Nous proposons une nouvelle structure de données distribuée et scalable appelée LH*RSP2P conçue pour les environnements pair à pair(P2P).Les données de l'application forment un fichier d’enregistrements identifiés par les clés primaires. Les enregistrements sont dans des cases mémoires sur des pairs, adressées par le hachage distribué (LH*). Des éclatements créent dynamiquement de nouvelles cases pour accommoder les insertions. L'accès par clé à un enregistrement comporte un seul renvoi au maximum. Le scan du fichier s’effectue au maximum en deux rounds. Ces résultats sont parmi les meilleurs à l'heure actuelle. Tout fichier LH*RSP2P est également protégé contre le Churn. Le calcul de parité protège toute indisponibilité jusqu’à k cases, où k ≥ 1 est un paramètre scalable. Un nouveau type de requêtes, qualifiées de sûres, protège également contre l’accès à toute case périmée. Nous prouvons les propriétés de notre SDDS formellement par une implémentation prototype et des expérimentations. LH*RSP2P apparaît utile aux applications Big Data, sur des RamClouds tout particulièrement / We propose a new scalable and distributed data structure termed LH*RSP2P designed for Peer-to-Peer environment (P2P). Application data forms a file of records identified by primary keys. Records are in buckets on peers, addressed by distributed linear hashing (LH*). Splits create new buckets dynamically, to accommodate inserts. Key access to a record uses at most one hop. Scan of the file proceeds in two rounds at most. These results are among best at present. An LH*RSP2P file is also protected against Churn. Parity calculation recovers from every unavailability of up to k≥1, k is a scalable parameter. A new type of queries, qualified as sure, protects also against access to any out-of-date bucket. We prove the properties of our SDDS formally, by a prototype implementation and experiments. LH*RSP2P appears useful for Big Data manipulations, over RamClouds especially.

Με τη διατριβή έγινε ανασκόπηση των πιο γνωστών θεωριών που διέπουν τις πολυφασικές ροές. Ακολούθησε συλλογή και κριτική αξιολόγηση των θεωρητικών μοντέλων. Εκπονήθηκαν δύο προγράμματα σε Η/Υ, τόσο για την ομογενή ροή όσο και τη χωριστή ροή. Έγινε σύγκριση των θεωρητικών με υπάρχοντα πειραματικά αποτελέσματα. Σχεδιάστηκε και κατασκευάστηκε πειραματική διάταξη. Λήφθηκαν πειραματικές μετρήσεις και έγινε αξιολόγησή τους και σύγκριση με θεωρητικά αποτελέσματα. Προτάθηκε μέθοδος διόρθωσης της πρόβλεψης λειτουργίας μιας υδροπνευματικής αντλίας με τη βοήθεια του μοντέλου ‘drift-flux’ (μέθοδος CoSM). Προτάθηκαν νέοι ροϊκοί χάρτες κατάλληλοι για την παρουσίαση της λειτουργίας των υδροπνευματικών αντλιών, τον υπολογισμό του κλάσματος κενού και την μετάβαση των ροϊκών καταστάσεων. Προτάθηκε τρόπος προσδιορισμού της μετάβασης ‘slug-churn’ με απλή φωτογραφική μέθοδο (camera) και χρήση του μοντέλου ‘drift-flux’. Τα αποτελέσματα της διατριβής (πειράματα, μέθοδος CoSM, ροϊκοί χάρτες και μετάβαση slug-churn) παρουσιάστηκαν σε συνέδρια και επιστημονικά περιοδικά. / This PhD thesis deals with multiphase flows and air-lift pumps. All well-known theories concerning these two scientific fields are presented and analyzed in detail. Two computational codes were developed for homogeneous two-phase and separated three-phase flow. A comparison between theoretical results and experimental data followed. An experimental investigation was performed in a lab scale air-lift pump installation at Fluid Mechanics Laboratory, University of Patras. A new method for the precise prediction of the performance of a two-phase air-lift pump with the aid of drift-flux model was presented (CoSM method). New regime maps were introduced suitable for air-lift pump presentations. That means the direct view of the flow behaviour inside the air-lift pump, void fraction calculation and the regime transitions. An experimental method was presented for the prediction of slug-churn transition in two-phase flow, using a camera and ‘drift-flux’ model. The results of this work were presented in International Conferences and Journals.

