Applied Time Series Analysis and Innovative Computing

Virtually any random process developing chronologically can be viewed as a time series. In economics closing prices of stocks, the cost of money, the jobless rate, and retail sales are just a few examples of many. Developed from course notes and extensively classroom-tested, Applied Time Series Analysis with R, Second Edition includes examples across a variety of fields, develops theory, and provides an R-based software package to aid in addressing time series problems in a broad spectrum of fields. The material is organized in an optimal format for graduate students in statistics as well as in the natural and social sciences to learn to use and understand the tools of applied time series analysis. Features Gives readers the ability to actually solve significant real-world problems Addresses many types of nonstationary time series and cutting-edge methodologies Promotes understanding of the data and associated models rather than viewing it as the output of a black box Provides the R package tswge available on CRAN which contains functions and over 100 real and simulated data sets to accompany the book. Extensive help regarding the use of tswge functions is provided in appendices and on an associated website. Over 150 exercises and extensive support for instructors The second edition includes additional real-data examples, uses R-based code that helps students easily analyze data, generate realizations from models, and explore the associated characteristics. It also adds discussion of new advances in the analysis of long memory data and data with time-varying frequencies (TVF).

Aquifers in mountain areas are a strategic resource for the people who live there. To optimise future management, it is vital to understand hydrogeological systems from both geological and hydrogeological perspectives. Historically, methods such as hydrograph and time series analyses have been applied to characterise large karst systems. The aim of this paper was to apply these methods to small mountain springs supplied by porous and shallow aquifers. Specifically were made: (1) a comparison to understand which method better fits the depletion curve of the aquifers and (2) an application of time series analysis both by auto-correlation (analysis of individual series) and by cross-correlation methods (analysis of interrelationships between time series) on all the three parameters monitored from the probe (discharge Q, temperature T, electrical conductivity EC). These techniques were applied on four mountain springs located in the Italy North-Western Alps in the Aosta Valley Region. The results suggested that spring hydrograph and time series analyses on Q, T and EC parameters are useful tools for understanding the hydrodynamic behaviour of porous and shallow aquifers and how to make a proper management of the resource.

The application of machine learning (ML) for use in generating insights and making predictions on new records continues to expand within the medical community. Despite this progress to date, the application of time series analysis has remained underexplored due to complexity of the underlying techniques. In this study, we have deployed a novel ML, called automated time series (AutoTS) machine learning, to automate data processing and the application of a multitude of models to assess which best forecasts future values. This rapid experimentation allows for and enables the selection of the most accurate model in order to perform time series predictions. By using the nation-wide ICD-10 (International Classification of Diseases, Tenth Revision) dataset of hospitalized patients of Romania, we have generated time series datasets over the period of 2008–2018 and performed highly accurate AutoTS predictions for the ten deadliest diseases. Forecast results for the years 2019 and 2020 were generated on a NUTS 2 (Nomenclature of Territorial Units for Statistics) regional level. This is the first study to our knowledge to perform time series forecasting of multiple diseases at a regional level using automated time series machine learning on a national ICD-10 dataset. The deployment of AutoTS technology can help decision makers in implementing targeted national health policies more efficiently.

Thyroid cancer, a prevalent and potentially life-threatening disease, demands early detection for effective treatment. This study proposes a novel approach for the early detection of thyroid cancer by employing Time Series Analysis (TSA) and Quadratic Discriminant Analysis (QDA). The integration of these techniques aims to enhance diagnostic accuracy and reliability, providing a valuable tool for clinicians to detect the disease in its early stages. In our approach, TSA was used to extract meaningful patterns and trends from temporal data, offering valuable insights into the evolving health status of the thyroid. Subsequently, QDA was applied to build a robust classification model, using the identified time series features to distinguish between cancerous and non-cancerous cases. The application of TSA and QDA yielded promising results, demonstrating high sensitivity and specificity, and outperforming traditional diagnostic methods. Our model achieved an accuracy of 97.72%, precision of 90.91%, sensitivity of 94.01%, and specificity of 98.2%. The incorporation of temporal dynamics through TSA provided a nuanced understanding of the evolving pathology, contributing to the enhanced accuracy of the diagnostic model. In conclusion, this study introduces a novel methodology for early thyroid cancer detection, combining the strengths of TSA and QDA. The results highlight the effectiveness of this integrated approach in improving diagnostic accuracy, particularly in identifying subtle temporal changes indicative of thyroid cancer. The proposed TSA-QDA model shows superior performance in terms of sensitivity, specificity, and classification accuracy for multi-class TIRADS classification.

