Search

The problem of modeling the share of small business in GRP based on ten years of statistical data, calculating the dynamic relationship between them and the structural shifts of their development parameters, developing the most convenient models and estimating forecast parameters is set in this research work. The relevance of the research work is that there are many scientific studies devoted to the study of the relationship of small business to economic processes, but not enough attention has been paid to modeling the impact of small business on GRP using time series. This includes the statistical study of the contribution of small businesses to GRP, the analysis of various models for forecasting their economic relevance using time series, and the development and evaluation of the most appropriate economic-mathematical models. [ABSTRACT FROM AUTHOR]

This paper discusses the empirical best linear unbiased prediction (EBLUP) method that is applied in linear mixed models used for small area estimation (SAE). Parameter estimation in the SAE model using the EBLUP method is obtained by minimizing the mean square error (MSE) of the estimator. The SAE method was initially developed by Fay and Herriot, employing a linear mixed model with random area effects. Rao and Yu developed the model by integrating a random area-time effect component with a first-order autoregressive process, enabling its use for time series and panel data analysis. Moreover, this article focuses on the application of the Rao-Yu model for estimating household consumption per capita expenditure (HCPE) of food and non-food by sub-districts in Langkat Regency. Both datasets were sourced from the National Socioeconomic Surveys (Susenas), which is held regularly by Statistics Indonesia. The application to real data showed that model estimates derived from EBLUP have a lower MSE than those obtained from direct estimates. According to the model, the estimated food expenditures by sub-district in Langkat Regency are significantly higher than the non-food expenditures. [ABSTRACT FROM AUTHOR]

Predicting equity market prices is a complex task due to the dynamic and unpredictable nature of financial markets. This research presents a comprehensive approach to developing an adaptive forecasting model for equity market price prediction using time series analysis. The proposed model integrates traditional time series techniques with modern machine learning methodologies to enhance predictive accuracy and adaptability. The research begins with the collection and preprocessing of historical price data for the target equity. Relevant features are selected and engineered, encompassing technical indicators, macroeconomic factors, and lagged variables of the target variable. The dataset is then partitioned into training, validation, and test sets to facilitate effective model development and evaluation. Various time series forecasting models are considered, including Autoregressive Integrated Moving Average (ARIMA), Seasonal Decomposition of Time Series (STL), Exponential Smoothing State Space Model (ETS), and deep learning architectures like Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRU). Each model's suitability for capturing trends, seasonality, and other relevant patterns is assessed. To achieve adaptability, rolling-window forecasting and online learning strategies are implemented. The rolling-window approach allows the model to be regularly retrained with the most recent data, enabling it to adapt to changing market conditions. Additionally, ensemble techniques are explored to combine predictions from multiple models, mitigating biases inherent to individual models. The developed model is rigorously evaluated using established performance metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and direction accuracy. The model is backtested using historical data to simulate real-world scenarios and to refine its predictive capabilities. The results demonstrate the potential of the proposed adaptive forecasting model to provide valuable insights into equity market price movements. However, it is crucial to acknowledge the inherent uncertainty and volatility of financial markets, which can impact the model's predictions. In conclusion, this research contributes to the field of equity market prediction by presenting a holistic approach to developing an adaptive forecasting model. By combining time-tested techniques with contemporary machine learning methods and adaptability strategies, the model offers a promising framework for enhancing the accuracy of equity market price predictions. [ABSTRACT FROM AUTHOR]

Agriculture is essential to achieving the country's food and nutritional security objectives. The Cauvery Delta Zone in Tamil Nadu contributes significantly to the state's agricultural productivity. However, productivity has decreased in recent years, and farming circumstances in the delta region have deteriorated since 2023. Farmers' incomes are low due to factors such as changes in soil water intrusion, increased irrigation patterns, and agricultural input prices. Furthermore, climate change has raised the cost of production for farmers by forcing them to purchase additional herbicides and insecticides to combat new crop diseases. Water scarcity and debts are the most prominent grounds for farmer suicides in the Cauvery Delta region. To appreciate the importance of agriculture to household income, one must first grasp the nature of landholding. Furthermore, the study used time series analysis to examine land use and cropping patterns in the Tamil Nadu district. The findings show a drop in net sown area and a shift in cropping patterns as a result of irregular rainfall patterns and insufficient irrigation water infrastructure in the district. [ABSTRACT FROM AUTHOR]

Fuzzy time series has gained widespread popularity as a forecasting method in recent years, with many researchers striving to enhance its performance by modifying clustering steps. This discussion focuses on the clustering process, specifically dividing it into two methods: the Kumar Method and the Cross Association Method. In the experiment, forecasts were computed using both methods, and the results, along with error values, were compared. The findings reveal that the Kumar Method demonstrates superior forecasting accuracy. This is substantiated by the smaller values of Mean Squared Error (MSE) and Average Forecasting Errors (AFER) at 197607.2 and 1.212%, respectively. In contrast, the Cross Association Method yielded higher MSE and AFER values at 304255.9163 and 1.128%, indicating a comparatively lower accuracy in forecasting. [ABSTRACT FROM AUTHOR]

The production of plastic waste from year to year has unstably decreased and increased. In 2019, the national average daily plastic waste production reached 5516 m3. Waste management that has not been maximized is a problem to be worried about. As much as 32% of all plastic waste is not recycled and damages natural ecosystems. The massive impact caused by plastic waste requires careful planning so that later an appropriate policy in waste management can be taken and minimize the negative impact of plastic waste. Based on the results of the study, it was found that the high order fuzzy time series method is better than conventional methods such as ARIMA and Fuzzy logic. First-order FTS has a good level of accuracy with the MSE value of 1013455.3 m3 and the AFER 34.12%. Forecasting plastic waste production data in Indonesia using the first-order fuzzy time series method produces forecasting results in 2020 it is 10500 m3, in 2021 it is 10500 m3, in 2022 it is 13900 m3, and in 2023 it is 11633.33 m3. [ABSTRACT FROM AUTHOR]

