The mobile traffic in cellular networks is increasing in a steady rate as we go intoa future where we are connected to the internet practically all the time in one wayor another. To map the mobile traffic and the volume pressure on the base stationduring different time periods, it is useful to have the ability to predict the trafficvolumes within cellular networks. The data in this work consists of 4G cellular trafficdata spanning over a 7 day coherent period, collected from cells in a moderately largecity. The proposed method in this work is ARIMA modeling, in both original formand with an extension where the coefficients of the ARIMA model are re-esimated byintroducing some user characteristic variables. The re-estimated coefficients produceslightly lower forecast errors in general than a isolated ARIMA model where thevolume forecasts only depends on time. This implies that the forecasts can besomewhat improved when we allow the influence of these variables to be a part ofthe model, and not only the time series itself.

In recent years, many research works such as Tiao and Tsay (1994), Stock and Watson (1999), Chen et al. (2001), Clements and Jeremy (2001), Marcellino (2002), Laurini and Vieira (2005) and others have described the dynamic features of many macroeconomic variables as nonlinear. Using the approach of Keenan (1985) and Tsay (1989) this study shown that Ghana inflation rates from January 1980 to December 2009 follow a threshold nonlinear process.  In order to take into account the nonlinearity in the inflation rates we then apply a two regime nonlinear SETAR model to the inflation rates and then study both in-sample and out-of-sample forecast performance of this model by comparing it with the linear SARIMA model. Based on the in-sample forecast assessment from the linear SARIMA and the nonlinear SETAR models, the forecast measure MAE and RMSE suggest that the nonlinear SETAR model outperform the linear SARIMA model. Also using multi-step-ahead forecast method we predicted and compared the out-of-sample forecast of the linear SARIMA and the nonlinear SETAR models over the forecast horizon of 12 months during the period of 2010:1 to 2010:12. From the results as suggested by MAE and RMSE, the forecast performance of the nonlinear SETAR models is superior to that of the linear SARIMA model in forecasting Ghana inflation rates. Thought the nonlinear SETAR model is superior to the SARIMA model according to MAE and RMSE measure but using Diebold-Mariano test, we found no significant difference in their forecast accuracy for both in-sample and out-of-sample.

Submitted by Marina Gruc Sztamfater (marinagruc@gmail.com) on 2015-02-27T18:33:07Z No. of bitstreams: 1 Tese_Marina_Gruc_Sztamfater.pdf: 909406 bytes, checksum: 53f3c92413eca5e92e2430768825432c (MD5) / Approved for entry into archive by Renata de Souza Nascimento (renata.souza@fgv.br) on 2015-02-27T18:34:48Z (GMT) No. of bitstreams: 1 Tese_Marina_Gruc_Sztamfater.pdf: 909406 bytes, checksum: 53f3c92413eca5e92e2430768825432c (MD5) / Made available in DSpace on 2015-02-27T19:27:56Z (GMT). No. of bitstreams: 1 Tese_Marina_Gruc_Sztamfater.pdf: 909406 bytes, checksum: 53f3c92413eca5e92e2430768825432c (MD5) Previous issue date: 2015-02-03 / This study aims to compare the forecasting efficiency of two different methodologies applied to the Brazilian shipments of corrugated board data. First the corrugated shipping data will be broken down by industrial categories of destination and for each category will be made univariate SARIMA models. The estimates of disaggregated series are then aggregated to form the prediction of the total shipment of corrugated board. The prediction made from the aggregation of industry categories will be compared with a univariate SARIMA aggregate model, in order to ascertain which of the two methods results in a model with better accuracy. This comparison will be made based on the methodology developed by Diebold and Mariano / O presente trabalho visa comparar o poder preditivo das previsões feitas a partir de diferentes metodologias aplicadas para a série de expedição brasileira de papelão ondulado. Os dados de expedição de papelão ondulado serão decompostos pelas categorias industriais de destino das caixas e serão feitos modelos do tipo SARIMA univariados para cada setor. As previsões das séries desagregadas serão então agregadas, para compor a previsão da série total de expedição. A previsão feita a partir da somatória das categorias industriais será comparada com um SARIMA univariado da série agregada, a fim de verificar qual dos dois métodos resulta em um modelo com melhor acurácia. Essa comparação será feita a partir da metodologia desenvolvida por Diebold e Mariano (1995).

