Dažnai iškyla būtinybė nustatyti ir giliau pažinti daugiamačių duomenų struktūrą: susidariusius klasterius, itin išsiskiriančius objektus, objektų tarpusavio panašumą ir skirtingumą. Vienas iš sprendimų būdų – duomenų dimensijos mažinimas ir jų vizualizavimas. Kai analizuojamos didelės duomenų aibės, tikslinga prieš vizualizavimą sumažinti ne tik dimensiją, bet ir duomenų skaičių. Šio darbo tyrimų sritis yra daugiamačių duomenų skaičiaus mažinimas ir duomenų atvaizdavimas plokštumoje. Disertacijoje nagrinėjami dirbtiniais neuroniniais tinklais grindžiami vektorių kvantavimo ir dimensijos mažinimu pagrįsti vizualizavimo metodai. Kaip alternatyva saviorganizuojančių neuroninių tinklų ir daugiamačių skalių junginiams, darbe pasiūlyti nuoseklus neuroninių dujų ir daugiamačių skalių junginys bei integruotas, atsižvelgiantis į neuroninių dujų metodo mokymosi eigą ir leidžiantis gauti tikslesnę daugiamačių vektorių projekciją plokštumoje. Junginiais gautų vaizdų kokybės vertinimui pasirinkti Konigo matas, Spirmano koeficientas bei MDS paklaida. Šie matai leidžia kiekybiškai įvertinti panašumų išlaikymą po daugiamačių duomenų transformavimo į mažesnės dimensijos erdvę. Taip pat pasiūlyti dvimačių vektorių pradinių koordinačių parinkimo būdai nuosekliame junginyje ir integruoto junginio pirmame mokymo bloke bei koordinačių reikšmių priskyrimo būdai integruoto junginio kituose mokymo blokuose. Eksperimentiškai nustatyta kvantavimo paklaidos priklausomybė nuo neuroninių dujų tinklo... [toliau žr. visą tekstą] / Often there is a need to establish and understand the structure of multidimensional data: their clusters, outliers, similarity and dissimilarity. One of solution ways is a dimensionality reduction and visualization of the data. If a huge datasets is analyzed, it is purposeful to reduce the number of the data items before visualization. The area of research is reduction of the number of the data analyzed and mapping the data in a plane. In the dissertation, vector quantization methods, based on artificial neural networks, and visualization methods, based on a dimensionality reduction, have been investigated. The consecutive and integrated combinations of neural gas and multidimensional scaling have been proposed here as an alternative to combinations of self-organizing maps and multidimensional scaling. The visualization quality is estimated by König’s topology preservation measure, Spearman’s rho and MDS error. The measures allow us to evaluate the similarity preservation quantitatively after a transformation of multidimensional data into a lower dimension space. The ways of selecting the initial values of two-dimensional vectors in the consecutive combination and the first training block of the integrated combination have been proposed and the ways of assigning the initial values of two-dimensional vectors in all the training blocks, except the first one, of the integrated combination have been developed. The dependence of the quantization error on the values of training... [to full text]

This thesis presents a semi-automatic method to analyze morphometric features and landscape elements based on Self Organizing Map (SOM) as an unsupervised Artificial Neural Network algorithm in two completely different environments: 1) the Man and Biosphere Reserve “Eastern Carpathians” (Central Europe) as a complex mountainous humid area and 2) Lut Desert, Iran, a hyper arid region characterized by repetition of wind-eroded features. In 2003, the National Aeronautics and Space Administration (NASA) released the SRTM/ SIR-C band data with 3 arc seconds (approx. 90 m resolution) grid for approximately 80 % of Earth’s land surface. The X-band SRTM data were processed with a 1 arc second (approx. 30 m resolution) grid by the German space agency, DLR and the Italian space agency ASI, but due to the smaller X-SAR ground swath, large areas are not covered. The latest version 3.0 SRTM/C DEM and SRTM/X band DEM were re-projected to 90 and 30 m UTM grid and used to generate morphometric parameters of first order (slope) and second order (cross-sectional curvature, maximum curvatures and minimum curvature) by using a bivariate quadratic surface. The morphometric parameters are then used in a SOM to identify morphometric features (or landform elements) e.g. planar, channel, ridge in mountainous areas or yardangs (ridge) and corridors (valley) in hyper-arid areas. Geomorphic phenomena and features are scale-dependent and the characteristics of features vary when measured over different spatial extents or different spatial resolution. Morphometric parameters were derived for nine window sizes of the 90 m DEM ranging from 5 × 5 to 55 ×55. Analysis of the SOM output represents landform entities with ground areas from 450 m to 4950 m that is local to regional scale features. Effect of two