Indonesia is the fourth most populous country in the world, after China, India, and the United States, with a population of 276,361,788. Regarding efforts to satisfy the need for rice as a staple food. To meet the demand for rice and achieve food independence, the Ministry of Agriculture promotes an increase in rice production. According to the Ministry of Agriculture, increasing rice production to achieve self-sufficiency will improve farmer welfare and decrease poverty. This study aims to investigate the impact of production policies on the growth and productivity of rice production. This investigation included both qualitative and quantitative analysis. This study employs both primary and secondary data as annual time series from 1980 to 2020. This study's model was created using the Two Stage Least Squares (2SLS) technique. The Ministry of Agriculture's efforts to increase rice production and productivity have resulted in a rice surplus over the course of four decades. These efforts have taken the form of expanding the harvest area, expanding the area of irrigated land, seed subsidies, and fertilizer subsidies. [ABSTRACT FROM AUTHOR]

Coronavirus Disease (COVID-19) is a new disease that first hit the world in 2020. This disease is thought to have originated in Wuhan, China [1]. WHO has designated COVID-19 as a pandemic which has successfully infected more than 190 countries in the world. DKI Jakarta is the capital city of Indonesia which is also one of the provinces with the highest number of positive confirmed cases of COVID-19 until the end of July 2020. This study aims to model the rate of COVID-19 cases in 15 sub-districts of DKI Jakarta with the highest intensity. The area of 15 sub-districts are chosen with the highest intensity to be the study area where more than 63.43 % of confirmed COVID-19 cases are in this area. The sub-districts are Gambir, Menteng, Sawah Besar, Kemayoran, Taman Sari, Senen, Tanah Abang, Johar Baru, Tambora, Grogol Petamburan, Cempaka Putih, Pademangan, Setia Budi, Matraman, and Palmerah. The rate of COVID-19 cases in this area is then analyzed using a GSTAR model. This model is one of the models in a stochastic time series that considers spatial component or location and time [2]. The binary weight matrix, uniform weight matrix, and distance weight matrix in this study were formed as a spatial dependency matrix among locations or called the W weight matrix. The results of Space Time Autocorrelation Function (STACF) and Space Time Partial Autocorrelation Function (STPACF) services for all spatial weighting matrices obtained the same model, that is GSTAR (3,1). Estimation of parameters of the GSTAR model (3,1) is carried out for each weighting matrix. The best GSTAR (3,1) model is based on a distance weighted matrix, with the smallest RMSE. [ABSTRACT FROM AUTHOR]

This paper presents an ARIMA model capable of accurately very short-term forecasting the global solar radiation in the period of 5 minutes time interval. Time series data deployed in analyzed method accounting for three years solar irradiances from 2011 to 2013 was collected from Naresuan University, Phitsanulok province of Thailand. The preparation data sequentially arranged and transformed into a stationary form suitable for a time series model using the Box-Jenkins procedure in the SPSS program. The correlation test of error technique was used to determine the ARIMA parameters. The results show the ARIMA(2, 1, 3) model efficiently forecasts very-short term global solar radiation. RMSE and MAPE are used to determine the model accuracy against the actual data showing 0.9125 and 0.059 respectively on the clear sky day because of smooth trend following sun position, while higher RMSE and MAPE on 6.1368 and 0.2373 respectively on the cloudy day. However, both characteristics show the lower RMSE and MAPE indicate better forecasts. The more forecasted irradiances accuracy is challenged which utilized to plan and regulate an electricity supply to meet the demand load and implementing in the solar power plant design for maximize benefits. The hybrid Neural Network and ARIMA address the limitations and can significantly improve forecasting accuracy. [ABSTRACT FROM AUTHOR]

We present in this paper the prediction of the NTP (Farmer's Exchange Rate) of Central Java Province by Chen's Fuzzy Time Series. Prediction of NTP needs to be done to describe the NTP index in the coming period so that related boards can prepare preventive action if the NTP index falls from the previous period. By using the data of NTP from January 2017 till August 2022, we could find that the value of NTP for September 2022 is 105.57 percent, with a MAPE of 0.37. Further, we could predict NTP for Central Java Province until December 2024. The prediction shows that the NTP is above 100 percent. So, the farmer will gain the profit until December 2024. [ABSTRACT FROM AUTHOR]

After Covid-19 pandemic, there has been no research that has modeled the number of plane passengers by considering the influence of seasonal effects. The purpose of the research is to predict domestic air passengers' traffic at Juanda Airport after the Covid-19 pandemic. The forecasting model used in this research is using the SARIMA-GLS-ANN hybrid model. SARIMA is used to capture linear data patterns while the GLS method excels in determining parameters. ANN excels in modeling data resulting from non-linear processes. The combination of linear approach in SARIMA, parameter estimation in GLS and nonlinear in ANN can be applied. A sample of 48 time series data from January 2019 to December 2022 was processed with the SARIMA-ANN method on SPSS 25, obtained valid data as many as 36 samples with 25 samples of training data (69.4%) and 11 samples of testing data (30.6%). The result of the difference in relative error in training data and testing data on SARIMA-ANN is 42.39%. Furthermore, data processing with the SARIMA-GLS-ANN method on SPSS 25 obtained valid data as many as 35 samples with training data as many as 23 samples (65.7%) and testing data as many as 12 samples (34.3%). The result of the relative error difference in training data and testing data on SARIMA-GLS-ANN is 4.29%. The SARIMA-GLS-ANN method is better at solving forecasting or predict domestic air passengers' traffic at Terminal I of Juanda Airport Surabaya. [ABSTRACT FROM AUTHOR]

