Data mining and machine learning techniques are becoming more popular in helping companies with decision-making, due to these processes’ ability to automatically search through very large amounts of data and discover patterns that can be hard to see with human eyes. Onslip is one of the companies looking to achieve more value from its data. They provide a cloud-based cash register to small businesses, with a primary focus on restaurants. Restaurants are heavily affected by variations in sales. They sell products with short expiration dates, low profit margins and much of their expenses are tied to personnel. By predicting future demand, it is possible to plan inventory levels and make more effective employee schedules, thus reducing food waste and putting less stress on workers. The project described in this report, examines how sales forecasts can be improved by incorporating factors known to affect sales in the training of machine learning models. Several different models are trained to predict the future sales of 130 different restaurants, using varying amounts of additional information. The accuracy of the predictions are then compared against each other. Factors known to impact sales have been chosen and categorized into restaurant information, sales history, calendar data and weather information. The results show that, by providing additional information, the vast majority of forecasts could be improved significantly. In 7 of 8 examined cases, the addition of more sales factors had an average positive effect on the predictions. The average improvement was 6.88% for product sales predictions, and 26.62% for total sales. The sales history information was most important to the models’ decisions, followed by the calendar category. It also became evident that not every factor that impacts sales had been captured, and further improvement is possible by examining each company individually.

Many people today use social media in one way or another, and many of these platforms have released APIs developers can use to integrate social media in their applications. As many of these platforms share a lot of functionality we see a need for developing a library, to contain these, and ease the development process when working with the platforms. The purpose of this paper is to find common functionality and explore the possibility of generalization in this regard. We first look for common denominators between the top social media networks, and using this information we attempt to make an implementation to evaluate the practicality. After the development process we analyze our findings and discuss the usability and maintainability of such a library. Our findings show that the current state of the studied APIs are not suitable for generalization.

Low resource languages possess a limited number of digitized texts, making it challenging togenerate a satisfactory language audio corpus and information retrieval services. Low resourcelanguages, especially those spoken exclusively in African countries, lack a well-defined andannotated language corpus, making it a big obstacle for experts to provide a comprehensive textprocessing system. In this study, I Found out the best practices for producing and collectingdata for such zero/low resource languages by means of crowd-sourcing. For the purpose of thisstudy, a number of research articles (n=260) were extracted from Google Scholar, MicrosoftAcademic, and science direct. From these articles, only 60 of them, which met the inclusioncriteria' demands, were considered to review for eligibility. A full-text version of these researcharticles was downloaded and then were carefully screened to ensure eligibility. On the result ofthe eligibility assessment from potentially eligible 60 full-text articles for inclusion, only 25were selected and qualified to include in the final review. The final pool of the selected articles,concerning data generation practices and collection of low resource languages, can beconcluded that speech-based audio data is one of the most common and accessible data types.It can be contended that the collection of audio data from speech-based resources such as nativespeakers of the intended language and available audio recording by taking the advantages ofnew technologies is the most practical, cost-effective, and common method for collecting datafor low resource languages.

Syftet med denna studie var att undersöka hur en webbapplikation kan tillfredsställa behovet av att snabbt och smidigt kunna ge enklare gåvor till vänner och bekanta på Linköpings universitet. En webbapplikation utvecklades med fokus på de tre nyckelaspekterna navigerbarhet, tillförlitlighet och effektiv köpprocess. Under studiens gång utfördes användartester av webbapplikationen under olika faser av utvecklingen med de ovannämnda aspekterna som mätvärden. Metoder som användes i samband med användartesterna var Smiths L-formel för att utvärdera navigerbarhetensamt Concurrent Think-Aloud Procedure och Retrospective Probing. Resultaten från test-personerna samlades sedan in och användes som underlag för framtida utveckling. Trots att testresultaten låg till grund för utformningen av webbapplikationen ansågs navigerbarheten vara som lägst efter det sista testet enligt resultaten vilket är något som diskuteras i diskussionskapitlet. Vidare var målet med en effektiv köpprocess att kunna klicka, mata in persondata och ta beslut så få gånger som möjligt för att utföra ett köp, vilket slutligen resulterade i fyra klick, sex inmatningar och två beslut. Slutligen var den upplevda tillförlitligheten för webbapplikationen hög och lyckades uppnå sitt högsta testresultat efter de sista testerna. Slutsatsen indikerar att det krävs kompromisser för att skapa en webbapplikation som maximerar de tre nyckelaspekterna. Vad som gynnar en parameter missgynnaren annan och prioriteringar kan krävas. I Baljangåvans fall blev resultatet en webbapplikation som upplevdes navigerbar och tillförlitlig, samtidigt som köpprocessen hölls så effektiv som möjligt. / The purpose of this study was to examine whether or not a web application can satisfy the need to give simple gifts to friends and acquiantences, studying at Linköping University, in a quick and simple manner. The web application was developed keeping three key-aspects in mind: navigability, trust and an efficient method of payment. During the stages of the development, user tests were performed using the afore-mentioned aspects as metrics. Some of the methods used during testing were Smith’s L-formula, to evaluate navigability, as well as Concurrent Think-Aloud Procedure and Restrospective Probing. The obtained test data was gathered and became the foundation for future development. Even though the development was premised on the test data, however, the experienced navigability showed an all time low in the last test. This is additionally explored in the discussion part of this thesis. Furthermore, the objective of an efficient method of payment was to keep the amount of clicks, inputs and decisions needed in or-der to buy a product as low as possible. This ended up being four clicks, six inputs, and two decisions. Lastly, the perceived trust of the web application was high and attain edits best result during the last test iteration. The conclusion indicates that compromises are needed in order to develop a web application that maximizes the three key-aspects. There is always a trade-off between the different paramters and priorities are crucial in order to reach a disired balance. The result of this study was a web application with high perceived navigability and trust as well as an efficient method of payment.