SRTM resolutions, C and X bands is studied on morphometric feature identification. The difference change analysis revealed the quantity of resolution dependency of morphometric features. Increasing the DEM spatial resolution from 90 to 30 m (corresponding to X band) by interpolation resulted in a significant improvement of terrain derivatives and morphometric feature identification. Integration of morphometric parameters with climate data (e.g. Sum of active temperature above 10 ° C) in SOM resulted in delineation of morphologically homogenous discrete geo-ecological units. These units were reclassified to produce a Potential Natural Vegetation map. Finally, we combined morphometric parameters and remotely sensed spectral data from Landsat ETM+ to identify and characterize landscape elements. The single integrated data set of geo-ecosystems shows the spatial distribution of geomorphic, climatic and biotic/cultural properties in the Eastern Carpathians. The results demonstrate that a SOM is a very efficient tool to analyze geo-morphometric features under diverse environmental conditions and at different scales and resolution. Finer resolution and decreasing window size reveals information that is more detailed while increasing window size and coarser resolution emphasizes more regional patterns. It was also successfully applied to integrate climatic, morphometric parameters and Landsat ETM+ data for landscape analysis. Despite the stochastic nature of SOM, the results are not sensitive to randomization of initial weight vectors if many iterations are used. This procedure is reproducible with consistent results. / Avhandlingen presenterar en halvautomatisk metod för att analysera morfometriska kännetecken och landskapselement som bygger på Self Organizing Map (SOM), en oövervakad Artificiell Neural Nätverk algoritm, i två helt skilda miljöer: 1) Man and Biosphere Reserve "Eastern Carpathians" (Centraleuropa) som är ett komplext, bergigt och humid område och 2) Lut öken, Iran, en extrem torr region som kännetecknas av återkommande vinderoderade objekt. Basen för undersökningen är det C-band SRTM digital höjd modell (DEM) med 3 bågsekunder rutnät som National Aeronautics and Space Administration släppte 2003 för ungefär 80 % av jordens yta. Dessutom används i ett mindre område X-band SRTM DEM med 1 bågsekund rutnät av den tyska rymdagenturen DLR. DEM transformerades till 90 och 30 m UTM nätet och därav genererades morfometriska parametrar av första (lutning) och andra ordning (tvärsnittböjning, största och minsta böjning). De morfometriska parametrar används sedan i en SOM för att identifiera morfometriska objekt (eller landform element) t.ex. plan yta, kanal, kam i bergsområden eller yardangs (kam) och korridorer (dalgångar) i extrem torra områden. Geomorfiska fenomen och objekt är skalberoende och kännetecken varierar med geografiska områden och upplösning. Morfometriska parametrar har härletts från 90 m DEM för nio fönsterstorlekar från 5 × 5 till 55 × 55. Resultaten representerar landform enheter för områden från 450 m till 4950 m på marken dvs. lokal till regional skala. Inflytande av två SRTM upplösningar i C och X-banden har studerats för identifikation av morfometriska objekt. Förändringsanalys visade storleken av upplösningsberoende av morfometriska objekt. Ökning av DEM upplösningen från 90 till 30 m (motsvarande X-bandet) genom interpolation resulterade i en betydande förbättring av terräng parametrar och identifiering av morfometriska objekt. Integration av morfometriska parametrar med klimatdata (t.ex. summan av aktiv temperatur över 10° C) i SOM resulterade i avgränsningen av homogena geoekologiska enheter. Dessa enheter ha används för att producera en karta av potentiell naturlig vegetation. Slutligen har vi kombinerat morfometriska parametrar och multispektrala fjärranalysdata från Landsat ETM för att identifiera och karaktärisera landskapselement. Dessa integrerade ekosystem data visar den geografiska fördelningen av morfometriska, klimatologiska och biotiska/kulturella egenskaper i östra Karpaterna. Resultaten visar att SOM är ett mycket effektivt verktyg för att analysera geomorfometriska egenskaper under skilda miljöförhållanden, i olika skalor och upplösningar. Finare upplösning och minskad fönsterstorlek visar information som är mer detaljerad. Ökad fönsterstorlek och grövre upplösning betonar mer regionala mönster. Det var också mycket framgångsrikt att integrera klimatiska och morfometriska parametrar med Landsat ETM data för landskapsanalys. Trots den stokastiska natur av SOM, är resultaten inte känsliga för slumpvisa värden i de ursprungliga viktvektorerna när många iterationer används. Detta förfarande är reproducerbart med bestående resultat. / QC 20100924