Space time or spatial time series data is time series data that is not only related to events at previous times, but also has a relationship to location. Generalized space time autoregressive integrated-autoregressive conditional heteroscedastic is a model used to analyze spatial time series data where this model considers the differentiation process. According to the law of probability, the process does not change over time, so if the data is not stationary, it can make the estimation results less precise, so a differentiation (integrated) process is needed. In reality, it is often found that the variance changes over time or is heteroscedastic. Autoregressive conditional heteroscedastic analyzes the error variance problem differently for each time series observation. In this condition, the variance error not only functions from the independent variable, but also depends on how big the previous error was. The aim of this research is to model Covid-19 cases with the GSTARI-ARCH model which is compared with the GSTARI model by estimating model parameters using two methods. Parameter estimation using the least squares method produces the linear form β ^ (i) = (X i ' X i) − 1 X i ' X i . The error in the autoregressive conditional heteroscedastic model is estimated using the maximum likelihood method assuming constant correlation, so that Σ = DtRDt is obtained. The modeling results show that the MAPE value of GSTARI-ARCH is smaller than GSTARI, so the GSTARI-ARCH model is the best model. [ABSTRACT FROM AUTHOR]

Time series clustering is a data mining technique where it involves grouping of time series data into a few separate clusters. In this paper, a shape-based clustering is applied on time series of PM10 data collected from 60 air quality monitoring stations in Malaysia from 5th July 2017 to 31st December 2018 obtained from Malaysian Department of Environment. k-shape clustering method along with shape-based distance (SBD) were used to cluster the PM10 time series data. Air quality patterns were analysed from the clusters formed. The results show that the clusters formed using the method used are very well separated and mainly influenced by the region and geographical locations instead of the station's categories or man activities at the locations. [ABSTRACT FROM AUTHOR]

Exchange rate instability poses a global economic challenge that impacts international trade and the overall economies of countries, affecting their human development as well. This paper proposes a method that clusters the exchange rate time series of countries and it's compared with the Human Development Index (HDI) categorization. By doing so, a deeper understanding of the relationship between currencies and development status can be gained. In this study, a fuzzy clustering framework was employed to group the time series of exchange rates for specific countries. A total of twenty-five currencies were chosen for analysis during the period from January 01, 2014, to December 31, 2018. Time series clustering was conducted based on the autocorrelation patterns of the time series. The fuzzy c-means approach was used to calculate the similarity between their autocorrelations. The analysis findings indicate the Autocorrelation-FCM approach demonstrated better cluster results compared with the FCM. Moreover, it was observed that the clustering of exchange rates significantly influenced the clustering of HDI. Based on these results, it is recommended that policymakers adopt a diverse range of policies to ensure stability in the exchange rate to foster human development. Such measures hold the potential to bring about significant changes and improvements in monetary policy across different nations. [ABSTRACT FROM AUTHOR]

A statistical modeling of air pollution index data can provide useful information about the characteristics and behaviors of air pollution events. This study evaluates the performance of the three parsimonious time series model on API data in Malaysia, which are: autoregressive integrated moving average, seasonal autoregressive integrated moving average, fuzzy time series, and artificial neural networks. A case study involving the analysis and modeling of API data for 16 locations in Malaysia is conducted. Results show that the API values fall between good and moderate levels. Some large fluctuations of the APIs are found to be mainly due to haze events. Modeling results indicate that the ARIMA method is highly applicable to the accurate prediction of API values in most cases. Nevertheless, the other forecasting methods could serve as good alternatives because of their similar performance to the ARIMA method. [ABSTRACT FROM AUTHOR]

In this study, hybrid approaches are proposed to deal with the complex features of nonstationary and nonlinear behaviour. The proposed model takes into account some features, such as the seasonal component and the trend as an input data. However, to the best of the authors' knowledge, statistical patterns, including trend, seasonality and nonlinear behaviour, have rarely been studied simultaneously in bitcoin data. Therefore, a hybrid of structural time series (STS) and nonlinear autoregressive with exogenous input (NARX) is developed to predict bitcoin closing price. The results reveal that the optimal model is the hybrid STS, combining Local Level with Drift Deterministic Seasonal and NARX, featuring two inputs, one output, two delay values, and ten neurons in a hidden layer. This model demonstrates superior performance with the lowest values across all three measurement errors when compared to a single linear STS in bitcoin price prediction. [ABSTRACT FROM AUTHOR]

S.R. Singh presented in 2009, a high-order fuzzy time series that served as the foundation for a computational forecasting approach. The new computational technique overcomes the drawbacks of previous high-order fuzzy time series models more effectively. Its simplicity is enhanced by the use of a w-step fuzzy predictor as the forecasting parameter, rather than costly computations of fuzzy logical relations and deviations in successive values of several orders. This approach is used to decide which order is preferable by evaluating the forecasting model using sMAPE on DKI Jakarta Province export value data from January 2017 to May 2022. When compared to the other orders, the seventh order has the lowest SMAPE of 2.626%. In this case, S.R. Singh Fuzzy Time Series order 7 delivers better and more accurate results than other orders. Furthermore, S.R. Singh Fuzzy Time Series may be used with data that has cyclical patterns as well as data that has monthly time intervals. [ABSTRACT FROM AUTHOR]

The Covid-19 pandemic and Russian invasion of Ukraine made food and energy prices in the world soar. This price spike has triggered inflation in various countries, such as Indonesia. Currently, the most developed forecasting method is the time series. The development of a multivariate time series, apart from looking at the time element, also involves an element of location. One model that involves time and location is the Generalized Space Time Autoregressive (GSTAR) model. The GSTAR model is a development of the Space Time Autoregressive (STAR) model which assumes that the location element is heterogeneous. The purpose of this study was to obtain the GSTAR model using Seemingly Unrelated Regression (SUR) on inflation data in five cities on the island of Sumatra. The methods used in this study are, 1) perform descriptive analysis, 2) perform stationarity test, 3) identify the GSTAR model, 4) calculate the location weight matrix, and 5) calculate parameter estimates for the GSTAR model using the SUR method. In this study, it is found that the inflation model for five cities on the island of Sumatra at time t is correlated with inflation data at the previous time and is influenced by inflation data from other cities. [ABSTRACT FROM AUTHOR]

