Theses deals with network traffic modeling focused on elaboration by time series analysis. The nature of network traffic is discussed above all http traffic. First three chapters are theoretical, which describes time series and basic models, linear AR, MA, ARMA, ARIMA and nonlinear ARCH. Other chapters define terms like self-similarity and long range dependence. It is demonstrated a failure of conventional models which cannot capture these specific properties of network data traffic. On the basis of study in chapter 6. is closely described the combined ARIMA/GARCH model and its parameter estimation procedure. Applied part of this theses deals with procedure of estimation and fitting the estimation model to observed network traffic. After an estimation a few future values are predicted on the basis of estimated model. These predicted values are consequently compared with real data.

The use of Normalized Difference Vegetation Index (NDVI) time series over the last decades has increased our understanding of vegetation change dynamics from global to regional scale through quantitative analysis of inter-annual trends in NDVI and climatological parameters (rainfall and temperature). Change in land cover induced by human activities such as livestock grazing and deforestation for large-scale farming (subsistence and mechanized) has influenced the ecological pattern of the Guinea savannah region (GSR) of Nigeria, thereby resulting in loss of biodiversity and changes in vegetation cover. In the context of the GSR of Nigeria where agriculture still plays a major role in people’s economy, it is important to identify the relationship between climatic variables, vegetation productivity and human activities which can be used to understand the on-going transition processes. This study, therefore, examines the spatial and temporal relationship between NDVI and climate parameters, land use land cover change (LULCC) and the perspective of local people on vegetation change dynamics in the study region. In order to do this, bi-monthly NDVI3g time series datasets from Global Inventory Modeling and Mapping Studies (GIMMS), monthly rainfall datasets from Tropical Applications of Meteorology Satellite (TAMSAT), monthly temperature datasets from Climate Research Unit (CRU), national land use land cover (LULC) data of Nigeria from Forestry Management Evaluation & Coordination Unit (FORMECU), global land cover datasets from European Space Agency, Landsat imagery and socio-economic field data collection were used in order to understand vegetation change dynamics across the Guinea savannah regions of Nigeria. Time series analysis (TSA) was applied to both NDVI and climate data used in order to examine the temporal dynamics of vegetation cover change and to detect NDVI-climate relationship during the period from 1983 through 2011. Both parametric and non-parametric statistical models were employed for the assessment of long-term inter-annual trend on the decomposed time series datasets for the whole region (Guinea savannah region) and selected locations. In addition to the TSA, harmonic regression analysis was performed on NDVI and rainfall datasets in order to examine change in seasonality and phyto-phenological characteristics of vegetation. Detection of change in land use and land cover was done by extracting information from existing land cover datasets (ancillary datasets). CLASlite was used for the assessment of the extent of deforestation, while linkage between remotely sensed data and social science was carried out via field surveys based on questionnaires in order to understand the drivers of vegetation change. The study reveals that about 90 % of the Guinea savannah region show positive NDVI trends which indicate greening over time, while about 10 % of the region shows negative trends. This greening trends are closely related to regions where intensive agriculture is being practiced (also along inland valleys) while regions with negative trends show significant loss in woodlands (forest and shrublands) as well as herbaceous vegetation cover due to over-grazing by agro-pastoralism. The result confirms that there is a good relationship (statistically significant positive correlation) between rainfall and NDVI both on intra-annual and inter annual time scale for some selected locations in the study region (> 65 %), while negative statistical correlation exists between NDVI and temperature in the selected locations. This implies that vegetation growth (productivity) in the region is highly dependent on rainfall. The result of the harmonic regression analysis reveals a shift in the seasonal NDVI pattern, indicating an earlier start and a more prolonged growing season in 2011 than in 1983. This study proves significant change in LULC with evidence of an increase in the spatial extent of agricultural land (+ 30 %) and loss of woodlands (- 55 %) between 2000 and 2009 for Kogi State. The results of the socio-economic analysis (people’s perception) highlight that vegetation change dynamics in the study region are the resultant effects of increased anthropogenic activities rather than climatic variability. This study couples data from remote sensing and ground survey (socio-economics) for a better understanding of greening trend phenomena across the Guinea savannah region of Nigeria, thus filling the gap of inadequate information on environmental condition and human perturbation which is essential for proper land use management and vegetation monitoring.