nÃo hà / Nesta tese aborda-se o problema de identificaÃÃo de sistemas dinÃmicos sobre a Ãtica dos modelos locais, em que o espaÃo de entrada à particionado em regiÃes de operaÃÃo menores sobre as quais sÃo construÃdos modelos de menor complexidade (em geral, lineares). Este tipo de modelo à uma alternativa aos chamados modelos globais em que a dinÃmica do sistema à identificada usando-se uma Ãnica estrutura (em geral, nÃo-linear) que cobre todo o espaÃo de entrada. Assim, o tema alvo desta tese à o projeto de modelos lineares locais cujo espaÃo de entrada à particionado por meio do uso de algoritmos de quantizaÃÃo vetorial, principalmente aqueles baseados em redes neurais competitivas. Para este fim, sÃo propostos trÃs novos modelos lineares locais baseados na rede SOM (self-organizing map), que sÃo avaliados na tarefa de identificaÃÃo do modelo inverso de quatro sistemas dinÃmicos comumente usados na literatura em benchmarks de desempenhos. Os modelos propostos sÃo tambÃm comparados com modelos globais baseados nas redes MLP (multilayer perceptron) e ELM (extreme learning machines), bem como com outros modelos lineares locais, tais como o modelo fuzzy Takagi-Sugeno e o modelo neural LLM (local linear mapping). Um amplo estudo à realizado visando comparar os desempenhos de todos os modelos supracitados segundo trÃs critÃrios de avaliaÃÃo, a saber: (i) erro mÃdio quadrÃtico normalizado, (ii) anÃlise dos resÃduos, e (iii) teste estatÃstico de Kolmogorov-Smirnov. De particular interesse para esta tese, à a avaliaÃÃo da robustez dos modelos locais propostos com relaÃÃo ao algoritmo de quantizaÃÃo vetorial usado no treinamento do modelo. Os resultados obtidos indicam que os desempenhos dos modelos locais propostos sÃo superiores aos dos modelos globais baseados na rede MLP e equivalentes aos modelos globais baseados na rede ELM. / In this thesis the problem of nonlinear system identification is approached from the viewpoint of local models. The input space is partitioned into smaller operational regions with lower complexity models (usually linear) built for each one. This type of model is an alternative to global models, for which the system dynamics is identified using a single structure (usually nonlinear ones) that covers the whole input space. The aim of this thesis is to design of local linear models whose input space is partitioned by means of vector quantization algorithms, special those based on competitive learning neural networks. For this purpose, three novel local linear modeling methods based on the SOM (self-organizing map) are introduced and evaluated on the identification of the inverse model of four dynamical systems commonly used in the literature for performance benchmarking. The proposed models are also compared with global models based on the MLP (multilayer perceptron) and ELM (extreme learning machines), as well as with alternative local linear models, such as the Takagi-Sugeno fuzzy model and the LLM(local linear mapping) neural model. A comprehensive study is carried out to compare the performances of all the aforementioned models according to three evaluation criteria, namely: (i) normalized mean squared error, (ii) residual analysis, and (iii) Kolmogorov-Smirnov test. Of particular interest to this thesis is the evaluation of the robustness of the proposed local models with respect to the vector quantization algorithm used to train the model. The obtained results indicates that the performance of the proposed local models are superior to those achieved by the MLP-based global models and equivalent to those achieved by ELM-based global models.

This study has been carried out in cooperation with Unilever and partly with the EC founded project, Smartdoc IST-2000-28137. In areas of statistics and image processing, both the amount of data and the dimensions are increasing rapidly and an interactive visualization tool that lets the user perform real-time analysis can save valuable time. Real-time cropping and drill-down considerably facilitate the analysis process and yield more accurate decisions. In the Smartdoc project, there has been a request for a component used for smart filtering in multidimensional data sets. As the Smartdoc project aims to develop smart, interactive components to be used on low-end systems, the implementation of the self-organizing map algorithm proposes which dimensions to visualize. Together with Dr. Robert Treloar at Unilever, the SOM Visualizer - an application for interactive visualization and analysis of multidimensional data - has been developed. The analytical part of the application is based on Kohonen’s self-organizing map algorithm. In cooperation with the Smartdoc project, a component has been developed that is used for smart filtering in multidimensional data sets. Microsoft Visual Basic and components from the graphics library AVS OpenViz are used as development tools.