This study aims to assess which supervised statistical learning method; random forest, logistic regression or K-nearest neighbor, that is the best at predicting banks customer churn. Additionally, the study evaluates which cross-validation set approach; k-Fold cross-validation or leave-one-out cross-validation that yields the most reliable results. Predicting customer churn has increased in popularity since new technology, regulation and changed demand has led to an increase in competition for banks. Thus, with greater reason, banks acknowledge the importance of maintaining their customer base.   The findings of this study are that unrestricted random forest model estimated using k-Fold is to prefer out of performance measurements, computational efficiency and a theoretical point of view. Albeit, k-Fold cross-validation and leave-one-out cross-validation yield similar results, k-Fold cross-validation is to prefer due to computational advantages.   For future research, methods that generate models with both good interpretability and high predictability would be beneficial. In order to combine the knowledge of which customers end their engagement as well as understanding why. Moreover, interesting future research would be to analyze at which dataset size leave-one-out cross-validation and k-Fold cross-validation yield the same results.

Given that employees are one of the most valuable assets of any organization, losing an employee has a detrimental impact on several aspects of business activities. Loss of competence, deteriorated productivity and increased hiring costs are just a small fraction of the consequences associated with high employee churn. To deal with this issue, organizations within many industries rely on machine learning and predictive analytics to model, predict and understand the cause of employee churn so that appropriate proactive retention strategies can be applied. However, up to this date, the problem of excessive churn prevalent in the healthcare industry has not been addressed. To fill this research gap, this study investigates the applicability of a machine learning-based employee churn prediction model for a Swedish healthcare organization. We start by extracting relevant features from real employee data followed by a comprehensive feature analysis using Recursive Feature Elimination (RFE) method. A wide range of prediction models including traditional classifiers, such as Random Forest, Support Vector Machine and Logistic Regression are then implemented. In addition, we explore the performance of ensemble machine learning model, XGBoost and neural networks, specifically Artificial Neural Network (ANN). The results of this study show superiority of an SVM model with a recall of 94.8% and a ROC-AUC accuracy of 91.1%. Additionally, to understand and identify the main churn contributors, model-agnostic interpretability methods are examined and applied on top of the predictions. The analysis has shown that wellness contribution, employment rate and number of vacations days as well as number of sick day are strong indicators of churn among healthcare employees. / Det sägs ofta att anställda är en verksamhets mest värdefulla tillgång. Att förlora en anställd har därmed ofta skadlig inverkan på flera aspekter av affärsverksamheter. Därtill hör bland annat kompetensförlust, försämrad produktivitet samt ökade anställningskostnader. Dessa täcker endast en bråkdel av konsekvenserna förknippade med en för hög personalomsättningshastighet. För att hantera och förstå hög personalomsättning har många verksamheter och organisationer börjat använda sig av maskininlärning och statistisk analys där de bland annat analyserar beteendedata i syfte att förutsäga personalomsättning samt för att proaktivt skapa en bättre arbetsmiljö där anställda väljer att stanna kvar. Trots att sjukvården är en bransch som präglas av hög personalomsättning finns det i dagsläget inga studier som adresserar detta uppenbara problem med utgångspunkt i maskininlärning. Denna studien undersöker tillämpbarheten av maskininlärningsmodeller för att modellera och förutsäga personalomsättning i en svensk sjukvårdsorganisation. Med utgångspunkt i relevanta variabler från faktisk data på anställda tillämpar vi Recursive Feature Elimination (RFE) som den primära analysmetoden. I nästa steg tillämpar vi flertalet prediktionsmodeller inklusive traditionella klassificerare såsom Random Forest, Support Vector Machine och Logistic Regression. Denna studien utvärderar också hur pass relevanta Neural Networks eller mer specifikt Artificial Neural Networks (ANN) är i syfte att förutse personalomsättning. Slutligen utvärderar vi precisionen av en sammansatt maskininlärningsmodell, Extreme Gradient Boost. Studiens resultat påvisar att SVM är en överlägsen model med 94.8% noggranhet. Resultaten från studien möjliggör även identifiering av variabler som mest bidrar till personalomsättning. Vår analys påvisar att variablerna relaterade till avhopp är friskvårdbidrag, sysselsättningsgrad, antal semesterdagar samt sjuktid är starkt korrelerade med personalomsättning i sjukvården.