Time Series Analysis

The application of time series analysis to describe the dynamic movements of suspension bridges

TSERIES is a computer program which was written to perform Box-Jenkins type analyses for univariate time series data. The program is capable of performing the analyses described in the first nine chapters of Box and Jenkins (1976) and it follows the general outline given there. TSERIES is both versatile and easy to use. Input is accomplished through a series of simple, compact commands, only a small number of which are needed for standard analyses. However, there are a large number of additional commands that can be used to modify standard assumptions and invoke special options. It is easy to learn to use the program. As a result, it has been very useful in practical applications as well as for giving students hands-on experience in courses on applied time series analysis. One of the important features of TSERIES is its easy-to-use free field command structure which was patterned after two other popular statistical packages, OMNITAB and MINITAB. For example, the following commands would be sufficient to produce outputs for identification of a time series model for a stationary time series of length 200:

Since the 2008 global economic crisis, the detection of financial instabilities has garnered extensive research attention, particularly through the application of time-series analysis. In this study, a novel time-series analysis method, integrating the Kullback-Leibler Divergence (KLD) metric with a sliding window technique, is proposed to detect instabilities in time-series data, especially in financial markets. Global financial time series from 2004 to 2022 were analyzed. The raw time series were preprocessed into return rate series and transformed into complex networks using the directed horizontal visibility graph (DHVG) algorithm, effectively preserving temporal variabilities in network topologies. The KLD method was evaluated through both retrospective analysis and real-time monitoring. It successfully identified idiosyncratic incidents in the financial market, correlating them with specific economic events. Compared to traditional metrics (e.g., moments) and econometric methods, KLD demonstrated superior performance in capturing sequence information and detecting anomalies without requiring linear regression models. Although initially designed for financial data, the KLD method is versatile and can be applied to other types of time series as well.

This study examines the application of time series analysis to predict future bike sales, focusing on two categories: road and mountain bikes. Using the Prophet module in JASP, the research explores trends and weekly seasonality in sales data to derive insights into potential future performance. The study aims to enhance decision-making for inventory management and marketing strategies. A quantitative methodology underpins the research, relying on historical data and the Prophet forecasting tool. The findings reveal no seasonal patterns, different trends for each bike category, enabling targeted business strategies. Practical implications include improved forecasting accuracy and resource allocation. While limited to two bike categories and reliant on the dataset’s quality, the research demonstrates the utility of time series forecasting for retail applications. The originality lies in applying modern forecasting tools to a specific retail context, contributing to both academic literature and industry practice. Our expectations are to have peak sales during weekends when customers have more free time for sports and recreation for both categories, albeit with varying intensities. But weekly patterns are not revealed. The MAPE values (for both categories-mountain and road bikes) are quite high and they do not increase slightly within the time frame of the forecast period. The conclusion is that if a trend was identified the forecast would not be reliable.

Time series analysis on clear-sky brightness temperature (TB) data observed in the morning with ground-based microwave radiometer for atmospheric remote sensing is adapted to judge the working state of the radiometer system according to meteorological data variation features in terms of radiative transfer. The TB data taken as the first example in this study was for the ground-based microwave radiometer at Nanjing during the period from Nov. 27, 2010 to May 29, 2011. The radiometer has 12 channels including five channels at 22.235, 23.035, 23.835, 26.235, and 30 GHz for sensing air humidity and liquid water content and seven channels at 51.25, 52.28, 53.85, 54.94, 56.66, 57.29, and 58.80 GHz for sensing air temperature profiles. The correlation coefficients between the TB readouts from the radiometer and the simulated TB with radiosonde temperature and humidity profiles as input to radiative transfer calculation are greater than 0.9 for the first five channels while the correlation coefficients for the last seven channels are quite poor, especially for the channels at lower frequency such as 51.25, 52.28, and 53.38 GHz, at which the TB readout values in time series do not show the right atmospheric temperature variation features as time goes from November (winter) through spring to early summer (late May). The results show that the first five channels worked well in the period while the last seven channels didn’t, implying that the radiometer need to be maintained or repaired by manufacture. The methodology suggested by this paper has been applied to the similar radiometers at Wuhan and Beijing for quality control and calibration on observed TB data.