In this study, three different algorithms are implemented to solve thecapacitated vehicle routing problem with and without time windows:ant colony optimization, a genetic algorithm and a genetic algorithmwith self-organizing map. For the capacitated vehicle routing problemthe Augerat et al’s benchmark problems were used and for the capaci-tated vehicle routing problem with time windows the Solomon’sbenchmark problems. All three algorithms were tuned over thirtyinstances per problem with the tuners SPOT and ParamILS. The tuningresults from all instances were combined to the final parameter valuesand tested on a larger set of instances. The test results were used tocompare the algorithms and tuners against each other. The ant colonyoptimization algorithm outperformed the other algorithms on bothproblems when considering all instances. The genetic algorithm withself-organizing map found more best known solutions than any otheralgorithm when using parameters, on the capacitated vehicle routingproblem. The algorithms performed well and several new best knownresults were discovered for the capacitated vehicle routing problem andnew best solutions found by heuristics were discovered for the 100customer Solomon problems. When comparing the tuners they bothworked well and no clear winner emerged.

Le but de cette thèse est de documenter la variabilité des teneurs en aérosols minéraux en Afrique de l’Ouest et de comprendre les mécanismes qui la contrôlent. L’analyse des concentrations mesurées au sol au Sahel montre que ce sont les transports d’aérosols sahariens qui sont responsables des maxima de concentration observés en saison sèche Marticorena et al. (2010). Ces évènements présentent une très forte variabilité à l’échelle intra-saisonnière et interannuelle mais une certaine persistance à l’échelle régionale. Quelles sont les conditions météorologiques qui expliquent cette variabilité ? La réponse à cette question ne peut être obtenue qu’au travers d’une analyse systématique et conjointe des conditions météorologiques régionales et des mesures d’aérosol désertiques sur le continent (concentration de surface, épaisseurs optiques en aérosols) et au large de l’Afrique de l’Ouest. Les mesures d’épaisseur optique en aérosols (AOT) permettent la surveillance globale du contenu atmosphérique en particules. Cependant la relation permettant de retrouver les concentrations massiques en (PM10) à partir des mesures d’épaisseur optique n’est pas toujours simple. Ainsi dans un premier temps l’analyse de la variabilité de la concentration en PM10 et des conditions météorologiques locales mesurées au sol, a permis de confirmer un cycle saisonnier régional présenté par Marticorena et al. (2010), caractérisé par un maximum de concentration en saison sèche et un minimum pendant la saison des pluies sur l’ensemble de trois stations. L’analyse du cycle diurne des concentrations a permis de voir qu’il est diffèrent selon les saisons. En saison sèche, le jet de basses couches (NLLJ en anglais) saharien semble moduler le cycle diurne des concentrations en particules mesurées au Sahel, tandis que l’évolution diurne des concentrations de poussière en saison des pluies apparaît être modulé par l’évolution des systèmes convectifs a l’échelle régionale.Dans un deuxième temps, une méthodologie basée sur une classification en types de temps a été proposée. Elle permet de mettre en évidence des situations météorologiques typiques pour lesquelles la relation AOT-PM10 est simplifiée, puis une modélisation de PM10 à partir de l’épaisseur optique. Pour cela deux approches ont été adoptées. D’abord une famille composée de six régimes de temps bruts a été définie décrivant la saisonnalité climatique. La mise en relation AOT-PM10 permet de voir que les régimes de temps caractéristiques du flux d’harmattan présentent une relation meilleure en terme de corrélation par rapport à la relation avant la séparation en type de temps. Ensuite six régimes de temps ayant un impact sur la variabilité interannuelle des aérosols ont été définis. Les régimes de temps obtenus sont caractérisés par une forte variabilité interannuelle et moins persistante que pour les régimes de temps bruts. Les régimes de temps sont caractérisés par des anomalies de circulation atmosphérique, soit cyclonique, soit anticyclonique, d’échelle synoptique et centrée sur le nord du domaine. L’intégration de l’ensemble de cette approche en régimes de temps permet d’obtenir des résultats satisfaisants dans la prévision des valeurs de PM10, ouvrant des perspectives en terme de système d’alerte précoce. / The aim of this thesis is to document the variability of the atmospheric dust content in West Africa and to understand the mechanisms that control it. The analysis of the concentrations measured ground in the Sahel