Over the last few years, Machine Learning has become a huge field with “Big Tech” companies sharing their experiences building machine learning infrastructure. Feature Stores, used as centralized data repositories for machine learning features, are seen as a central component to operational and scalable machine learning. With the growth in machine learning, there is, naturally, a tremendous growth in data used for training. Most of this data tends to sit in Parquet files in cloud object stores or data lakes and is used either directly from files or in-memory where it is used in exploratory data analysis and small batches of training. A majority of the data science involved in machine learning is done in Python, but the infrastructure surrounding it is not always directly compatible with Python. Often, query processing engines and feature stores end up having their own Domain Specific Language or require data scientists to write SQL code, thus leading to some level of ‘transpilation’ overhead across the system. This overhead can not only introduce errors but can also add up to significant time and productivity cost down the line. In this thesis, we conduct a systems research on the performance of offline feature stores and identify ways that allow us to pull out data from feature stores in a fast and efficient way. We conduct a model evaluation based on benchmark tests that address common exploratory data analysis and training use cases. We find that in the Hopsworks feature store, with the use of state-of-the-art, storage-optimized, format-aware, and vector execution-based query processing engine as well as using Arrow protocol from start to finish, we are able to see significant improvements in both creating batch training data (feature value reads) and creating Point-In-Time Correct training data. For batch training data created in-memory, Hopsworks shows an average speedup of 27x over Databricks (5M and 10M scale factors), 18x over Vertex, and 8x over Sagemaker across all scale factors. For batch training data as parquet files, Hopsworks shows a speedup of 5x over Databricks (5M, 10M, and 20M scale factors), 13x over Vertex, and 6x over Sagemaker across all scale factors. For creating in-memory Point-In-Time Correct training data, Hopsworks shows an average speedup of 8x over Databricks, 6x over Vertex, and 3x over Sagemaker across all scale factors. Similary for PIT-Correct training data created as file, Hopsworks shows an average speedup of 9x over Databricks, 8x over Vertex, and 6x over Sagemaker across all scale factors. Through the analysis of these experimental results and the underlying infrastructure, we identify the reasons for this performance gap and examine the strengths and limitations of the design. / Under de senaste åren har maskininlärning blivit ett stort område där ”Big Tech”-företag delar med sig av sina erfarenheter av att bygga infrastruktur för maskininlärning. Feature Stores, som används som centraliserade datalager för maskininlärningsfunktioner, ses som en central komponent för operativ och skalbar maskininlärning. Med tillväxten inom maskininlärning följer naturligtvis en enorm tillväxt av data som används för utbildning. De flesta av dessa data finns i Parquet-filer som lagras i molnobjektsbutiker eller datasjöar och används antingen direkt från filer eller i minnet där de används i explorativ dataanalys och små utbildningsbatcher. En majoritet av datavetenskapen inom maskininlärning görs i Python, men den omgivande infrastrukturen är inte alltid direkt kompatibel med Python. Ofta har motorer för frågebehandling och feature stores sina egna domänspecifika språk eller kräver att datavetare skriver SQL-kod, vilket leder till en viss nivå av `transpileringsoverhead' i hela systemet. Denna overhead kan inte bara leda till fel utan också till betydande tids- och produktivitetskostnader i slutändan. I den här avhandlingen genomför vi en systemstudie av prestandan hos offline feature stores och identifierar sätt som gör att vi kan ta fram data från feature stores på ett snabbt och effektivt sätt. Vi genomför en modellutvärdering baserad på benchmarktester som tar upp vanliga användningsfall för explorativ dataanalys och utbildning. Vi konstaterar att vi i Hopsworks feature store, med hjälp av en toppmodern, lagringsoptimerad, formatmedveten och vektorexekveringsbaserad frågebehandlingsmotor samt Arrow-protokoll från början till slut, kan se betydande förbättringar både när det gäller att skapa batchutbildningsdata (läsa featurevärden) och skapa Point-In-Time Correct-utbildningsdata. För batchutbildningsdata som skapats i minnet visar Hopsworks en genomsnittlig hastighet på 27x över Databricks (5M och 10M skalfaktorer), 18x över Vertex och 8x över Sagemaker över alla skalfaktorer. För batch-träningsdata som parkettfiler visar Hopsworks en hastighetsökning på 5x över Databricks (5M, 10M och 20M skalfaktorer), 13x över Vertex och 6x över Sagemaker över alla skalfaktorer. För att skapa Point-In-Time Correct-träningsdata i minnet visar Hopsworks en genomsnittlig hastighet på 8x över Databricks, 6x över Vertex och 3x över Sagemaker över alla skalfaktorer. På samma sätt för PIT-Correct träningsdata som skapats som fil, visar Hopsworks en genomsnittlig hastighet på 9x över Databricks, 8x över Vertex och 6x över Sagemaker över alla skalfaktorer. Genom att analysera dessa experimentella resultat och den underliggande infrastrukturen identifierar vi orsakerna till denna prestandaklyfta och undersöker styrkorna och begränsningarna i designen.