Computational ECG (electrocardiogram) analysis enables accurate and faster diagnosis and early prediction of heart failure related symptoms (heart failure worsening). Machine learning, particularly deep learning, has been applied for ECG data successfully. The previous applications, however, either mainly focused on classifying occurrent, known patterns of on-going heart failure or heart failure related diseases such arrhythmia, which have undesirable predictability beforehand, or emphasizing on data from pre-processed public database data. In this dissertation, we developed an approach, however, does not fully capitalize on the potential of deep learning, which directly learns important features from raw input data without relying on a priori knowledge. Here, we present a deep transfer learning pipeline which combines an image-based pretrained deep neural network model with manifold learning to predict the precursors of heart failure (heart failure-worsening and recurrent heart failure related re-hospitalization) using raw ECG time series from wearable devices. In this dissertation, we used the unprocessed real-life ECG data from the SENTINEL-HF study by Dovancescu, et al. to predict the precursors of heart failure worsening. To extract rich features from ECG time series, we took a deep transfer learning approach where 1D time-series of five heartbeats were transformed to 2D images by Gramian Angular Summation Field (GASF) and then the pretrained models, VGG19 were used for feature extraction. Then, we applied UMAP (Uniform Manifold Approximation and Projection) to capture the manifold of the standardized feature space and reduce the dimension, followed by SVM (Support Vector Machine) training. Using our pipeline, we demonstrated that our classifier was able to predict heart failure worsening with 92.1% accuracy, 92.9% precision, 92.6% recall and F1 score of 0.93 bypassing the detection of known abnormal ECG patterns. In conclusion, we demonstrate the feasibility of early alerts of heart failure by predicting the precursor of heart failure worsening based on raw ECG signals. We expected that our approached provided an innovative method to assess the recovery and successfulness for the treatment patient received during the first hospitalization, to predict whether recurrent heart failure is likely to occur, and to evaluate whether the patient should be discharged.

Cell-to-cell variations and intracellular processes such as cytoskeletal organization and organelle dynamics exhibit massive heterogeneity. Advances in imaging and optics have enabled researchers to access spatiotemporal information in living cells efficiently. Even though current imaging technologies allow us to acquire an unprecedented amount of cell images, it is challenging to extract valuable information from the massive and complex dataset to interpret heterogeneous biological processes. Machine learning (ML), referring to a set of computational tools to acquire knowledge from data, provides promising solutions to meet this challenge. In this dissertation, we developed ML pipelines for deconvolution of subcellular protrusion heterogeneity from live cell imaging and molecular diagnostic from lens-free digital in-line holography (LDIH) imaging. Cell protrusion is driven by spatiotemporally fluctuating actin assembly processes and is morphodynamically heterogeneous at the subcellular level. Elucidating the underlying molecular dynamics associated with subcellular protrusion heterogeneity is crucial to understanding the biology of cellular movement. Traditional ensemble averaging methods without characterizing the heterogeneity could mask important activities. Therefore, we established an ACF (auto-correlation function) based time series clustering pipeline called HACKS (deconvolution of heterogeneous activities in coordination of cytoskeleton at the subcellular level) to identify distinct subcellular lamellipodial protrusion phenotypes with their underlying actin regulator dynamics from live cell imaging. Using our method, we discover “accelerating protrusion”, which is driven by the temporally ordered coordination of Arp2/3 and VASP activities. Furthermore, deriving the merits of ML, especially Deep Learning (DL) to learn features automatically, we advanced our pipeline to learn fine-grained temporal features by integrating the prior ML analysis results with bi-LSTM (bi-direction long-short term memory) autoencoders to dissect variable-length time series protrusion heterogeneity. By applying it to subcellular protrusion dynamics in pharmacologically and metabolically perturbed epithelial cells, we discovered fine differential response of protrusion dynamics specific to each perturbation. This provides an analytical framework for detailed and quantitative understanding of molecular mechanisms hidden in their heterogeneity. Lens-free digital in-line holography (LDIH) is a promising microscopic tool that overcomes several drawbacks (e.g., limited field of view) of traditional lens-based microscopy. Numerical reconstruction for hologram images from large-field-of-view LDIH is extremely time-consuming. Until now, there are no effective manual-design features to interpret the lateral and depth information from complex diffraction patterns in hologram images directly, which limits LDIH utility for point-of-care applications. Inherited from advantages of DL to learn generalized features automatically, we proposed a deep transfer learning (DTL)-based approach to process LDIH images without reconstruction in the context of cellular analysis. Specifically, using the raw holograms as input, the features extracted from a well-trained network were able to classify cell categories according to the number of cell-bounded microbeads, which performance was comparable with that of object images as input. Combined with the developed DTL approach, LDIH could be realized as a low-cost, portable tool for point-of-care diagnostics. In summary, this dissertation demonstrate that ML applied to cell imaging can successfully dissect subcellular heterogeneity and perform cell-based diagnosis. We expect that our study will be able to make significant contributions to data-driven cell biological research.

Forecasting demand is one of the processes which greatly influences the decision making within a company, and it is also one of the greatest sources of uncertainty. Inaccurate forecasts force companies to find ways to compensate for the uncertainty, often by building inventories. On the other hand, accurate forecasts help companies to achieve better customer service and lower inventory levels. Cytiva is a global life science leader who manufactures high-technology laboratory instruments at their site in Umeå. To make sure that their transportation and storage spaces are sufficient, their down-stream suppliers require information about the quantity of the final products beforehand. But as of today, the company has inconsistent outbound inventory volumes. Thus, there is a great demand for increased visibility and predictability at the Umeå site’s outbound logistics. Further, Cytiva in Umeå bases their forecasting on manually calculated estimations which is both inefficient and can create errors due to human factors. These intrinsic information inconsistencies related to the outbound logistics is prone to creating bottlenecks in their overall supply chain. The main goal of this project is to increase the accuracy of these forecasts by developing a model. The outcome will be better estimations and clearer connection between the site in Umeå and the 3PL in Rosersberg. Additionaly, a good model makes the supply chain more efficient by creating better preconditions for managing the transportation and inventory at the receiving 3PL. To make forecasts for Cytiva’s outbound inventory, we chose to focus on two of the most common families of univariate time series models, namely the ARIMA and the Exponential Smoothing family. Based on these two families we have implemented, evaluated and compared six forecasting models. Initially, the modeling was done using daily observations in order to examine whether the models could improve the company’s demand for short forecast horizons. However, except for modeling on daily observations, we also widened the time interval by merging the observations into weeks to extend the modeling perspective even further. The results showed that the use of models can noticeably improve the estimations of the inventory and transportation spaces. We conclude that, among our models, the Holt-Winters using additive seasonality is the most optimal model when the forecasts are made on a daily time frequency, while the SARIMA model performs better on the weekly data.