shows that it is the transport of Saharan aerosols that are responsible for the maximum concentration observed in the dry season. These events have a very high variability at intra-seasonal and interannual scale with some persistence on a regional scale. What are the weather conditions that explain this variability? The answer to this question can be obtained only through a systematic analysis of regional weather and desert aerosol measurements on the continent (surface concentration, aerosol optical thickness) and of the West Africa. The aerosol optical thickness measurements (AOT) allow global monitoring of atmospheric particulate content. However the relationship allowing to find the concentrations (PM10) from the optical thickness measurement is not always simple. So at first the analysis of the variability of the PM10 concentration and local meteorological conditions measured on the ground, confirmed a regional seasonal cycle presented by Marticorena et al. (2010), characterized by a maximum concentration during the dry season and a minimum during the rainy season on all three stations. The analysis of the diurnal cycle concentrations allowed seeing that it differ depending on the season. In the dry season, the Saharian low level jet (NLLJ) appears to modulate the diurnal cycle of particulate concentrations in the Sahel, while the diurnal evolution of concentrations of the rainy season appears to be modulated by changes in convective systems has regionally. Secondly, a methodology based on a classification into weather types was proposed. It allows to highlight typical meteorological situations in which the AOT- PM10 relationship is simplified and modeling PM10 from the Aerosol Optical Thickness. For this, two approaches have been adopted. At first a family of six weather types has been defined describing the climate seasonality. The link between AOT and PM10 can allowed to see that the weather types characterise to harmattan flow have a better relationship in terms of correlation to the relationship before separating in weather type. Then six weather types affecting the interannual variability of aerosols were defined. The resulting weather patterns are characterized by high interannual variability and less persistent than for rough weather regimes. The weather patterns are characterized by atmospheric circulation anomalies, either cyclonic or anticyclonic at synoptic scale and centered on the northern area. The integration of all of this approach in weather types provides satisfactory results in forecasting PM10 values, opening perspectives in terms of early warning system.

Abstract A future vision of enabling technologies for the needs of energy conservation as well as energy efficiency based on the most important megatrends identified, namely climate change, urbanization, and digitalization. In the United States and in the European Union, about 40% of total energy consumption goes into energy use by buildings. Moreover, indoor climate quality is recognized as a distinct health hazard. On account of these two factors, energy efficiency and healthy housing are active topics in international research. The main aims of this thesis are to study which elements affect indoor climate quality, how energy consumption describes building energy efficiency and to analyse the measured data using intelligent computational methods. The data acquisition technology used in the studies relies heavily on smart metering technologies based on Building Automation Systems (BAS), big data and the Internet of Things (IoT). The data refining process presented and used is called Knowledge Discovery in Databases (KDD). It contains methods for data acquisition, pre-processing, data mining, visualisation and interpretation of results, and transformation into knowledge and new information for end users. In this thesis, four examples of data analysis and knowledge deployment concerning small houses and school buildings are presented. The results of the case studies show that the data mining methods used in building energy efficiency and indoor climate quality analysis have a great potential for processing a large amount of multivariate data effectively. An innovative use of computational methods provides a good basis for researching and developing new information services. In the KDD process, researchers should co-operate with end users, such as building management and maintenance personnel as well as residents, to achieve better analysis results, easier interpretation and correct conclusions for exploiting the knowledge. / Tiivistelmä Tulevaisuuden visio energiansäästön sekä energiatehokkuuden mahdollistavista teknologioista pohjautuu tärkeimpiin tunnistettuihin megatrendeihin, ilmastonmuutokseen, kaupungistumiseen ja digitalisoitumiseen. Yhdysvalloissa