Dagens hälso- och sjukvårdssystem är byggda som en traditionell monolit men det finns fler sätt att utveckla dessa system på. Händelsedriven arkitektur med händel-sekällor för lagring av data är ett modernt sätt att utveckla system. Syftet med det här arbetet är att undersöka om den modernare arkitekturen skulle kunna vara lämpligt och säkert alternativ för hälso- och sjukvårdssystem. För att undersöka detta gjordes en litteraturstudie inom relevanta områden som händelsedriven arki-tektur, händelsekällor, mikrotjänster och monolitiska system. Resultatet ifrån litteraturstudien gav att det finns andra beprövade koncept från andra problemområden, så som säker kommunikation och säker lagring med spår-barhet, som om de användes med händelsedriven arkitektur skulle uppnå kraven för hälsosektorn. Forskningsfrågan kan besvaras positivt baserat på den genomförda lit-teraturstudien. Därtill utvecklades en enkel prototyp, som utan att innehålla säker-hetsaspekter, gör det lätt att observera för och nackdelar med händelsedrivna lös-ningar jämfört med exempelvis traditionella monolitiska lösningar. / Today's healthcare systems are built as a traditional monolith, but there are more ways to develop these systems. Event driven architecture with event sourcing for sto-ring data is a modern way of developing systems. The purpose of this work is to in-vestigate whether the more modern architecture could be a suitable and safe alter-native for healthcare systems. To investigate this, a literature study was conducted in relevant areas such as event driven architecture, event sources, microservices and monolithic systems. The result of the literature study showed that there are other proven concepts from other problem areas, such as secure communication and secure storage with tracea-bility, which if used with event driven architecture would meet the requirements of the health sector. The research question can be answered positively based on the completed literature study. In addition, a simple prototype was developed, which, without containing security aspects, makes it easy to observe the pros and cons of event driven solutions compared to, for example, traditional monolithic solutions.

No description available.

No description available.

No description available.

No description available.

[pt] Nos últimos anos as empresas varejistas começaram a investir em modelos menores de lojas, voltando-se para o atendimento de conveniência, que busca atender os novos hábitos de consumo da população. No entanto, esse modelo de loja representa um grande desafio logístico para as empresas, uma vez que por serem pequenas essas lojas normalmente não possuem espaço para estoque, necessitando de abastecimentos frequentes. A logística urbana apresenta alguns obstáculos como as restrições de circulação por tipo de veículo e as janelas de tempo, tornando a roteirização adequada fundamental para entregar o produto correto, no momento correto e na quantidade correta, garantindo assim a redução de veículos, distância percorrida e custos logísticos. Desta forma, este trabalho busca contribuir para o reabastecimento ótimo de lojas de conveniência localizadas em grandes centros urbanos através da otimização de rotas, considerando restrições de circulação e janelas de tempo disponíveis para abastecimento. Para isto, um estudo de caso foi feito em uma empresa do setor varejista utilizando uma ferramenta de código aberto baseada no Excel para resolver a variante do problema de roteirização de veículos. A ferramenta utilizada para resolver casos reais da empresa em estudo, considera o planejamento médio da demanda diária assim como a capacidade e disponibilidade dos veículos, tempos de descarga e restrições de horário para recebimento e circulação de veículos. Os resultados do estudo para os cenários discutidos apresentam uma redução de até 32 por cento no frete da empresa, além de oportunidades de ganhos operacionais frente à atual operação da empresa. / [en] In recent years retail companies have begun to invest in smaller store models, turning to the convenience service, which seeks to meet the new consumption habits of the population. However, this store model represents a major logistical challenge for companies, once they are small, usually do not have space for inventory and need frequent replenishments. City Logistics presents some obstacles such as traffic restrictions by vehicle type and time windows, making proper routing essential to deliver the right product at the right time and in the right amount, thus ensuring vehicle reduction, distance traveled and logistics costs. In this way, the objective of this paper is to contribute to the optimal replenishment of convenience stores located in large urban centers through an optimization of the routes, considering the traffic restrictions and the time windows available for supply. For this, a case study was done in a retail company using an Excel open source tool to solve the variant of the vehicle routing problem. The tool is used to solve real cases of the company under study, considering the average planning of daily demand as well as the capacity and availability of vehicles, unloading times and time constraints for receiving and circulating vehicles. The study results for the scenarios discussed present a reduction of up to 32 per cent in the company s freight as well as opportunities for operating gains against the company s current operation.

Through the introduction and popularisation of e-books and e-readers, the way books are read is changing. This paper aims to investigate the reading habits of five Swedish-based Kindle users to understand their reading ecologies and what place the Kindle has in their reading ecologies. The Kindle proves an interesting research focus as it is one of the most sold e-readers in the world, but has yet to establish itself on the Swedish market. The research focuses on three main themes: the reading ecologies and habits of the interviewees; why they use the Kindle; and how they use reviews on Kindle Store. The research uses the methods of communicative ecology mapping and qualitative interviews for collecting empirical data, which is then contextualised and analysed through the theories of communicative ecology, mediatization, and media as practice. The research shows that the interviewees prefer reading on digital devices, and that particular practices of reading are done in specific spatial dimensions. Three practices of reading are visible in the interviewees’ reading ecologies: news-reading, social media-reading, and Kindlereading. The interviewees use the Kindle as a replacement of the physical book, which is shown in the way the interviewees list the e-ink technology and lack of backlit screens as motivations for using the device, in addition to the vast amount of niched literature available on Kindle Store. Moreover, reviews on Kindle Store are valuable to the interviewees when buying books, but the type of book changes how much validity the reviews hold. The reviews, no matter if they are being read or written by the interviewees, are viewed as helping the community of readers who use Kindle in finding ‘good’ literature.