Denna rapport undersöker hur en e-butik kan utvecklas för att uppnå god navigerbarhet med avseende på användarens subjektiva upplevelse. Med grund i befintlig vetenskaplig teori inom området användbarhet med fokus på navigerbarhet har en e-handelsplattform för försäljning och handel av överproducerad mat i form av matlådor utvecklats. Under utvecklingen som genomförts med en iterativ projektmetod har användartest genomförts. Resultaten från dessa användartest har sedan tillsammans med den vetenskapliga teorin använts i den fortsatta utvecklingen av webbapplikationen. Med utgångspunkt i resultat från användartester och vetenskaplig teori har frågeställningen ”Hur kan en webbapplikation ämnad som handelsplattform mellan restauranger och konsumenter för försäljning av överproducerad lunch utformas för att uppnå god navigerbarhet med avseende på användarens subjektiva upplevelse?” undersökts och besvarats. Det konkluderades att en viktig aspekt för navigerbarhet är att användaren ständigt är medveten om var i navigationsträdet denne befinner sig. Samtidigt bör det vara tydligt när och var användaren navigerar på applikationen vilket bland annat kan underlättas genom animationer eller färgmarkeringar i navigationsfältet. Det visade sig också vara viktigt att omdirigera användaren till relevanta platser på applikationen efter nyttjandet av viss funktionalitet. Vidare drogs slutsatsen att trasiga eller otydliga länkar reducerade den upplevda navigerbarheten. Länkar bör alltså utföra förväntad funktionalitet. En annan viktig faktor som påverkar navigerbarhet är vilken information som delges användaren och på vilket sätt denna presenteras.

Biometrics are distinctive for each person, and can not be given away or hacked like a password. Keystroke dynamics is a behavioral biometric characteristic that can be used as a complementary authentication step [1]. In online examinations it is difficult to make sure that each student write their own work. Keystroke dynamics from these examinations could be used to detect attempted cheating. To detect cheating attempts, a Gaussian Mixture Models with Universal Background Model (GMM-UBM) was implemented, and tested on benchmark data set recorded from online examinations written in free text. The use of a Universal Background Model (UBM) allows students to be enrolled using a limited amount of data, making the sug- gested approach suitable for the use case. The use of a GMM-UBM resulted in an Equal Error Rate (ERR) of 5.4% and an accuracy of 94.5%.

A parallel and concurrent garbage collector offers low latency spikes. A common approach in such collectors is to move objects around in memory without stopping the application. This imposes additional overhead on an application in the form of tracking objects' movements, so that all pointers to them, can eventually be updated to the new locations. Typical ways of storing this information suffer from pathological cases where the size of this "forwarding information" can theoretically become as big as the heap itself. If we dimension the application for the pathological case this would be a waste of resources, since the memory usage is usually significantly less. This makes it hard to determine an application's memory requirements. In this thesis, we propose a new design that trades memory for CPU, with a maximum memory overhead of less than 3.2% memory overhead. To evaluate the impact of this trade-off, measurements on application execution time was performed using the benchmarks from the DaCapo suite and SPECjbb2015. For 6 configurations in DaCapo a statistically significant negative effect on execution time in the range of 1-3% was found for the new design. For 10 configurations in DaCapo no change in execution times was shown in statistically significant data and two configurations in DaCapo showed a statistically significant shorter execution times for the new design on 6% and 22%, respectively. In SPECjbb2015, both max-jOPS and critical-jOPS has, for the new design, a statistically significant performance regression of ~2%. This suggests that for applications were stable and predictable memory footprint is important, this approach could be something to consider.