ja Euroopan unionissa käytetään noin 40 % kokonaisenergiankulutuksesta rakennusten käytön energiatarpeeseen. Myös rakennusten sisäilmaston on havaittu olevan ilmeinen terveysriski. Perustuen kahteen edellä mainittuun tekijään, energiatehokkuus ja asumisterveys ovat aktiivisia tutkimusaiheita kansainvälisessä tutkimuksessa. Tämän väitöskirjan päätavoitteena on ollut tutkia, mitkä elementit vaikuttavat sisäilmastoon ja rakennusten energiatehokkuuteen pääasiassa analysoimalla mittausdataa käyttäen älykkäitä laskennallisia menetelmiä. Tutkimuksissa käytetyt tiedonkeruuteknologiat perustuvat etäluentaan ja rakennusautomaatioon, big datan hyödyntämiseen ja esineiden internetiin (IoT). Väitöskirjassa esiteltävä tietämyksen muodostusprosessi (KDD) koostuu tiedonkeruusta,datan esikäsittelystä, tiedonlouhinnasta, visualisoinnista ja tutkimustulosten tulkinnasta sekä tietämyksen muodostamisesta ja oleellisen informaation esittämisestä loppukäyttäjille. Tässä väitöstutkimuksessa esitellään neljän data-analyysin ja niiden pohjalta muodostetun tietämyksen hyödyntämisen esimerkkiä, jotka liittyvät pientaloihin ja koulurakennuksiin. Esimerkkitapausten tulokset osoittavat, että käytetyillä tiedonlouhinnan menetelmillä sovellettuna rakennusten energiatehokkuus- ja sisäilmastoanalyyseihin on mahdollista jalostaa suuria monimuuttuja-aineistoja tehokkaasti. Laskennallisten menetelmien innovatiivinen käyttö antaa hyvät perusteet tutkia ja kehittää uusia informaatiopalveluja. Tutkijoiden tulee tehdä yhteistyötä loppukäyttäjinä toimivien kiinteistöhallinnan ja -ylläpidon henkilöstön sekä asukkaiden kanssa saavuttaakseen parempia analyysituloksia, helpompaa tulosten tulkintaa ja oikeita johtopäätöksiä tietämyksen hyödyntämiseksi.

Abstract The effect of alcohol intake on the pathophysiology of atherosclerotic cardiovascular disease is controversial, especially with respect to heavy alcohol intake. The pathobiology behind atherosclerosis is a complex and multiparametric phenomenon, therefore a self-organizing map (SOM), an unsupervised learning based artificial neural network technique, was applied in the present work. This study was carried out to investigate the effect of heavy alcohol intake on the pathophysiology of atherosclerosis, including several lipoproteins and adiponectin, an adipocyte-derived cytokine that may ameliorate atherosclerosis. Firstly, the effect of heavy alcohol intake on the capacity of HDL and its subclasses (HDL2 and HDL3) to mediate cholesterol efflux from macrophages was studied. Secondly, data of ultracentrifugally isolated lipoproteins were fed into SOM analysis to investigate whether this method can find diverse lipoprotein phenotypes from the heterogeneous lipoprotein data. Thirdly, the aforementioned method was applied to multivariate data of alcohol drinkers to study whether distinct metabolic profiles are associated to heavy alcohol consumption. The results revealed that HDL2, not HDL3, of heavy alcohol drinkers had an enhanced capacity to remove cholesterol from macrophages when compared with control persons. SOM analysis enhanced the ultracentrifugally based lipoprotein data and depicted several novel lipoprotein phenotypes. In addition, lipoprotein-based SOM analysis found two distinct metabolic profiles in heavy alcohol drinkers: an anti-atherogenic and a metabolic syndrome-like profile with opposite metabolic features, such as characteristics of lipoproteins, plasma concentration of adiponectin and prevalence of metabolic syndrome. These profiles also tended to differ in their CV risk. In conclusion, the enhanced cholesterol efflux capacity of HDL2 in heavy drinkers is an anti-atherogenic change linked to alcohol drinking. However, clinically it may be important to be aware that although heavy alcohol drinkers have a low LDL-C level, they differ in their other lipoprotein measures, forming distinct phenotypes with potentially different CV risks. Finally, SOM analysis of ultracentrifugally based lipoprotein data generates in silico classification of lipoprotein particles and thereby offers a new tool for lipoprotein research. / Tiivistelmä Alkoholinkäytön vaikutus ateroskleroottisen sydän- ja verisuonitaudin patofysiologiaan on kiistanalainen, etenkin runsaan alkoholinkäytön kohdalla. Koska patobiologia ateroskleroosin taustalla on monimutkainen ilmiö, tässä työssä sovellettiin menetelmänä itseorganisoituvaa karttaa, joka on ohjaamattomaan oppimiseen perustuva neuroverkkomalli. Tutkimuksen tavoitteena oli selvittää runsaan alkoholinkäytön vaikutusta ateroskleroosin patofysiologisiin merkkiaineisiin, mukaan lukien useita lipoproteiineja sekä adiponektiini, rasvasoluperäinen sytokiini, joka voi lievittää ateroskleroosia. Ensimmäisessä osatyössä tutkittiin runsaan alkoholinkäytön vaikutusta HDL:n ja sen alafraktioiden (HDL2 ja HDL3) kykyyn poistaa kolesterolia makrofageista. Toisessa osatyössä ultrasentrifuugaukseen perustuva lipoproteiiniaineisto syötettiin itseorganisoituvaan karttaan. Työssä selvitettiin löytäisikö menetelmä erilaisia lipoproteiinifenotyyppejä heterogeenisestä aineistosta. Kolmannessa osatyössä em. menetelmää sovellettiin monimuuttuja-aineistoon, joka koostui runsaasti alkoholia käyttävistä ja verrokeista. Tutkittiin, liittyykö runsaaseen alkoholinkäyttöön erilaisia metabolisia profiileja. Tulokset osoittivat, että suurkuluttajien HDL2-hiukkasen kolesterolinpoistokyky makrofageista oli suurempi kuin verrokeilla. Itseorganisoituvaan karttaan perustuva lipoproteiinien luokittelumenetelmä löysi useita uusia lipoproteiinifenotyyppejä. Lisäksi, em. menetelmä löysi suurkuluttajilta kaksi erilaista metabolista profiilia: anti-aterogeeninen ja metabolisen syndrooman kaltainen. Näillä oli vastakkaiset metaboliset piirteet, kuten lipoproteiinien ominaisuudet, adiponektiinin pitoisuus plasmassa ja metabolisen syndrooman esiintyvyys. Profiileihin liittyi mahdollisesti myös erilainen sydän- ja verisuonitautiriski. Tutkimus osoittaa, että alkoholin suurkuluttajilla havaittu parempi HDL2:n kyky poistaa kolesterolia soluista on anti-aterogeeninen muutos, joka liittyy alkoholin käyttöön. Kliinisesti voi olla merkittävää, että vaikka alkoholin suurkuluttajilla oli pieni LDL-C pitoisuus, he jakaantuivat muiden lipoproteiiniperäisten muuttujien perusteella kahteen eri fenotyyppiryhmään, joihin liittyi erilainen sydäntautiriski. Lisäksi itseorganisoituva kartta loi ultrasentrifugoinnilla eristetyille lipoproteiineille in silico -luokittelun, joten se tarjoaa uuden työkalun lipoproteiinitutkimukseen.

Abstract The amount of information available has increased due to the development of our modern digital society. This has caused an information overflow, meaning that there is lot of data available but the meaningful information or knowledge is hidden inside the overwhelming data smog. Nevertheless, the large amount of data together with the increased capabilities of computers provides a great opportunity to learn the behaviour of different kinds of phenomena at a more detailed level. The quality of life, well-being and a healthy living environment, for example, are fields where new information services can assist the creation of proactive decisions to avoid environmental problems caused by industrial activity, traffic, or extraordinary weather conditions. The combination of data coming from different sources such as public registers, companies’ operational information systems, online sensors and process monitoring systems provides a fruitful basis for creating new valuable information for citizens, decision makers or other end users. The aim of this thesis is to present the concept of intelligent information services and a methodological background in order to add intelligence using computational methods for the enrichment of multidimensional data. Moreover, novel examples are presented where new significant information is created and then provided for end users. The data refining process used is called data mining and contains methods for data collection, pre-processing, modelling, visualizing and interpreting the results and sharing the new information thus created. Information systems are a base for the creation of information services, meaning that stakeholder groups have access only to information but they do not own the whole information system that contains measurement systems, data collecting, and a technological platform. Intelligence in information services comes from the use of computational intelligent methods in data processing, modelling and visualization. In this thesis the general concept of such services is presented and concretized using five cases that focus on environmental and industrial examples. The results of these case studies show that the combination of different data sources provides fertile ground for developing new information services. The data mining methods used such as clustering and predictive modelling together with effective pre-processing methods have great potential to handle the large amount of multivariate data in this environmental context also. A self-organizing map combined with k-means clustering is useful for creating more detailed information about personal energy use. Predictive modelling using a multilayer perceptron (MLP) is well suited for estimating the number of tourists visiting a leisure centre and to find the correspondence between pulp process characteristics and the chemicals used. These results have many indirect effects on reducing negative concerns regarding our surroundings and maintaining a healthy living environment. The innovative use of stored data is one of the main elements in the