Generally, service industries require a rapid innovation of service portfolios to gain and maintain a competitive advantage. In this context, service elimination is a tool of portfolio renewal, where customer retention is a strategic priority for companies. This is especially so because service elimination usually causes higher churn rates than an average churn in telecommunications. Thus, customer retention is seen as a major aspect in enhancing service elimination success. The purpose of this paper is to investigate the factors that increase customer churn in the case of service elimination. We use one of the three Hungarian telecommunication Operator's databases containing usage data three months before and after Service elimination in the course of a major service package reform. Contract-related information and demographics of 10 065 customers are used to differentiate between high and low churn factors, taking care of a possible sample selection problem. The results show that in the course of service elimination there is a significant positive relationship between price decrease, tenure, interaction intensity on the one, and customer retention on the other side. Besides these, demographics (age and residence) also play an important role in explaining churn rates during service elimination. Furthermore, we find that a higher monthly fee after elimination increases the customer´s usage intensity. This research aims to contribute both to service elimination, as well as to customer retention literature, by hierarchical modeling of retention and usage during service elimination with practical implications for decision-makers in rapidly innovating telecommunication markets. / Series: Department of Economics Working Paper Series

[pt] O tema central deste projeto é a retenção de clientes como estratégia competitiva para aumentar a lucratividade da empresa. O objetivo é desenvolver um modelo estatístico que relacione variáveis transacionais, demográficas e dados sobre o histórico de eventos com a probabilidade de cancelamento dos clientes assinantes de jornal e definir o perfil dos clientes com maior risco de desligamento. Em uma primeira etapa, este estudo fornece uma revisão teórica sobre lealdade, satisfação e marketing de relacionamento, a fim de buscar uma relação com a retenção de clientes. Em seguida, a revisão de literatura levantou as variáveis mais usadas na segmentação de clientes tais como: variáveis transacionais, geográficas, demográficas, psicográficas e comportamentais para definir o perfil dos clientes que cancelam e dos que não cancelam sua assinatura. Depois de construir um modelo teórico, a regressão logística foi utilizada como técnica estatística para desenvolver um modelo de previsão de cancelamento. Os resultados foram analisados com o auxílio do programa estatístico SPSS e conclui-se que o perfil do cliente que cancela a assinatura do jornal é o jovem de até 30 anos; com baixo nível sócio-demográfico; morador da baixada, subúrbio e outros estados que não o Rio de Janeiro; que tenha adquirido sua assinatura através do canal telemarketing ativo; com a assinatura da modalidade anual e forma de pagamento em boleto ou débito em conta corrente; clientes que adquiriram sua assinatura mais recentemente; que comprem menos de 3 produtos da empresa e que não tenham feito reclamações através da central de atendimento. O modelo final de previsão de cancelamento contou com 11 variáveis e a tabela de classificação mostrou uma taxa de acerto geral de 75,3%. A última etapa apresenta algumas conclusões, implicações e sugestões para pesquisas futuras. / [en] The core subject of this project is the customers´ retention as a competitive strategy to increase the company´s profitability. The goal is to develop a statistical model that links transactional and demographic variables and customer´s history data with the subscribers´ churn of a certain publication. In the first part, this study provides a revision on loyalty, satisfaction and relationship marketing theory in order to find a relation with customers´ retention. After that, the literature revision raised the most used variables for the segmentation of customers, such as: transactional, geographic, demographic, psycological and behavior variables to define the profile of the customer who churns and the profile of that one who doesn´t. After constructing a theoretical model, the logistic regression was used as a statistical technique to develop a model of cancellation forecasting. The results has been analyzed with the aid of statistical program SPSS and conclude that the profile of the customer who cancels the subscription of the publication is young up to 30 years old; with low social- demographic level; living at Baixada, Suburb, and other states than Rio De Janeiro; that bought the subscription through the outbound telemarketing sales channel; with one year subscription and payment through invoice or direct debit in current account; customers who has bought its signature more recently; that do not buy less than 3 other products of the company and that have not made complaints through the customer service. The final model of churn forecasting uses 11 variables and the classification table showed an accuracy of 75,3%. The last part presents some conclusions, implications and suggestions for future research.

Many interesting applications involve decision prediction based on a time-series sequence or a set of time-series sequences, which are referred to as time-series classification problems. Past classification analysis research predominately focused on constructing a classification model from training instances whose attributes are atomic and independent. Direct application of traditional classification analysis techniques to time-series classification problems requires the transformation of time-series data into non-time-series data attributes by applying some statistical operations (e.g., average, sum, etc). However, such statistical transformation often results in information loss. In this thesis, we proposed the Time-Series Classification (TSC) technique, based on the nearest neighbor classification approach. The result of empirical evaluation showed that the proposed time-series classification technique had better performance than the statistical-transformation-based approach.

