Urinary tract infections (UTI) are a common bacterial infection. Diagnosing UTI can be done through urine culturing. While precise, culturing is both time and money consuming. Flow cytometry analysis (FCA) is a different technique that can calculate different attributes in a urine sample. This is both faster and cheaper. Though, the problem with FCA is that it cannot reliably diagnose patients. The aim of this thesis is to investigate how different screening methods perform when applied before culturing. The screening methods uses FCA and some general characteristics to predict UTI. Using machine learning algorithms, different screening methods were compared. The methods were altered using a sensitivity correction such that the sensitivity exceeded 95%. The performance was measured using obtained real life data consisting of 1316 samples and cross validation. The best savings achieved was obtained using random forest. It managed to save up to 46% of the load on the culturing process while keeping a sensitivity of 95.15%. The specificity were 72%. Even though the data set obtained was too small to reliable declare the real performance, the savings looks really promising.

Studiens syfte är att utifrån ett sociokulturellt perspektiv ta reda på hur lärare använder fysiska och intellektuella redskap i aritmetikundervisningen med barn i åldrarna 5–7 år. Syftet ska besvaras genom följande frågeställning: Hur gör lärare för att vägleda barn i att använda strukturerade fingermönster som ett verktyg för att lösa enklare aritmetikproblem? Studien utgår ifrån det sociokulturella perspektivet som ser lärande som en process som konstrueras i en social miljö där läraren vägleder barn i hur kulturella redskap kan användas för att förstå omvärlden. Studien genomfördes med en kvalitativ metod där data samlades in via videoobservationer. De elva videoinspelningarna som observerades innehöll olika lärares undervisningsaktiviteter som syftade till att utveckla barns taluppfattning. I undervisningsaktiviteterna utgjorde fingermönster en viktig del. Under observation gjordes anteckningar i ett observationsschema som skapades för studien. Materialet analyserades sedan genom kodning. Resultatet visar på olika sätt vilket lärare gör för att vägleda barn i att använda strukturerade fingermönster när de löser enklare aritmetikproblem. / The purpose of the study is to determine, from a sociocultural perspective, how teachers use physical and intellectual tools in arithmetic teaching with children aged 5-7. The purpose should be answered by the following question: How do teachers guide children in using structured finger patterns as a tool to solve an arithmetic problem? The study is based on the sociocultural perspective, which sees learning as a process constructed in a social environment where the teacher guides children in how cultural tools can be used to understand the outside world. The study was conducted using a qualitative method where data was collected with video observations. The eleven videos recordings observed contained the teaching activities of various teachers aimed at developing preventing speech perception. In the teaching activities finger patterns formed an important part. During the observations, notes were made in an observation schedule created for the study. The material was then analyzed by coding. The results show in different ways what teacher do to guide children in using structured finger patterns when solving simpler arithmetic problems.

No description available.

On the financial markets, investors search to achieve their economical goals while simultaneously being exposed to minimal risk. Volatility surfaces are used for estimating options' implied volatilities and corresponding option prices, which are used for various risk calculations. Currently, volatility surfaces are constructed based on yesterday's market information and are used for estimating options' implied volatilities today. Such a construction gets redundant very fast during periods of high volatility, which leads to inaccurate risk calculations. With an aim to reduce volatility surfaces' estimation errors, this thesis explores the possibilities of calibrating volatility surfaces intraday using incomplete market information. Through statistical analysis of the volatility surfaces' historical movements, characteristics are identified showing sections with resembling motion patterns. These insights are used to adjust the volatility surfaces intraday. The results of this thesis show that calibrating the volatility surfaces intraday can reduce the estimation errors significantly during periods of both high and low volatility. However, these results highly depend on the conducted choices when constructing and analyzing the volatility surfaces which leave room for further reasearch.