In his article, "Cycles of Deviance" (1996), Hawdon demonstrates how varying rates of social mobility correspond to cyclical patterns of drug use in the United States between 1880 and 1990. He proposes that social mobility alters the "deviance structure" of a society by changing the rate at which certain behaviors are labeled deviant, and thus, the rate at which people engage in those behaviors. This study provides an updated assessment of the cycles of deviance model to determine whether it can account for rates of violent and economic crime. I use social mobility to predict homicide, burglary, and overall rates of drug use from 1970 through 2010 using a time-series analysis. Crime data are obtained from the FBI\'s Uniform Crime Reports and Monitoring the Future. Social mobility data are obtained from the Bureau of Labor Statistics, the Bureau of Justice Statistics, and the U.S. Census Bureau. I also control for several well- established correlates of crime -- namely, economic and demographic factors, police size, illicit drug market activity, and firearm availability. Results show moderate support for the cycles of deviance model in predicting rates of homicide and burglary. However, social mobility\'s influence with respect to drug use appears to vary with the size of the youth population. / Master of Science

How can fluctuations in one-dimensional time series data be characterized and how can detected effects be decomposed into their dynamical origins or causes? In the context of these questions, a variety of problems are discussed and solutions are introduced. The first issue concerns the causes of anomalous diffusion. A previously proposed framework decomposes the Hurst exponent into the Joseph, Noah, and Moses effects. They represent violations of the three premises of the central limit theorem. Here the framework is applied to an intermittent deterministic system, which exhibits a rich combination of all three effects. Nevertheless, the results provide an intuitive interpretation of the dynamics. In addition, the framework is theoretically discussed and connected to a calculation that proves its validity for a large class of systems. Once the type of anomalous statistical behavior is classified, one might ask what the dynamical origin of the effects is. Especially the property of long range temporal correlations (the Joseph effect) is discussed in detail. In measurements, they might arise from different dynamical origins or can be explained as an emerging phenomenon. A collection of different routes to the observed behavior is established here. A popular tool for detecting long range correlations is detrended fluctuation analysis. Its advantages over traditional methods are stability and smoothness for timescales up to one fourth of the measurement time and the ability to neglect the slow dynamics and trends. Recently, a theory for an analytical understanding of this method was introduced. In this thesis, the method is further analyzed and developed. An approach is presented that enables scientists to use this method for short range correlated data, even if the dynamics is very complex. Fluctuations can be decomposed into a superposition of linear models that explain its features. Therefore, on the one hand, this thesis is about understanding the effects of anomalous diffusion. On the other hand, it is about widening the applicability of one of its detection methods such that it becomes useful for understanding normal or complex statistical behavior. A good example of a complex system, where the proposed stochastic methods are useful, is the atmosphere. Here it is shown how detrended fluctuation analysis can be used to uncover oscillatory modes and determine their periods. One of them is the El Ni\~no southern oscillation. A less well known and more challenging application is a 7--8 year mode in European temperature fluctuations. A power grid is a very different type of complex system. However, using the new method, it is possible to generate a data model that incorporates the important features of the grid frequency.

Includes bibliography. / Ocean waves represent an important design factor in many coastal engineering applications. Although extreme wave height is usually considered the single most important of these factors there are other important aspects that require consideration. These include the probability distribution of wave heights, the seasonal variation and the persistence, or duration, of calm and storm periods. If one is primarily interested in extreme wave height then it is possible to restrict one's attention to events which are sufficiently separated in time to be effectively independently (and possibly even identically) distributed. However the independence assumption is not tenable for the description of many other aspects of wave height behaviour, such as the persistence of calm periods. For this one has to take account of the serial correlation structure of observed wave heights, the seasonal behaviour of the important statistics, such as mean and standard deviation, and in fact the entire seasonal probability distribution of wave heights. In other words the observations have to be regarded as a time series.

Thesis (M.S.)--Massachusetts Institute of Technology, Dept. of Meteorology and Physical Oceanography, 1982. / Microfiche copy available in Archives and Science. / Bibliography: leaves 75-77. / by John Roberts Anderson. / M.S.

Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering, 1981. / MICROFICHE COPY AVAILABLE IN ARCHIVES AND ENGINEERING. / Bibliography: leaves 74-76. / by Stephen W. Lang. / Ph.D.