Developing user interfaces that consist of lists on native mobile platforms is complex. This project aims at reducing the complexity for application developers when developing dynamic interactive lists on the iOS-platform by creating an abstraction that lets the application developer write code on a higher abstraction level. The result is a library that creates an abstraction that developers can use to build list user interfaces in a declarative way.

In this paper, predictions of future price movements of a major American stock index was made by analysing past movements of the same and other correlated indices. A model that has shown very good results in speech recognition was modified to suit the analysis of financial data and was then compared to a base model, restricted by assumptions made for an efficient market. The performance of any model, that is trained by looking at past observations, is heavily influenced by how the division of the data into train, validation and test sets is made. This is further exaggerated by the temporal structure of the financial data, which means that the causal relationship between the predictors and the response is dependent in time. The complexity of the financial system further increases the struggle to make accurate predictions, but the model suggested here was still able to outperform the naive base model by more than 20 percent. The model is, however, too primitive to be used as a trading system, but suitable modifications, in order to turn the model into one, will be discussed in the end of the paper.

This thesis addresses the question of how to deploy the infrastructure for wireless networks carrying high data rate services. It starts from the assumption that a major part of the costs for installing wireless infrastructure is caused by antenna site acquisition, planning of network coverage and capacity as well as manpower for setting up and wiring the antenna sites. Specific attention is therefore paid to simplifying the network installation and hence reducing the overall costs. The results show that the proposed user-deployment approach, where Access Points (APs) are set up by the customers themselves, can achieve coverage and data rates comparable to pre-planned networks with properly placed and wired APs. Typical for user-deployment is that APs are set up in an ad-hoc fashion, wherever wireless access is needed and a wired backbone infrastructure is available. A number of typical usage scenarios are developed for identifying characteristic situations and places where high data-rate applications are likely to be used. Service requirements and essential technical parameters are derived and motivated, based on these scenarios. A system design is proposed, featuring an air-interface with multi-carrier modulation and slow link-adaptation. Both coverage and capacity of this system, which achieves link-layer data rates between 40 and 130Mbps, are then studied in the different usage scenarios. The user-deployment approach, a core supposition in this thesis, requires the considered networks to be operated in an unlicensed fashion. Sufficient spectrum for unlicensed wireless services is allocated around 5, 17, 24 and 60GHz. The 5GHz band has been studied thoroughly in conjunction with the development of HiperLAN/2 and IEEE 802.11. This thesis focuses on the 17 and 60GHz bands and assumes that the performance of a system operating at 24GHz can be to some extent approximated from these results. An in-depth investigation of propagation properties at 17 and 60GHz shows that achievable cell radii are rather small. Shadowing severely impairs coverage and achievable data rates of a wireless network. A large number of APs is therefore necessary for providing sufficiently high signal levels to transmit high data rates. The shadowing problem is particularly severe for ad-hoc installations. Two different deployment scenarios and their impact on the system performance are investigated: arbitrary placed, wall-mounted APs and regularly placed, ceiling-mounted APs. The first represents the user-deployment paradigm and is certainly the cheapest method; the latter requires coarse network planning and suitable wiring and will hence be more costly. Results show that both installation methods achieve comparable performance for dense infrastructures, e.g. indoor environments. Since user-deployment is simpler and cheaper, it should be preferred in this case. Sparse networks are typical for large, open buildings or outdoor areas. In that case, regular installation should be favored, since cells typically overlap very little and achieving coverage is difficult. Consequently, some form of network planning is needed. The 60GHz band is best suited for indoor applications with a dense infrastructure, since achievable cell radii are very limited. However, very high data rates and capacities can be offered due to the large amount of bandwidth allocated at 60GHz. If capacity is less important, the 17GHz band should be preferred. The better propagation characteristics allow larger cells and fewer APs are required for reliable coverage, but the attainable network capacity is limited by the rather small amount of unlicensed spectrum. User-deployment is generally suitable for indoor applications. A slightly denser infrastructure will be required to compensate for the lack of network planning, but the costs for additional hardware will be likely insignificant compared to the potential savings by avoiding coverage planning and additional wiring. / QC 20110308