China
Africa
Explore chapters and articles related to this topic
Time Series Forecasting Techniques for Infectious Disease Prediction
Published in Monika Mangla, Subhash K. Shinde, Vaishali Mehta, Nonita Sharma, Sachi Nandan Mohanty, Handbook of Research on Machine Learning, 2022
Jaiditya Dev, Monika Mangla, Nonita Sharma, K. P. Sharma
The rising prevalence of IoT and sensor networks leads to the collection of a vast amount of time-series data that enables forecasting for many IoT applications [1]. The multitude of use cases for time series applications in IoT exists from optimization problems to anomaly detection, prediction, and many more. The prime focus of these applications is to automate a conventional manual system. This automation enables the evaluation of numerous data points ranging from tens to millions along a data series [2]. In particular, time series forecasting applies a statistical model to predict future values based on past results. In such applications, data is collected over time, and the time series model uses that data to forecast future values [3]. The objective of data learning is to analyze the observed environment and/or extract useful information that supports decision-making [4]. The ability to predict the next value in a time series helps to estimate how a specific factor will evolve over time. A higher level of confidence in the prediction model helps to take appropriate action or business decisions accordingly [5].
Time Series Analysis for Modeling the Transmission of Dengue Disease
Published in Dinesh C. S. Bisht, Mangey Ram, Recent Advances in Time Series Forecasting, 2021
A.M.C.H. Attanayake, S.S.N. Perera
This chapter deals with multiple time series modeling techniques, namely, Autoregressive Integrated Moving Average (ARIMA), exponential smoothing, decomposition, Alpha-Sutte modeling, Autoregressive Integrated Moving Average with Explanatory variables (ARIMAX) and exponential smoothing with explanatory variables. We illustrated the application of each technique using dengue cases reported in Jakarta and/or dengue cases reported in Colombo, Sri Lanka, as test cases. Monthly rainfall, number of rainy days within a month, average wind speed, and average maximum temperature data in Colombo were used as explanatory variables, whereas monthly rainfall and average humidity data in Jakarta were utilized for this purpose. Finally, we summarized several combined forecasting approaches and demonstrated how the forecasts of ARIMA and exponential smoothing techniques can be improved by combined forecasting approaches.
Background and Exploration in Time Series
Published in Simon Washington, Matthew Karlaftis, Fred Mannering, Panagiotis Anastasopoulos, Statistical and Econometric Methods for Transportation Data Analysis, 2020
Simon Washington, Matthew Karlaftis, Fred Mannering, Panagiotis Anastasopoulos
Trend in a time series is usually a gradual change in some property of the series and is frequently referred to as a long-term (or secular) movement that corresponds to the general direction in which a time series graph moves over time. The movement is indicated by a trend curve, frequently represented by a simple trend line. Some of the appropriate methods for estimating trends in a time series are the usual least squares approach, the moving average method, and the semi-averages method (Spiegel and Stephens, 1998). Although the notion of a trend is rather simple and straightforward, there are many methods in the time series literature that assume the lack of a trend (stationarity assumption; see Section 7.2.1 for more details), or where a trend actually distorts the statistical relationship of interest, and hence its removal is an important process in time-series analyses (see Section 7.1.2).
Monitoring of Carbon Monoxide (CO) changes in the atmosphere and urban environmental indices extracted from remote sensing images for 932 Iran cities from 2019 to 2021
Published in International Journal of Digital Earth, 2023
Mohammad Mansourmoghaddam, Iman Rousta, Haraldur Olafsson, Przemysław Tkaczyk, Stanisław Chmiel, Piotr Baranowski, Jaromir Krzyszczak
The Time Series Smoothing Algorithm (TES) is a simple approach for smoothing the time series data. Its usefulness for predicting climatic parameters has already been proven in various studies (Mansourmoghaddam et al. 2022a; Indriani et al. 2020). In this method, the weight reduction is attributed to the data exponentially over time (Gardner 1985). The present study has used the TES algorithm to predict the change in the spatially averaged values of used parameters for Iran territory. To smooth the time series and remove the high-frequency information, the TES algorithm was applied to the data three times (Kalekar 2004). For this purpose, a sequence of research time-series data was considered as the KT algorithm. Then, the seasonal change cycle L was selected according to the predicted period of the research, which was 2 years, and the prediction process started from T = 0. Thus, the TES algorithm determined the most optimal estimate of future time data at time T + 1 (i.e. smoothed value) as output, using the time series data available at time T, which was used as KT input, and the data trend line calculation. The final output was the predicted estimate for the value of K at time T + M, where M had to be greater than zero, based on the data up to time T (with T included) (Dev et al. 2018). To evaluate the accuracy of the algorithm, the area of defined land cover classes (the classes for which the area was known) in 2019 and 2020 was predicted by the time series data for previous years, and the results were evaluated by the statistical criterion of Root Mean Square Error (RMSE).
An effective weather forecasting method using a deep long–short-term memory network based on time-series data with sparse fuzzy c-means clustering
Published in Engineering Optimization, 2023
Vasavi Ravuri, Dr. S. Vasundra
Time-series forecasting and analysis of future values have become a major area of research in recent decades. Time-series analysis using time-series information has gained importance in various applications, such as electricity demand, weather, business, stock markets and the use of products, e.g. electricity and fuel. However, forecasting using time-series data provides an organization with an enormous amount of data, which requires some important decisions to be made (Mahalakshmi, Sridevi, and Rajaram 2016). Meteorological data comprise a series of data concerning time periods, and therefore weather forecasting can be analysed by time-series mining (Krishna 2015). The secure transmission of the gathered data is useful for the authentication of the information (Alazeez et al.2016; Mohsin, Li, and Abdalla 2020). Because of the impact of weather on human life, weather prediction has gained more attention from researchers in recent years (Mehrkanoon 2019). Weather forecasting is the scientific procedure of forecasting the condition of the atmosphere considering the location and unique time frames (Hayati and Mohebi 2007; Hewage et al.2020). It predicts future weather conditions, such as wind, pressure, temperature and precipitation, (Sønderby et al.2020; Vasundra 2019a). Improving the warning times regarding climatic disasters could save hundreds of human lives every year. However, the importance of weather prediction can be seen from the point of view of agriculture, for example in the planning of farm operations and the storage and transportation of food grains (Mehrkanoon 2019).
Tracking the evolution of crisis processes and mental health on social media during the COVID-19 pandemic
Published in Behaviour & Information Technology, 2022
Antonela Tommasel, Andrés Diaz-Pace, Daniela Godoy, Juan Manuel Rodriguez
Once the matching for a marker over the entire time span was computed, we obtained a time series distribution for the marker, which models the sequence of observations of a given marker (or dimension) during the defined time window. On average, in the collected dataset, 189.260 tweets were shared per day, with a maximum monthly average of 396.228 in June and a minimum average of 800 in March. To reduce the impact of day-to-day variations and weekly periodicity in the time series, for each marker we computed the gradient over the one-week smoothed time series. This smoothing better exposes the characteristics of the time series, as it responds more slowly to recent changes, which favours the observation of more consistent behaviours over longer periods (in opposition to instantaneous shifts).