiten der beiden Übergänge des 3NS die maßgeblichen Parameter darstellen. Abschließend ergründe ich die Emissionsdynamik zweier im Abstand L in den Wellenleiter eingebetteter 2NS. Diese Dynamik wird insbesondere durch kavitätsartige und polaritonische Zustände des Systems aus Wellenleiter und Emitter ausschlaggebend beeinflusst. Bei der kollektiven Emission der 2NS treten - abhängig vom Abstand L - Sub- bzw. Superradianz auf. Dabei nimmt die Intensität dieser Effekte mit längerem Abstand L zu. Diese Eigenart lässt sich auf die Eindimensionalität des Wellenleiters zurückführen. / The field of waveguide quantum electrodynamics (WQED) deals with the physics of quantised light in one-dimensional (1D) waveguides coupled to single emitters. In this thesis, I investigate WQED effects for single three-level systems (3LS) and pairs of two-level systems (2LS), respectively, which are embedded in the waveguide. To this end, I utilise numerical techniques and consider all model systems within the rotating wave approximation. I investigate the dynamics of single-photon scattering by single, embedded 3LS. In doing so, I analyse the influence of dark and almost-dark states of the 3LS on the scattering dynamics. I also show, how stationary electrical driving fields can control the outcome of the scattering. I quantify entanglement between the waveguide''s light field and single emitters by utilising the Schmidt decomposition. I apply this formalism to a lambda-system embedded in a 1D waveguide and study the generation of entanglement by scattering single-photon pulses with different envelopes on the emitter. I show that this entanglement generation is mainly determined by the photon''s width in k-space and the 3LS''s emission times. Finally, I explore the emission dynamics of a pair of 2LS embedded by a distance L into the waveguide. These dynamics are primarily governed by bound states in the continuum and by polaritonic atom-photon bound-states. For collective emission processes of the two 2LS, sub- and superradiance appear and depend strongly on the 2LS''s distance: the effects increase for larger L. This is an exclusive property of the 1D nature of the waveguide.

This work describes static and dynamic problems with one messenger or multiple number of messengers and suggests a possibility of solving such problems with modified heuristic methods. To solve messenger problem, modified nearest neighbor heuristic, modified insertion heuristic and modified exchange heuristic are used. The main contribution of this work are applications, developed in MS Excel, programmed with Visual Basic for Application, that can solve static and dynamic problems with one messenger or multiple number of messengers and that could be beneficial for companies that do business in messenger services.

Les travaux présentés dans cette thèse ont porté sur l’étude de faisabilité technique et logicielle du système i-GUIDE, système interactif de guidage des personnes dans les hypermarchés. Nous avons détaillé l’analyse fonctionnelle du besoin du système. Ensuite, nous avons étudié l’impact de l’intégration du système dans le magasin à travers le diagramme BPMN. Nous avons opté pour l’approche UML pour décrire les principales fonctionnalités de notre système ainsi que les objets nécessaires pour son bon fonctionnement. Une architecture du système i-GUIDE, basée sur la technologie RFID avec une application sous Android, a été présentée. Par ailleurs, nous avons proposé des approches d’optimisation de parcours dans un hypermarché basées sur la méthode de recherche tabou pour deux problèmes. Pour le premier problème, nous avons choisi le critère de la plus courte distance pour la détermination du chemin et pour le deuxième nous avons ajouté une contrainte de temps pour des articles en promotion. Avant de chercher le chemin le plus court à parcourir pour trouver les articles existants dans la liste de courses, nous avons proposé une méthode pour ladétermination des distances entre les articles de l’hypermarché pris deux à deux / The present work focuses on the technical feasibility study of i-GUIDE system which is a real time indoor navigation system dedicated to assist persons inside hypermarkets. We detailed its functional analysis. Then, we studied the impact of integrating the system inside hypermarkets. We opted for an UML design to describe its main functionalities and objects required. We presented architecture of i-GUIDE system based on RFID technology with an Android application. Furthermore, we introduced optimization approaches based on tabu search to compute the route visiting items existing in a shopping list for two problems. The first one treats the shortest path to pick up items and the second one adds a time constraint for promotional items. Before computing the shortest path, we introduced a method to determine distance between each two items existing in the hypermarket

No description available.