This thesis deals about using models of neural networks like alternative of time series model based on Box-Jenkins methodology. The work is divided into two parts according to the model construction method. Each of the parts contains a theory that explains the individual processes and the progress of the model construction. This is followed by two experiments demonstrating the difference in approach to the design of a given model and creating a forecast by estimated values. for the following year. The last part expertly evaluates the quality of the predictions and considers the use of neural networks against prediction models as an alternative to Box-Jenkins methodology based models

This work concerns the creation of automatized procedure of ARIMA and SARIMA class model choice according to Box-Jenkins methodology and in this connection, also deals with force testing of unit roots and analysis of applying of informatics criteria when choosing a model. The goal of this work is to create an application in the environment R that can automatically choose a model of time array generating process. The procedure is verified by a simulation study. In this work an effect of values of generating ARMA (1,1) model processes parameters is examined, for his choice and power of KPSS test, augmented Dickey-Fuller and Phillips-Peron test of unit roots.

This study aims to investigate short-to-medium forecasting and simulation of streamflow, water levels, and sediments in Canada using Seasonal Autoregressive Integrated Moving Average (SARIMA) time series models. The methodology can account for linear trends in the time series that may result from climate and environmental changes. A Universal Canadian forecast Application using python web interface was developed to generate short-term forecasts using SARIMA. The Akaike information criteria was used as performance criteria for generating efficient SARIMA models. The developed models were validated by analyzing the residuals. Several stations from the Canadian Hydrometric Database (HYDAT) displaying a linear upward or downward trend were identified to validate the methodology. Trends were detected using the Man-Kendall test. The Nash-Sutcliffe efficiency coefficients (Nash ad Sutcliffe, 1970) of the developed models indicate that they are acceptable. The models can be used for short term (1 to 7 days) and medium-term (7 days to six months) forecasting of streamflow, water levels and sediments at all Canadian hydrometric stations. Such a forecast can be used for water resources management and help mitigate the effects of floods and droughts. The models can also be used to generate long time-series that can be used to test the performance of water resources systems. Finally, we have automated the process of analysis, model-building and forecasting streamflow, water levels, and sediments by building a python-based application easily extendable and user-friendly. Therefore, automating the SARIMA calibration and forecasting process for all Canadian stations for the HYDAT database will prove to be a very useful tool for decision-makers and other entities in the field of hydrological study.

In this paper we conduct an out-of-sample forecasting exercise for monthly Swedish air traveler volumes. The models considered are multiplicative seasonal ARIMA, Neural network autoregression, Exponential smoothing, the Prophet model and a Random Walk as a benchmark model. We divide the out-of-sample data into three different evaluation periods: Pre-COVID-19, during COVID-19 and Post-COVID-19 for which we calculate the MAE, MAPE and RMSE for each model in each of these evaluation periods. The results show that for the Pre-COVID-19 period all models produce accurate forecasts, in comparison to the Random Walk model. For the period during COVID-19, no model outperforms the Random Walk, with only Exponential smoothing performing as well as the Random Walk. For the period Post-COVID-19, the best performing models are Random Walk, SARIMA and Exponential smoothing, with all aforementioned models having similar performance.