To analyze economic data, it would be better to use panel data (a combination of time series and cross section) because there is a lot of information obtained. One model that is widely used in panel data analysis is the PVAR model because it was able to identify the dynamic effects of the heterogeneity shock of the unobserved variables and their causal effects. PVAR is widely used in data on the economic growth of a region or country. Many factors affect economic growth, one of which is inflation. Inflation and economic growth are two things that influence each other. Economic growth it self can be calculated based on the value of GDP. In this study, Granger causality shows that there is a two-way influence of the two variables. The PVAR model shows that the two variables influence each other but the time effect has a greater influence. Where the GDP rate in the previous year had more influence on the current GDP rate than the inflation rate. [ABSTRACT FROM AUTHOR]

The Generalised Space-Time Autoregressive Moving Average-Exogenous (GSTARMA-X) model has been implemented to model spatial time series where the observed values at each location depend not only on the past observed values at neighbouring locations, but also on past values of an exogenous variable at neighbouring locations. Since space time series models involve the value of an observed variable at multiple locations, any such model can be viewed as a vector time series model, which enables the usage of methods commonly utilised to analyse vector time series. In this paper, datasets following a GSTARMA-X (11,11)-(11,11) were generated and analysed by first identifying an appropriate vector ARMA-X (VARMA-X) model for the space-time model, then representing the VARMA-X model as a state-space model. Finally, the state-space model is analysed using the Kalman filter method. The datasets generated fall into either one of three categories: stationary and invertible data, non-invertible but stationary data, and nonstationary but invertible data. The Kalman filter performed significantly better for the nonstationary but invertible data than for any of the other categories, however the Kalman forecasting only performed moderately well for the stationary and invertible data and not so much for the other categories. [ABSTRACT FROM AUTHOR]

Forecasting is one of the important elements for making a decision. One of the most frequently used forecasting methods is fuzzy time series. In recent years, a number of methods have been proposed for forecasting using the fuzzy time series method. This study aims to explore how to determine the division of the length of the interval. The division of intervals in the fuzzy time series usually divides the intervals into equal lengths, whereas in this study, the division of intervals will be carried out using the ratio interval method and then combined with the frequency density method. The ratio interval method does not divide the intervals equally in length, but divides the intervals based on the ratio. The obtained intervals are then partitioned again using the frequency density method. This study focuses on the comparison between the fuzzy time series method based on the interval ratio, the fuzzy time series method based on the frequency density, and the combination of the two methods. These methods will be applied to Indonesian rubber production data for 20 years, from 2000 to 2020. The result shows that the combination of the two methods gives the best accuracy. [ABSTRACT FROM AUTHOR]

Activation functions are important hyperparameters in neural networks, applied to calculate the weighted sum of inputs and biases and determine whether a neuron can be activated. Choosing the most suitable activation function can assist neural networks in training faster without sacrificing accuracy. This study aims to evaluate the performance of three activation functions, Sigmoid, Hyperbolic Tangent (Tanh), and Rectified Linear Unit (ReLU) in the hidden layer of Multilayer Perceptron (MLP) for time series forecasting. To evaluate the activation functions, three simulated non-linear time series were generated using the Threshold Autoregressive (TAR) model, and two real datasets, the Canadian Lynx series and Wolf's Sunspot data, were employed. The Mean Square Error (MSE) and Mean Absolute Error (MAE) were computed to measure the performance accuracy. The analysis of the real data revealed that the Tanh function exhibited the lowest MSE and MAE, with values of 1.345 and 0.945, respectively. The Sigmoid function yielded MSE and MAE values of 1.520 and 1.005, while the ReLU function resulted in values of 1.562 and 1.018. These findings align with the simulation results, confirming that the Tanh function is the most effective for time series forecasting. Therefore, it is recommended to replace the commonly used Sigmoid function with Tanh for an accurate forecast. [ABSTRACT FROM AUTHOR]

The accurate operation and planning of power plants depend heavily on the ability of utilities to estimate load demand, which also enables them to allocate resources more efficiently and guarantee a steady flow of electricity. Modern models have limited relevance in real-world situations and demonstrate severe non-linearity in load information from forecasts that are now accessible. But for practical use, the field of energy forecasting needs more flexibility, enhanced prediction accuracy, and resilience. This study proposes a brand-new method for anticipating short-term load demand that has been uniquely developed for Bangladesh. An extensive dataset made up of historical load demand data, weather factors, and Bangladesh-specific calendar data is used to train and test the model that is suggested. The dataset spans six years of load demand data, covering several seasonal trends. The proposed model uses a parallel CNN-BiLSTM architecture. In this architecture, the BiLSTM module keeps track of temporal relationships and learns long-term trends, while the CNN module pulls out spatial features from the merged dataset. The parallel nature of the model enables it to capture the intricate correlations visible in the load demand time series data. The model also makes use of parallel processing's advantages, which boost the effectiveness of computation. Based on the results of experiments, the proposed CNN-BiLSTM hybrid computational neural network system is better at making accurate predictions than baseline models and traditional forecasting methods. The model successfully captures short-term load demand changes, enabling providers to build facilities, balance load, and schedule resources accordingly. The Mean Absolute Percentage Error (MAPE) and R-squared score, which surpass the existing state-of-the-art models, were taken into consideration for evaluation. [ABSTRACT FROM AUTHOR]