This thesis work examines anomaly detection methods on large data sets related to insurance funds. Starting from requirements of low time complexity, ease of implementation and thorough definitions of contextual- and collective anomalies, different modelling frameworks are examined. Twelve time series models and a replicator neural network are presented in detail, while other modelling methods that are considered promising are mentioned more briefly. It is shown that a very simple time series-model can be widely applied to detect collective anomalies. It is also shown that a relatively simple replicator neural network can be applied to detect contextual anomalies. / I detta examensarbete undersöks metoder för att upptäcka anomalier i stora datamängder relaterade till försäkringsbestånd. Med ansats från krav om låg tidskomplexitet, enkel implementation och grundliga definitioner av kontextuella- och kollektiva anomalier undersöks olika modelleringsramverk. Tolv tidsseriemodeller och ett neuralt replikatornätverk är presenterade i detalj, medans andra metoder som anses lovande nämns mer kortfattat. Det visas att en väldigt enkel tidsserie-modell kan appliceras brett för att upptäcka kollektiva anomalier. Det visas även att ett relativt simpelt neuralt replikatornätverk kan appliceras för att upptäcka kontextuella anomalier.

Although Sweden's geographical location entails a relatively low outdoor temperature for much of the year, the cooling of buildings is starting to become a problem. In this thesis it is showed that different types of machine learning models can be used to predict if Vasakronan's use of district cooling will pass a decided level of kW (binary models) and if it is possible to predict the next coming days power peak, that is, the highest amount of power of district cooling used, and when it will occur. A comparison is presented between logistic regression and support vector machine, as well as an LSTM model. The explanatory variables are mainly the internal measurements of the flow in the district cooling network and weather data. Logistic regression with reduced number of explanatory variables was chosen as the resulting binary model and for the continuous part the resulting LSTM had an RMSE of 160 and MAE of 103. The result shows that it is possible predict usage of tomorrow's power with relatively high certainty.

No description available.

The purpose of the study was to investigate how teachers teach to facilitate students' understanding of mathematical concepts. To answer our research question, the qualitative observation method was used on six different teachers who teach 4–6 in the subject of mathematics. The data collected from the study was processed through the thematic analysis which gave us results in the form of different categories. These categories are as follows: The teacher's use of the mathematical language, the teacher's use of aids during the lesson and the teacher's use of other words / synonyms. Through the study's previous research and theory, the socio-cultural perspective, we have analyzed the study's results. The results showed that the teachers who were observed used mostly the mathematical concepts in their teaching. They did this in the form of various aids, for example concrete material, the blackboard, the body language, digital aids.

I examensarbetet undersöks metoder för anomalidetektion i tidsserie data. Givet data för overnight index swaps (SEK), så har syntetiskt data skapats med olika ty-per av anomalier. Jämförelse mellan algoritmerna Isolation forest och Local outlierfactor görs genom att mäta respektive prestande för de syntetiska dataseten mot Accuracy, Precision, Recall, F-measure och Matthews correlation coefficient. / In the thesis, methods for anomaly detection in time series data are investigated. Given data for overnight index swaps (SEK), synthetic data has been created with different types of anomalies. Comparison between the Isolation forest and Local outlier factor algorithms is done by measuring the respective performances for the synthetic data sets against Accuracy, Precision, Recall, F-Measure and Matthews correlation coefficient.

Purchasing Mangers' Index (PMI) is an index published by the Institute for Supply Management. The index is based on a monthly survey answered by hundreds of purchasing managers in the manufacturing business all over the United States. The questions capture the activities of companies when it comes to orders, production, employment, supplier deliveries and inventories. PMI is an important and reliable macro economic factor and has shown a high correlation with the GDP which means that it may be of interest for a bank as Nordea to predict it. The purpose of this thesis is to evaluate whether it is possible to use neural networks, more precisely sequence-to-sequence models and recurrent neural networks, to forecast PMI with a multivariate data set. Simpler methods and models are also evaluated and compared to the neural networks. In total, there was two naive models, one Autoregressive Moving Average (ARMA) model, two linear regression models, and three neural network models that was implemented and compared to each other. All models were evaluated by their mean squared error (MSE) and mean absolute error (MAE).  By analyzing MSE and MAE for the different models it is shown that the ARMA-model predicted only slightly better than the persistence algorithm which assumes that PMI next month equals PMI current month. Only one neural network model performed better compared to the linear regression models, and the same model also gave a better prediction compared to the persistence algorithm.