Time series is a collection of points gathered at regular intervals. Time series analysis explores the time correlations and tries to model it according to trend and seasonality. One of the most relevant tasks, in time series analysis, is forecasting future values, which is considered fundamental in many real-world scenarios. Nowadays, many companies forecast using hand-written models or naive statistical models. Call centers are the front end of the organization, managing the relationship with the customers. A key challenge for call centers remains the call load forecast and the optimization of the schedule. Call load indicates the number of calls a call center receives. The call load forecast is mostly exploited to schedule the staff. They are interested in the short term forecast to handle the unforeseen and to optimize the staff schedule, and in the long term forecast to hire or assign staff to other tasks. Machine learning has been applied to several fields reporting excellent results, and recently, time series forecasting problems have gained a high-interest thanks to the new recurrent network, named Long-short Term Memory. This thesis has explored the capabilities of machine learning in modeling and forecasting call load time series, characterized by a strong seasonality, both at daily and hourly scale. We compare Seasonal Artificial Neural Network (ANN) and a Long-Short Term Memory (LSTM) models with Seasonal Autoregressive Integrated Moving Average (SARIMA) model, which is one of the most common statistical method utilized by call centers. The primary metric used to evaluate the results is the Normalized Mean Squared Error (NMSE), the secondary is the Symmetric Mean Absolute Percentage Error (SMAPE), utilized to calculate the accuracy of the models. We carried out our experiments on three different datasets provided by the Teleopti. Experimental results have proven SARIMA to be more accurate in forecasting at daily scale across the three datasets. It performs better than the Seasonal ANN and the LSTM with a limited amount of data points. At hourly scale, Seasonal ANN and LSTM outperform SARIMA, showing robustness across a forecasting horizon of 160 points. Finally, SARIMA has shown no correlation between the quality of the model and the number of data points, while both SANN and LSTM improves together with the number of sample / Tidsserie är en samling punkter som samlas in med jämna mellanrum. Tidsseriens analys undersöker tidskorrelationerna och försöker modellera den enligt trend och säsongsbetonade. En av de mest relevanta uppgifterna, i tidsserieranalys, är att förutse framtida värden, som anses vara grundläggande i många verkliga scenarier. Numera förutspår många företag med handskrivna modeller eller naiva statistiska modeller. Callcenter är organisationens främre del och hanterar relationen med kunderna. En viktig utmaning för callcentra är fortfarande samtalslastprognosen och optimeringen av schemat. Samtalslast indikerar antalet samtal ett callcenter tar emot. Samtalslastprognosen utnyttjas mest för att schemalägga personalen. De är intresserade av den kortsiktiga prognosen för att hantera det oförutsedda och för att optimera personalplanen och på långsiktigt prognos för att anställa eller tilldela personal till andra uppgifter. Maskininlärning har använts på flera fält som rapporterar utmärkta resultat, och nyligen har prognosproblem i tidsserier fått ett stort intresse tack vare det nya återkommande nätverket, som heter Long-short Term Memory. Den här avhandlingen har undersökt kapaciteten för maskininlärning i modellering och prognoser samtalsbelastningstidsserier, kännetecknad av en stark säsongsbetonning, både på daglig och timskala. Vi jämför modeller med säsongsmässigt artificiellt neuralt nätverk (ANN) och ett LSTM-modell (Long- Short Term Memory) med Seasonal Autoregressive Integrated Moving Average (SARIMA)-modell, som är en av de vanligaste statistiska metoderna som används av callcenter. Den primära metriken som används för att utvärdera resultaten är det normaliserade medelkvadratfelet (NMSE), det sekundära är det symmetriska genomsnittet absolut procentuellt fel (SMAPE), som används för att beräkna modellernas noggrannhet. Vi genomförde våra experiment på tre olika datasätt från Teleopti. Experimentella resultat har visat att SARIMA är mer exakt när det gäller prognoser i daglig skala över de tre datasätten. Det presterar bättre än Seasonal ANN och LSTM med en begränsad mängd datapoäng. På timskala överträffar Seasonal ANN och LSTM SARIMA och visar robusthet över en prognoshorisont på 160 poäng. SARIMA har slutligen inte visat någon korrelation mellan modellens kvalitet och antalet datapunkter, medan både SANN och LSTM förbättras tillsammans med antalet sampel.