In this age of growing Functional Data Analysis (FDA) procedures, it might be challenging to express time series data using the basis functions that are often used in the FDA smoothing techniques due to the rigidity of existing methods. Thus, it is essential to develop advanced smoothing techniques to interpret massive data sets flexibly and to portray the data accurately in functional form. In this study, a comprehensive systematic literature review was done between 2018 and 2023 by focusing on smoothing techniques applied in FDA. The contributions of this study are on (1) the advantages and disadvantages of the techniques and (2) the current trend and future direction of the study. In conclusion, this study demonstrates the significance of the FDA in a wide range of sectors, the advantages, and disadvantages of smoothing techniques for time series data, and its significance issues. Since the FDA makes no assumptions regarding the data, applying it to all time-series studies should improve the modeling technique. [ABSTRACT FROM AUTHOR]

Bitcoin price can be challenging to predict due to the presence of non-stationary and non-linear patterns in the data series. It can take time to make accurate predictions in such circumstances. The fact of trends and seasonal fluctuations rarely makes the modeling process more accessible. Therefore, most researchers isolate hidden behavior, such as trends, seasonal and irregular components from the actual series. As a result, information about the data is lost, and the prediction becomes inaccurate. Therefore, this study used the linear structural time series (STS) model to cope with uncertainties by using previous information about the model structure. However, the linearity of the underlying STS model and the accurate knowledge of it are often not available in practice. This is the disadvantage of a single model that cannot handle the different patterns. Therefore, this study proposes the hybrid model STS-NARX to account for non-stationary and non-linear behavior data series. The result shows that the proposed STS-NARX hybrid model has higher accuracy in predicting the Bitcoin price than a single linear STS model due to the greater abundance of information. [ABSTRACT FROM AUTHOR]

The service sector has been seeing growth over the past few years, and this trend is anticipated to continue. The ability to serve customers is one of the key characteristics of a business. The most important thing is to use the available resource efficiently. To do this, businesses are utilizing cutting-edge technology. An intelligent sequence model is furthermore needed for the prediction of taxi demand when coupled with a time series forecasting issue. We discuss the challenges in predicting taxi demand, such as data sparsity,and propose solutions to address these challenges. Our outcome shows that our model outperforms existing methods in terms of prediction accuracy and computational efficiency. [ABSTRACT FROM AUTHOR]

Given that G is a finite, connected, and undirected graph. Let the vertex set and edge set be denoted as V (G) and E(G), respectively. If every vertex in graph G is either in D or adjacent to a vertex in D, then subset D of V (G) is an efficient dominating set of graph G. If any vertex in G can be uniquely identified from every other vertex in an ordered set W through its distinct representation, then subset W of V (G) is a resolving set of G. Let (W = w1, w2, w3, ..., wk) be a subset of V (G). Assuming that G represents the resolving efficient dominating set, and γre(G) represents its minimum cardinality. The resolving efficient domination number of graphs is what we refer to as γre(G). In this study, we obtained γre(G) for the Dutch Windmill graph Wn,m, groundfield graph Gdn and the Greenfield graph G fn. Furthermore, to see the application of resolving efficient dominating set, at the end of this paper will illustrate the implementation of it on graph neural network (GNN)multi-step time series forecasting on the forecasting of natrium, phosphorus, and potassium (NPK) nutrients and water pH in hydroponic plants. [ABSTRACT FROM AUTHOR]

Let G = (V, E) be a graph with vertex set V and edge set E. A proper k − vertex coloring of a graph G = (V, E) is a partition P = {V1,V2, ...Vk} of V into independent set. The graph G is b-coloring if there is a vertex in each color class, which is adjacent to at least one vertex in every other color class. The b-chromatic number of graph G, denoted by ϕ(G), is the largest integer k such that there is a b-coloring with k colors. In this paper, we will study b−coloring of pencil graph Pcn, diamond snake graph Dsn and triangular ladder graph Tln. Furthermore, to see the robust application of b-coloring, at the end of this paper we will illustrate the implementation of it on analyzing the multi-step time series forecasting on soil moisture, temperature and pH of companion farming. [ABSTRACT FROM AUTHOR]

The cryptocurrency market is a rapidly changing environment that presents considerable problems for investors, traders, and analysts. Predicting cryptocurrency prices has been a popular topic in recent years. These currencies are used by millions of people worldwide, and their value fluctuate according to demand. Researchers and investors are making progress in constructing predicting models for cryptocurrency values, thanks to increased availability of historical data and breakthroughs in machine learning techniques. In this study we took four different models (Auto regressive integrated moving average, Support Vector Regression, Long Short Term Memory and Gated Recurrent Unit) to find the most effective way to predict price of cryptocurrency and then deploying into a web application for users to predict price. The study compares the mean absolute percentage error of these three models. For our study we have taken Litecoin a well-known cryptocurrency for model training and testing. The data has been fetched from yahoo finance which then goes through transformation and finally fitted into the ml model. The most effective model turns out to be LSTM with a MAPE of 37.45 to produce a prediction. [ABSTRACT FROM AUTHOR]

The exponential growth of big data has necessitated the development of effective analytical methods for real-time processing of massive amounts of datasets & samples. The design of an effective VARMAx (Vector Autoregressive Moving Average with Exogenous Inputs) model that takes advantage of in-memory computing significantly improves the speed and computational efficiency of big data analytics. This work is required because big data streams must be processed in real-time, as delays can negatively impact business decisions and operations. By anticipatorily processing real-time samples using the VARMAx model, we demonstrate a significant reduction in delay, achieving a 10.5% improvement over recently proposed big data analysis methods. The analysis of sensor data, optimization of the supply chain, and financial forecasting are just a few of the many domains and applications that can benefit greatly from our methodology & operational process. The VARMAx model is a versatile instrument for the analysis of big data because it can handle a variety of data types, including time series, textual, and categorical datasets & samples. Our research also demonstrates the computational efficiency advantage of the VARMAx model over other approaches. By utilizing in-memory computing, which enables faster processing and analysis of large data streams, we achieved a remarkable 8.3% increase in computational efficiency levels. The ability of the VARMAx modeling process to capture the intricate dependencies and relationships present in big data streams justifies its application. Considering both the autoregressive and moving average components as well as exogenous inputs, the VARMAx model successfully captures temporal patterns, seasonality, and the effect of exogenous factors on the datasets & samples. This comprehensive modeling strategy improves the analytics' precision and predictive capabilities, making it the best option for utilizing in-memory computing for large-scale data analysis. [ABSTRACT FROM AUTHOR]