creation of future information services. Thus, more emphasis should be placed on the development of data integration and effective data processing methods. Furthermore, it is noted that final end users, such as citizens or decision makers, should be involved in the data refining process at the very first stage. In this way, the approach is truly customer-oriented and the results fulfil the concrete need of specific end users. / Tiivistelmä Informaation määrä on kasvanut merkittävästi tietoyhteiskunnan kehittymisen myötä. Käytössämme onkin huomattava määrä erimuotoista tietoa, josta voimme hyödyntää kuitenkin vain osan. Jatkuvasti mitattavan datan suuri määrä ja sijoittuminen hajalleen asettavat osaltaan haasteita tiedon hyödyntämiselle. Tietoyhteiskunnassa hyvinvointi ja terveellisen elinympäristön säilyminen koetaan aiempaa tärkeämmäksi. Toisaalta yritysten toiminnan tehostaminen ja kestävän kehityksen edistäminen vaativat jatkuvaa parantamista. Informaatioteknologian avulla moniulotteista mittaus- ja rekisteritietoa voidaan hyödyntää esimerkiksi ennakoivaan päätöksentekoon jolla voidaan edistää edellä mainittuja tavoitteita. Tässä työssä on esitetty ympäristöalan älykkäiden informaatiopalveluiden konsepti, jossa oleellista on loppukäyttäjien tarpeiden tunnistaminen ja ongelmien ratkaiseminen jalostetun informaation avulla. Älykkäiden informaatiopalvelujen taustalla on yhtenäinen tiedonlouhintaan perustuva tiedonjalostusprosessi, jossa raakatieto jalostetaan loppukäyttäjille soveltuvaan muotoon. Tiedonjalostusprosessi koostuu datan keräämisestä ja esikäsittelystä, mallintamisesta, tiedon visualisoinnista, tulosten tulkitsemisesta sekä oleellisen tiedon jakamisesta loppukäyttäjäryhmille. Datan käsittelyyn ja analysointiin on käytetty laskennallisesti älykkäitä menetelmiä, josta juontuu työn otsikko; älykkäät informaatiopalvelut. Väitöskirja pohjautuu viiteen artikkeliin, joissa osoitetaan tiedonjalostusprosessin toimivuus erilaisissa tapauksissa ja esitetään esimerkkejä kuhunkin prosessin vaiheeseen soveltuvista laskennallisista menetelmistä. Artikkeleissa on kuvattu matkailualueen kävijämäärien ennakointiin ja kotitalouksien sähköenergian kulutuksen pienentämiseen liittyvät informaatiopalvelut sekä analyysi selluprosessissa käytettävien kemikaalien määrän pienentämiseksi. Näistä saadut kokemukset ja tulokset on yleistetty älykkään informaatiopalvelun konseptiksi. Väitöskirjan toisena tavoitteena on rohkaista organisaatioita hyödyntämään tietovarantoja aiempaa tehokkaammin ja monipuolisemmin sekä rohkaista tarkastelemaan myös oman organisaation ulkopuolelta saatavien tietolähteiden käyttämistä. Toisaalta, uudenlaisten informaatiopalvelujen ja liiketoimintojen kehittämistä tukisi julkisilla varoilla kerättyjen, ja osin yritysten hallussa olevien, tietovarantojen julkaiseminen avoimiksi.

For scientific or socio-economic reasons it is often necessary or desirable that biological material be identified. Given that there are an estimated 10 million living organisms on Earth, the identification of biological material can be problematic. Consequently the services of taxonomist specialists are often required. However, if such expertise is not readily available it is necessary to attempt an identification using an alternative method. Some of these alternative methods are unsatisfactory or can lead to a wrong identification. One of the most common problems encountered when identifying specimens is that important diagnostic features are often not easily observed, or may even be completely absent. A number of techniques can be used to try to overcome this problem, one of which, the Self Organizing Map (or SOM), is a particularly appealing technique because of its ability to handle missing data. This thesis explores the use of SOMs as a technique for the identification of indigenous trees of the Acacia species in KwaZulu-Natal, South Africa. The ability of the SOM technique to perform exploratory data analysis through data clustering is utilized and assessed, as is its usefulness for visualizing the results of the analysis of numerical, multivariate botanical data sets. The SOM’s ability to investigate, discover and interpret relationships within these data sets is examined, and the technique’s ability to identify tree species successfully is tested. These data sets are also tested using the C5 and CN2 classification techniques. Results from both these techniques are compared with the results obtained by using a SOM commercial package. These results indicate that the application of the SOM to the problem of biological identification could provide the start of the long-awaited breakthrough in computerized identification that biologists have eagerly been seeking. / Dissertation (MSc)--University of Pretoria, 2011. / Computer Science / unrestricted