[ES] La diabetes es un importante problema de salud mundial, siendo una de las enfermedades no transmisibles más graves después de las enfermedades cardiovasculares, el cáncer y las enfermedades respiratorias crónicas. La prevalencia de la diabetes ha aumentado constantemente en las últimas décadas, especialmente en países de ingresos bajos y medios. Se estima que 425 millones de personas en todo el mundo tenían diabetes en 2017, y para 2045 este número puede aumentar a 629 millones. Alrededor del 10% de las personas con diabetes padecen diabetes tipo 1, caracterizada por una destrucción autoinmune de las células beta en el páncreas, responsables de la secreción de la hormona insulina. Sin insulina, la glucosa plasmática aumenta a niveles nocivos, provocando complicaciones vasculares a largo plazo. Hasta que se encuentre una cura, el manejo de la diabetes depende de los avances tecnológicos para terapias de reemplazo de insulina. Con la llegada de los monitores continuos de glucosa, la tecnología ha evolucionado hacia sistemas automatizados. Acuñados como "páncreas artificial", los dispositivos de control de glucosa en lazo cerrado suponen hoy en día un cambio de juego en el manejo de la diabetes. La investigación en las últimas décadas ha sido intensa, dando lugar al primer sistema comercial a fines de 2017, y muchos más están siendo desarrollados por las principales industrias de dispositivos médicos. Sin embargo, como dispositivo de primera generación, muchos problemas aún permanecen abiertos y nuevos avances tecnológicos conducirán a mejoras del sistema para obtener mejores resultados de control glucémico y reducir la carga del paciente, mejorando significativamente la calidad de vida de las personas con diabetes tipo 1. En el centro de cualquier sistema de páncreas artificial se encuentra la predicción de glucosa, tema abordado en esta tesis. La capacidad de predecir la glucosa a lo largo de un horizonte de predicción dado, y la estimación de las tendencias futuras de glucosa, es la característica más importante de cualquier sistema de páncreas artificial, para poder tomar medidas preventivas que eviten por completo el riesgo para el paciente. La predicción de glucosa puede aparecer como parte del algoritmo de control en sí, como en sistemas basados en técnicas de control predictivo basado en modelo (MPC), o como parte de un sistema de supervisión para evitar episodios de hipoglucemia. Sin embargo, predecir la glucosa es un problema muy desafiante debido a la gran variabilidad inter e intra-sujeto que sufren los pacientes, cuyas fuentes solo se entienden parcialmente. Esto limita las prestaciones predictivas de los modelos, imponiendo horizontes de predicción relativamente cortos, independientemente de la técnica de modelado utilizada (modelos fisiológicos, basados en datos o híbridos). La hipótesis de partida de esta tesis es que la complejidad de la dinámica de la glucosa requiere la capacidad de caracterizar grupos de comportamientos en los datos históricos del paciente que llevan naturalmente al concepto de modelado local. Además, la similitud de las respuestas en un grupo puede aprovecharse aún más para introducir el concepto clásico de estacionalidad en la predicción de glucosa. Como resultado, los modelos locales estacionales están en el centro de esta tesis. Se utilizan varias bases de datos clínicas que incluyen comidas mixtas y ejercicio para demostrar la viabilidad y superioridad de las prestaciones de este enfoque. / [CA] La diabetisés un important problema de salut mundial, sent una de les malalties no transmissibles més greus després de les malalties cardiovasculars, el càncer i les malalties respiratòries cròniques. La prevalença de la diabetis ha augmentat constantment en les últimes dècades, especialment en països d'ingressos baixos i mitjans. S'estima que 425 milions de persones a tot el món tenien diabetis en 2017, i per 2045 aquest nombre pot augmentar a 629 milions. Al voltant del 10% de les persones amb diabetis pateixen diabetis tipus 1, caracteritzada per una destrucció autoimmune de les cèl·lules beta en el pàncrees, responsables de la secreció de l'hormona insulina. Sense insulina, la glucosa plasmàtica augmenta a nivells nocius, provocant complicacions vasculars a llarg termini. Fins que es trobi una cura, el maneig de la diabetis depén dels avenços tecnològics per a teràpies de reemplaçament d'insulina. Amb l'arribada dels monitors continus de glucosa, la tecnologia ha evolucionat cap a sistemes automatitzats. Encunyats com "pàncrees artificial", els dispositius de control de glucosa en llaç tancat suposen avui dia un canvi de joc en el maneig de la diabetis. La investigació en les últimes dècades ha estat intensa, donant lloc al primer sistema comercial a finals de 2017, i molts més estan sent desenvolupats per les principals indústries de dispositius mèdics. No obstant això, com a dispositiu de primera generació, molts problemes encara romanen oberts i nous avenços tecnològics conduiran a millores del sistema per obtenir millors resultats de control glucèmic i reduir la càrrega del pacient, millorant significativament la qualitat de vida de les persones amb diabetis tipus 1. Al centre de qualsevol sistema de pàncrees artificial es troba la predicció de glucosa, tema abordat en aquesta tesi. La capacitat de predir la glucosa al llarg d'un horitzó de predicció donat, i l'estimació de les tendències futures de glucosa, és la característica més important de qualsevol sistema de pàncrees artificial, per poder prendre mesures preventives que evitin completament el risc per el pacient. La predicció de glucosa pot aparèixer com a part de l'algoritme de control en