Train delays can have a significant impact on transport systems, causing passengers inconvenience and operational challenges for train operators. One of the causes of railway delays is seasonality, which includes holidays and weather patterns. ARIMA model is a well-known statistical model for analyzing and forecasting time series data. However, the ARIMA model is unable to deal with seasonal time series data. Furthermore, time series data frequently exhibit stochastic patterns that may include outliers, whereas the traditional ARIMA model fails to produce accurate results when the data contains outliers. This study aims to forecast train delays based on Network Rail dataset using Facebook (FB) Prophet. FB prophet is a flexible time series model that is renowned for its ability to deal with seasonality and outliers. The number of 713 daily delayed occurrences (in minutes) from Edinburgh, Scotland was taken into account and divided at a ratio of 70:30 for model fitting and evaluation, respectively. The testing dataset was fitted with ARIMA, FB Prophet with and without holidays effect. The mean absolute error (MAE) and root mean squared error (RMSE) were used to evaluate the models. The RMSE and MAE of the FB prophet model with holidays feature deliver best performance in term of forecasting accuracy with the lowest rate at 968.446 and 638.558 respectively, as compared to ARIMA and FB prophet without holidays feature. The results are further improved after removing the outliers. This study highlights the potential of FB Prophet model with holidays feature as a predictive analytic tool to enhance train services. [ABSTRACT FROM AUTHOR]

On data that are not normally distributed, classical linear regression is developed into the Generalized Linear Model (GLM). One of them is logistic regression which is used for binary discrete data. A different strategy is to consider all variables (response and predictors) to be random observations from the joint probability distribution. A model with dependencies should be used because some factors are dependent. The copula model, one of which is the Gaussian copula, is a popular technique for creating multivariate models with dependencies. The Gaussian copula regression model (GCMR) is a general structure that can be used to model any form of dependent response. The Gaussian copula blends the ease of interpretation of marginal modeling with the flexibility of the dependency structure specification. In this case, the estimated parameter is the regression coefficient and the error is expressed in the dependency structure. The model estimation uses the maximum likelihood method. This paper aims to apply the GCMR model to time series data where the dependent variable is discrete with a binomial distribution. A case study was conducted to see the effect of weather factors on Covid cases in DKI Jakarta. The regression model used is binomial data logistic regression with the dependency structure expressed in the correlation matrix with the assumption that the error is obtained from the ARMA(p,q) process. The dependent variable is the number of Covid-19 cases and the independent variable is the maximum temperature, average temperature and humidity. The data are a daily time series with a range of March 1, 2020 to April 30, 2022. Based on the GCMR model, Maximum temperature, average temperature and Humidity have a significant influence on the number of Covid-19 cases in Jakarta. For this case study, the results show that the GCMR model is better than logistic regression analysis. [ABSTRACT FROM AUTHOR]

The nonlinear system of a photovoltaic (PV) module was studied in this article. For PV panel temperature prediction, three types of Artificial Neural Network (ANN) time series are used: Nonlinear Autoregressive with External Input (NARX), Nonlinear Autoregressive (NAR), and Nonlinear Input-Output modeling. The model's input data parameters were (ambient temperature, humidity, wind speed, and irradiance), while the model's output data parameter was the temperature of the photovoltaic panel. All modeling techniques are evaluated using Mean Square Error (MSE), and their results are compared to determine which technique is the best. The results indicated that the MSE for the NARX method was 0.1196, while the MSE for the NAR and nonlinear input-output methods were 0.1385 and 0.2806, respectively. [ABSTRACT FROM AUTHOR]

The definition of the design storm is of pivotal importance in hydrological studies, such as for the design of urban drainage networks and for flood risk management. The design storm is usually defined throughout frequency analysis of the observed annual maxima of rainfall depth, by defining the probability distribution that best fits the sample of observations. Due to the recent increasing frequency of urban floods and to the gained interest in climate change effects on precipitation features, many local water authorities are asking water engineers to employ not just long time series, but also recent samples to estimate design storm values. However, although it is well recognized the role of sample size in frequency analysis, it has not been investigated yet the influence of the recentness of samples in the estimation of design rainfall. In this study we analyze the influence of both sample features on the estimation of a 100 years return period and 1-hour duration design storm, by using rainfall time series of rain gauges located in central Italy. The study reveals that in most of cases the time frame from which samples are extracted is relevant only for short sized samples, while for opportunely large samples the values of design rainfall computed using recent observations is similar to those obtained with older observations. [ABSTRACT FROM AUTHOR]

Demand forecasting is one of the key aspects of operations management, particularly sales forecasting, as it plays a fundamental role in the retailer's resource planning system, which impacts consumer satisfaction. Traditional time series models are widely used in the context of sales forecasting, while machine learning approaches have been playing an important role in recent years. Different approaches should be tested and compared to support practical knowledge. In this work, we compare different prediction methods to forecast diesel fuel sales in a small business in order to choose the most suitable method, in terms of the usual evaluation measures, for presenting an inventory planning for weakly fuel sales. [ABSTRACT FROM AUTHOR]