Nous proposons une nouvelle structure de données distribuée et scalable appelée LH*RSP2P conçue pour les environnements pair à pair(P2P).Les données de l'application forment un fichier d'enregistrements identifiés par les clés primaires. Les enregistrements sont dans des cases mémoires sur des pairs, adressées par le hachage distribué (LH*). Des éclatements créent dynamiquement de nouvelles cases pour accommoder les insertions. L'accès par clé à un enregistrement comporte un seul renvoi au maximum. Le scan du fichier s'effectue au maximum en deux rounds. Ces résultats sont parmi les meilleurs à l'heure actuelle. Tout fichier LH*RSP2P est également protégé contre le Churn. Le calcul de parité protège toute indisponibilité jusqu'à k cases, où k ≥ 1 est un paramètre scalable. Un nouveau type de requêtes, qualifiées de sûres, protège également contre l'accès à toute case périmée. Nous prouvons les propriétés de notre SDDS formellement par une implémentation prototype et des expérimentations. LH*RSP2P apparaît utile aux applications Big Data, sur des RamClouds tout particulièrement

Submitted by Roberta Lorenzon (roberta.lorenzon@fgv.br) on 2011-06-10T16:45:29Z No. of bitstreams: 1 68090200023.pdf: 4324154 bytes, checksum: 9b6b7dca8ee0cb46eaf78ab5cafb94a2 (MD5) / Approved for entry into archive by Gisele Isaura Hannickel(gisele.hannickel@fgv.br) on 2011-06-10T17:13:24Z (GMT) No. of bitstreams: 1 68090200023.pdf: 4324154 bytes, checksum: 9b6b7dca8ee0cb46eaf78ab5cafb94a2 (MD5) / Approved for entry into archive by Gisele Isaura Hannickel(gisele.hannickel@fgv.br) on 2011-06-10T17:14:39Z (GMT) No. of bitstreams: 1 68090200023.pdf: 4324154 bytes, checksum: 9b6b7dca8ee0cb46eaf78ab5cafb94a2 (MD5) / Made available in DSpace on 2011-06-10T17:17:49Z (GMT). No. of bitstreams: 1 68090200023.pdf: 4324154 bytes, checksum: 9b6b7dca8ee0cb46eaf78ab5cafb94a2 (MD5) Previous issue date: 2011-01-11 / O objetivo deste trabalho é criar um modelo que permita predizer quais clientes possuem tendência a abandonar a empresa (churn) através da base de dados de solicitações de reparo dos clientes de banda larga. Este tema ganha importância à medida que a concorrência neste mercado se acirra e novas tecnologias para acesso à banda larga são disponibilizadas aos clientes. Para este estudo foram utilizadas as bases de dados de solicitações de reparo e de solicitações de desligamento do serviço. O primeiro modelo criado utilizou as variáveis tempo até o reparo e a quantidade de solicitações de reparo. Na busca por um modelo com maior nível de acerto, foi criado um segundo modelo no qual se consideraram as variáveis motivo da reclamação e causa do defeito. O resultados do estudo indicam que as variáveis tempo até o reparo e quantidade de solicitações de reparo influenciam na taxa de solicitações de desligamento, assim como alguns motivos de solicitação de reparo e causa de defeito são mais relevantes para a decisão de desligamento. No entanto, o nível de acerto dos modelos criados é baixo. Portanto, a utilização destes modelos para identificação e abordagem de clientes propensos a solicitar desligamento implica em ajustes no nível de certeza do modelo e consequente redução na quantidade de clientes a serem abordados.

How can we predict key decisions made by organizations in the presence of big data and on-demand information? In this dissertation we exploit a large repository of B2B real-time transactional data with service quality indicators and present evidence that organizational decision analytics apply both rational and boundedly-rational (i.e. behavioral) economic models. The dissertation’s findings demonstrate that both utility and heuristic models, respectively, play significant roles in predicting organizational decisions on churn, a key decision in this context. In the presence of a large data set the assumed rationality of organizations appears to provide accurate predictions in uncontrolled experiences and selected boundedly-rational decision rules appear to cause somatic states that make organizations more sensitive to past total qualities of service. This dissertation makes significant new contributions to the understanding of how organizations can effectively use big data to make key operational decisions. As a managerial implication, organizations must be alert to heuristics that might exacerbate the impact of total service pain on customer’s decision to churn.