The time series analysis approach has the ability to recognize the characteristics of a karst aquifer by analyzing the response of the discharge from rainfall in a karst area. The karst aquifer is considered to be a system that influences the response to the conduit, fissure, and diffuse flow. This research was conducted in Mudal, Kiskendo, and Anjani Springs located in the Jonggrangan Karst, Java Island, Indonesia. The objective of this study was to determine the differences in the response of karst aquifers in releasing the flow in these three springs. The data used in this study are paired data of discharge and rainfall, which were obtained from January to November 2018 (N data = 31,408). Discharge data were obtained by installing an automatic water level logger, while rainfall data were obtained by installing automatic rain gauges in each spring. The time series analysis methods that were used are cross-correlation, auto-correlation, cross-amplitude, spectral density, phase functions, and gain functions. The master recession curve (MRC) calculation was performed to confirm the results from the time series analysis. The study reveals that each spring has a different response in releasing the flow components. Anjani Spring has the fastest response to conduit and fissure flows. However, this spring has the longest release duration of the diffuse flow component, indicating that a well-developed karst aquifer can also have high system memory and large storage capacity. Kiskendo Spring has a faster conduit and diffuse flow response than Mudal Spring but has the slowest fissure flow response, reflecting the void fissure that dominates in this site. In addition, MRC confirmed the results of the time series analysis calculation. This MRC analysis also shows that the study sites were categorized as a high sensitivity degree of karst aquifer. The karst aquifer’s characterization in this study can be used as basic data in the water resources management in the Jonggrangan Karst.

The globalization with the time-space compression caused by the improvement of transportation and communication technology has obscured the territorial organization scales (nation, region and local). It caused the centralization of population and industries in cities, and made urban governance become an important issue. The commission of urban governance has evolved into enhancing governance itself under the shortage of finance through making proper policies and investing efficient expenditures. Thus, we applied time series analysis and relevant variables to analyze the interaction among public expenditures, household disposable income and employment rate in the long term, then compared the governance efficiency between two municipal governments. The results revealed that the public expenditures has positive effect on disposable income and employment rate, in Taipei, however it merely has positive effect on employment rate but no effect on disposable income in Kaohsiung. The results also implied that the governance efficiency could be affected by industrial structures.

joj sam naziv govori, predstavlja skup pojmova, definicija, modela i njihovih primjena u podruju modeliranja i prognoziranja vremenskih serija. Naime, u dananje doba prognoze koje su jako bliske sadanjosti (jako bliska budunost i prolost) (engl. nowcasting 1 ) i prognoziranje openito (engl. forecasting) veoma je vano u poslovnoj ekonomiji

Correlation techniques are important tools for investigating relationships between crop growth and environment. However when applied to a time series the presence of autocorrelation affects the estimates of correlations between two observed series, an effect which is sometimes overlooked in studies reported in the literature. Appropriate statistical techniques are needed to ensure that proper inferences can be made from these observations. Empirical time series modeling has the potential for removing autocorrelations in many of these cases. To test the feasibility of this technique the relationship between boll retention in cotton (Gossypium hirsutum L.) and 5‐day average minimum temperature during the growing season was investigated with 3 years of data. The best fitting autoregressive moving average models were selected for the observed time series for each of the 3 years. Residual series, observed minus predicted values, were computed from the estimated time series models for boll retention and minimum temperature and were free of autocorrelation. Negative crosscorrelations were found between the residual series for boll retention and average minimum temperature for the 5‐day periods beginning with 10 days and 1 day prior to anthesis, respectively. The negative relationship suggests that high nighttime temperatures accumulated over 5‐day periods prior to and during anthesis increase boll shedding. We conclude that time series modeling is an effective technique to aid the identification of relationships between agronomic variables measured in time sequences.

Applications of Least Mean Square (LMS) Algorithm Regression in Time-Series Analysis