Since air pollution is a major public health concern, collecting and analyzing air quality indicators data is very important for monitoring pollution. Data on these indicators are generally collected through air quality monitoring stations located in specific areas. In this study, 12 monitoring sites belonging to the Beijing Municipal Environmental Monitoring Center air pollution monitoring network are clustered based on their similarity in terms of hourly or monthly concentrations of some air pollutants. The hourly data was collected from these stations between March 2013 and February 2017. The clustering procedure was performed through average linkage and partitioning around medoids algorithms. The preliminary results obtained from the hierarchical algorithm show the presence of two clusters that are well distinguished with meaningful interpretation while the two obtained from the partitional algorithm are not well distinguished. [ABSTRACT FROM AUTHOR]

The Seasonal Autoregressive Integrated Moving Average (SARIMA) models which built upon the ARIMA models to support seasonality, are used when it encounter the periodic time series data. It has been widely used in infectious disease prediction and other fields where data indicate a seasonal pattern. However, there are lacks of research focuses on the impact of the length of training and test data on the forecasting accuracy. In this study, the data is split into two parts: Training and Test data. The selected SARIMA model will be fitted for different lengths of training data and forecast the observations of the length of test data. This study aims to investigate how the length of training and test data affect the forecasting accuracy. In order to check it, the Mean Absolute Percentage Error (MAPE) for different lengths of training data are calculated and compared to study their relationship. [ABSTRACT FROM AUTHOR]

Large community of high-energy physicists share their data all around world making it necessary to ship a large number of files over wide- area networks. Regional disk caches such as the Southern California Petabyte Scale Cache have been deployed to reduce the data access latency. We observe that about 94% of the requested data volume were served from this cache, without remote transfers, between Sep. 2022 and July 2023. In this paper, we show the predictability of the resource utilization by exploring the trends of recent cache usage. The time series based prediction is made with a machine learning approach and the prediction errors are small relative to the variation in the input data. This work would help understanding the characteristics of the resource utilization and plan for additional deployments of caches in the future. [ABSTRACT FROM AUTHOR]

Wheat is the main growing cereal crop in India. For the country, designing, estimating productivity is the important tool for predicting the production of wheat to determine the circumstances that would be the value of production upcoming year. Fuzzy time series deals with the problems associated with unpredictability ambiguousness and cloudiness. In this study enlargement of fuzzy time series variant model is used in forecasting wheat production. The data of previous years from 2000 to 2021 was taken for analysis and the proposed method on fuzzy time series estimates future events' precision rate. The suggested method used the percentage change as the Universe of Discourse and Average-based Partitioning. [ABSTRACT FROM AUTHOR]

Time series model is one of the best tools which is used to describe and forecast of phenomena behavior. The determination of the order of the model is an important step to build a model for a certain phenomenon. Different criteria are proposed and derived for this purpose. Each criterion is based on is based on a respectable scientific logical point of view. The most popular criterion in this regard is the Akiaki final prediction error. Unfortunately, the Akiaki criterion has tasteless performance in finite series case. A criterion to fix this issue is suggested by Karimi 2007. Here, we will evaluate the Karimi criterion and its performance under different conditions and cases. An empirical study is used to satisfy this aim with different distribution of residual variate. [ABSTRACT FROM AUTHOR]

Time series data is used as a description of the state or nature of variables in the past for forecasting the value of that variable in the future period. Forecasting is certainly very important to use in various fields, one of which is in the economic field. Exports from Central Java show an increasing trend, which indicates that the economy of Central Java will continue to grow. The growth of the economy certainly needs to be supported also through sources of funds. One of them is to attract Foreign Investment to invest in Indonesia. Export value needs to be controlled by developing the right strategy. This strategy can be realized if investors, policy makers, and economic actors know the condition of the future export value. Fuzzy Time Series forecasting method is a method without the need to fulfill assumptions as in other methods. The author is interested in comparing the Fuzzy Time Series method that has been developed, namely Ruey Chyn Tsaur and Stevenson Porter using export value data in Central Java in January 2012-February 2022. Forecasting results from the Fuzzy Time Series Ruey Chyn Tsaur method one period ahead in March 2022 amounted to 1052.5 million US$. As for the Fuzzy Time Series, Stevenson Porter has not been able to predict outside the data period, so it must be combined with the DES method from Holt with forecasting results of 823,8476648 million US$. Fuzzy Time Series Ruey Chyn Tsaur method is the best method because it has a smaller MAPE value than the Stevenson Porter method of 6.91% and 11.54%, respectively. [ABSTRACT FROM AUTHOR]

Along with the development of data science, of course, there are also various innovative solutions to support data processing. These solutions often bring with them new data science trends that are rapidly changing the way we work and live our lives. The trend of data science is using a small dataset, which has a smaller volume and is easy to access and process. It is energy-efficient, has lower costs, and can work without an internet connection. Some of the research is to create tools and platforms that anyone can use to build their own Machine Learning (ML) applications. Currently, there are many innovations in big data processing that are rooted in geospatial data. Everyone carries at least one piece of hardware that monitors their movements that generate spatial data. Also, the drone and IoT industries are highly developed and collect spatial data in real-time. So, there is a lot of data available that is rooted in spatial data. The approach in this research will be utilizing the Geospatial tools for data science to work with vector and raster data types and how to manipulate them, implement time series analysis on maps and perform geospatial analysis using vector data. [ABSTRACT FROM AUTHOR]

There have been many attempts to improve how well chaotic time-series can be used to make predictions. Predicting a chaotic time series beyond the short time limit is extremely difficult as chaotic trajectories diverge exponentially. Various approaches, including machine learning techniques, have been attempted in the literature to attain better prediction accuracy. A complex network has recently been suggested as a way to look at how a certain time series changes over time. Here, we propose a technique for forecasting time series based on an induced weighted state transition network. We demonstrate that the suggested approach can successfully predict chaotic time series up to a reasonable time limit. The proposed method is simple and computationally efficient. [ABSTRACT FROM AUTHOR]