The Diploma thesis ,Selected aspects of the retail supplier-purchaser relations' defines British grocery retailing. It presents government bodies which regulate retail industry and important independent not-for-profit organizations which have a strong influence on the market situation. It introduces the reader into legislation which regulates supplier-purchaser relations, particularly the Supermarket Code of Practice and its revised version the Groceries Supply Code of Practice. The analytical part examines the British market at the macroeconomical level, outlines the grocery retailing structure, describes the supply chain, customer behaviour, points out certain pricing and product policy trends and presents subjects which are bonded by the provisions set out in the Groceries Supply Code of Practice. The findings on local market competition and barriers which prevent the entry by a competitor are of a high significance as well. Despite the introduction of legal regulation the supplier-purchaser relations remain unequal and unfair. The thesis proposes a forecast of possible future grocery market development in the United Kingdom.

Idag lever vi i ett samhälle som digitaliseras allt mer och mer. Från att handeln för 20 år sedan enbart utfördes i fysiska butiker till att konsumenter idag hemma från tv-sofforna enkelt kan beställa hem produkter via e-handel. Digitaliseringen har främjat handel på internet och gjort att e-handeln har etablerats på modemarknaden. Även sättet som konsumenter handlar, alltså köpprocessen, har påverkats av samhällets digitalisering. Enligt Workman och Studak (2006) består köpprocessen av följande fem steg: behovsidentifikation, informationssökning, alternativutvärdering, köp och utvärdering av köp. En annan aspekt som uppsatsen undersöker är livsstil och då relaterat till familje- och arbetssituation för Generation X och Y. Generation X och Y är ett utgångsläge i uppsatsen. Generation X är födda under mitten av 1960-talet till början av 1980-talet och kännetecknas av att vara karriärsinriktade med begränsad fritid. Internetanvändningen för generationen sker främst i samband med informationssökning och som verktyg, exempelvis e-mail. Generation Y är födda i början av 1980-talet till 1990-talets mitt. Generationen karaktäriseras av att vara flitiga internetanvändare som är relativt högutbildade, då majoriteten har eftergymnasial utbildning. Uppsatsens syfte är att undersöka om köpprocessen skiljer sig mellan Generation X och Y utifrån livsstil, samt om det påverkar konsumenters val av köp i fysisk butik eller via e- handel. Resultatet av undersökningen visar att köpprocessen ser likvärdig ut för de båda generationerna. Däremot skiljer sig generationerna åt när det kommer till hur faktorerna i köpprocessens tredje steg, alternativutvärdering, prioriteras. Å andra sidan är båda generationerna eniga om att pris är den högst prioriterade faktorn. När det kommer till val mellan köp i fysiska butiker och via e-handel visar resultatet att valet inte enbart har med generation och livsstil att göra. Däremot kan slutsats dras att vanor, värderingar och individuella preferenser har större påverkan på vart köp utförs, i fysisk butik eller via e- handel, än vad livsstil samt konsumenters generation har. / Today we are living in a society that is becoming more and more digitalized. Only 20 years ago all kind of fashion trading took place in physical stores. Today, 20 years later, consumers can easily through the Internet purchase products from their couches in front of the TV. The digitalization has encouraged commerce on the Internet, which has established e-commerce on the fashion market. The way consumers shop, in other words the buying process, has been affected by the digitalization in society. According to Workman and Studak (2006) the buying process consists of the five steps: need identification, information search, alternative evaluation, purchase and evaluation of purchase. Another aspect that this essay examines is the lifestyle of Generation X and Y related to family- and work situation. This essay is based on Generation X and Y. Generation X was born in the middle of the 1960 to the beginning of the 1980. The generation is characterized for being very focused on their career, which limit their leisure time. The Internet usage of the generation is mainly in relation to information search and as a tool, for example e-mail. Generation Y was born in the beginning of the 1980 to the middle of the 1990. The generation is known for using Internet frequently and is highly educated, since the majority of the generation has a post-secondary education. The purpose of this essay is to examine if the buying process differentiates between Generation X and Y based on lifestyle, and if this influences the choice consumers make between purchase in physical store or through e-commerce. The result of the study demonstrates that the buying processes of both generations are equal. However, the generations differentiate in the third step of the buying process, alternative evaluation, when it comes to priority of different variables. On the other hand, the generations were united about that price is the highest prioritized variable. When it comes to the choice between purchases in physical stores or through e-commerce the result presents that the choice does not only depend on what generation or lifestyle consumers has. Nevertheless, a conclusion can be that habit, values and individual preferences have bigger impact on where the purchases are executed, in physical stores and through e-commerce, than what lifestyle and generation have.