si, com en sistemes basats en técniques de control predictiu basat en model (MPC), o com a part d'un sistema de supervisió per evitar episodis d'hipoglucèmia. No obstant això, predir la glucosa és un problema molt desafiant degut a la gran variabilitat inter i intra-subjecte que pateixen els pacients, les fonts només s'entenen parcialment. Això limita les prestacions predictives dels models, imposant horitzons de predicció relativament curts, independentment de la tècnica de modelatge utilitzada (models fisiològics, basats en dades o híbrids). La hipòtesi de partida d'aquesta tesi és que la complexitat de la dinàmica de la glucosa requereix la capacitat de caracteritzar grups de comportaments en les dades històriques del pacient que porten naturalment al concepte de modelatge local. A més, la similitud de les respostes en un grup pot aprofitar-se encara més per introduir el concepte clàssic d'estacionalitat en la predicció de glucosa. Com a resultat, els models locals estacionals estan al centre d'aquesta tesi. S'utilitzen diverses bases de dades clíniques que inclouen menjars mixtes i exercici per demostrar la viabilitat i superioritat de les prestacions d'aquest enfocament. / [EN] Diabetes is a significant global health problem, one of the most serious noncommunicable diseases after cardiovascular diseases, cancer and chronic respiratory diseases. Diabetes prevalence has been steadily increasing over the past decades, especially in low- and middle-income countries. It is estimated that 425 million people worldwide had diabetes in 2017, and by 2045 this number may rise to 629 million. About 10% of people with diabetes suffer from type 1 diabetes, characterized by autoimmune destruction of the beta-cells in the pancreas, responsible for the secretion of the hormone insulin. Without insulin, plasma glucose rises to deleterious levels, provoking long-term vascular complications. Until a cure is found, the management of diabetes relies on technological developments for insulin replacement therapies. With the advent of continuous glucose monitors, technology has been evolving towards automated systems. Coined as "artificial pancreas", closed-loop glucose control devices are nowadays a game-changer in diabetes management. Research in the last decades has been intense, yielding a first commercial system in late 2017 and many more are in the pipeline of the main medical devices industry. However, as a first-generation device, many issues still remain open and new technological advancements will lead to system improvements for better glycemic control outputs and reduced patient's burden, improving significantly the quality of life of people with type 1 diabetes. At the core of any artificial pancreas system is glucose prediction, the topic addressed in this thesis. The ability to predict glucose along a given prediction horizon, and estimation of future glucose trends, is the most important feature of any artificial pancreas system, in order to be able to take preventive actions to entirely avoid risk to the patient. Glucose prediction can appear as part of the control algorithm itself, such as in systems based on model predictive control (MPC) techniques, or as part of a monitoring system to avoid hypoglycemic episodes. However, predicting glucose is a very challenging problem due to the large inter- and intra-subject variability that patients suffer, whose sources are only partially understood. These limits models forecasting performance, imposing relatively short prediction horizons, despite the modeling technique used (physiological, data-driven or hybrid approaches). The starting hypothesis of this thesis is that the complexity of glucose dynamics requires the ability to characterize clusters of behaviors in the patient's historical data naturally yielding to the concept of local modeling. Besides, the similarity of responses in a cluster can be further exploited to introduce the classical concept of seasonality into glucose prediction. As a result, seasonal local models are at the core of this thesis. Several clinical databases including mixed meals and exercise are used to demonstrate the feasibility and superiority of the performance of this approach. / This work has been supported by the Spanish Ministry of Economy and Competitiveness (MINECO) under the FPI grant BES-2014-069253 and projects DPI2013-46982-C2-1-R and DPI2016-78831-C2-1-R. Moreover, with relation to this grant, a short stay was done at the end of 2017 at the Illinois Institute of Technology, Chicago, United States of America, under the supervision of Prof. Ali Cinar, for four months from 01/09/2017 to 29/12/2017. / Montaser Roushdi Ali, E. (2020). STOCHASTIC SEASONAL MODELS FOR GLUCOSE PREDICTION IN TYPE 1 DIABETES [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/136574

This thesis is focused on analysis of the development of gross month wage and in particular on development and examining seasonality. There are also described the institutions and their surveys of wages, such as the Czech Statistical Office, Ministry of Finance and the Ministry of Labor and Social Affairs, which administers the Information System of Average Earnings. The monthly income is compared between regions and between major classes KZAM. The development of time series is modeled by the Box-Jenkins methodology, further charts of seasonal values and seasonal indexes . For comparison the average relative wage growth in regions are used cartograms. The base for these analyses is data obtained from business statistical return systems and structural statistics from the site of the Czech Statistical Office and the Ministry of Labor and Social Affairs.