Precision agriculture is an ever-growing domain. Fruit yield estimation and forecasting is a necessary step towards precision agriculture. Reliable and accurate estimation of fruit yield in an orchard help the farmers to make suitable arrangements. Fruit yield estimation is performed on video footage in which fruits are detected and tracked in each frame of the input video. YOLO V5 (You Only Look Once Version 5) neural network are applied to detect fruits in an orchard. Bounding boxes are extracted from detection and Non-Maximum Suppression (NMS) is performed to avoid multiple detections. Then the results are input into the tracking pipeline. DeepSORT algorithm is used for tracking the fruits. By keeping track of the fruits throughout the video frames, it is ensured that we are estimating the total fruit count appropriately when they are detected. Future fruit yield prediction is also carried out using time series forecasting by ARIMA model. This work may greatly help the needy farmers in yield estimation and predicting the future yields for sustainable growth. [ABSTRACT FROM AUTHOR]

In this work, the task was set to predict how the indicators of the trading network (sales volume, customer loyalty, etc.) will change due to the use of marketing campaigns-discounts, bonuses, newsletters, etc. A multidimensional time series modeled the values of indicators and the presence of specific stocks. The LSTM recurrent neural network was used as a prediction tool. Although for most indicators, the network showed a reasonably high predictive ability, the magnitude of the forecast error is several orders of magnitude lower than the sample variance of the indicators. This result indicates the possibility of using recurrent neural networks to predict the effect of marketing campaigns. [ABSTRACT FROM AUTHOR]

An approach to obtain autoregressive models for generating synthetic time series in a single point with a pre-defined von Kármán spectrum is presented. Statistical stationarity is assumed. Under the premise of the proposed approach a theoretical expression for the power spectral density of the autoregressive model is used to obtain the regression coefficients. This approach is proposed to avoid the aliasing effect which may appear when the autoregressive model is obtained to reproduce a given target autocovariance function. A genetic algorithm is used to achieve an optimal autoregressive model. This approach is compared to the approach proposed by Gallego-Castillo et al. where the autocovariance function of the von kármán model is employed as a target. The theoretical autoregressive spectrum obtained using the proposed approach is not affected by aliasing effect, however, it can also present overestimation of the spectrum values at high frequencies in certain circumstances explained in the present work. The proposed approach could facilitate the use of autoregressive models to generate synthetic wind velocity time series when the target statistical information is defined in the frequency domain, as it is the usual case with atmospheric turbulence models. [ABSTRACT FROM AUTHOR]

Exponential smoothing models are used to smooth time series data and treat the impact of Trend or Seasonal in the time series. There are several models including of smoothing: Simple Exponential Smoothing, Double Smoothing and Holt-Winters smoothing. Robust statistical methods in time series models or in linear models (Regression, Experiment Design) they are a suitable to treat the problem of outlier values in data in order to obtain more reliable forecast results. In this paper, we applied the Robust Exponential smoothing method Compared with some anther exponential as :(Double smoothing, Holt-Winters Additive/Multiplicative smoothing, EST method) for estimation model parameters and forecasting are presented. and application of all these methods in Data of corona Pandemic (COVID-19) from Iraq in period (Mar. 2020 – Mar 2022), the robust exponential method is presented that applies the standard technique to preclean data from the outlier values in time series.We depended on R-packages functions as [ses(),EST, robiest(), hw(),...],it use to obtain smoothing parameters are robust. In this research we found Robust method have a good forecast performance for time series with an outlier values. [ABSTRACT FROM AUTHOR]

Wind is one of the most important sustainable energy sources because it uses the kinetic energy provided by moving air to generate electrical renewable energy. To generate the electrical energy, the wind turbine is used to collect the wind's kinetic energy. To generate more wind power, the wind turbine needs to be located at the areas which have strong wind speed and air density. Hence, realizing the significant of wind speed, the study of forecasting wind speed needs to be done to determine the best location for wind turbine and what early risk management plan can be created for the convenience of power grid dispatcher. In this study, the Autoregressive Integrated Moving Average (ARIMA) Fourier time series model was used to forecast the wind speed at Senai station from the year 1985 to 2000. In the analysis, ARIMA Fourier model was compared with ARIMA model in terms of in-sample and out-sample measurement errors. The best fitted model will have the lowest measurement errors of Mean Absolute Percentage Error (MAPE) and Root Mean Square Absolute Error (RMSE). The results showed that ARIMA Fourier model outperformed the ARIMA model. Thus, ARIMA Fourier model is selected as the best forecasting model as compared to ARIMA for the management plan of wind turbine's site. [ABSTRACT FROM AUTHOR]

This study compares three different deep learning approaches for forecasting the air quality time series data. All the approaches were based on the convolutional neural network (CNN), which is one of the deep learning algorithms. The first approach, known as the univariate CNN (UCNN), is based solely on the time series past values. Meanwhile, the second approach, known as the multi-parallel CNN (MPCNN), will use the same variable time series from different observations as the input for the model. The last approach is the multivariate CNN (MCNN), where the forecast will incorporate other variables' time series as the input for the model. The time series data used in this study are the hourly particulate matter, PM10 and PM2.5, concentrations from eight air quality monitoring stations in Peninsular Malaysia. The duration of the time series is from July 2017 until June 2019. The results show that the best air quality predictive model is a model which is built using UCNN since it shows the highest accuracy for both time series PM10 and PM2.5. It is also shown that the forecast will have better accuracy if shorter past values are used as the input for the model. [ABSTRACT FROM AUTHOR]

The prevalence of gaps and outliers within datasets presents substantial challenges, particularly in the realm of time series fore-casting and various other predictive machine learning (ML) tasks. This paper, introduces an effective technique for correcting gaps and outliers in data and validates the approach by applying it to datasets with outlier zones drawn from three diverse contexts. This innovative technique holds promising potential to enhance the performance of machine learning models by treating the data to alleviate the complications posed by these issues and in doing so contributes a valuable tool to the data science toolbox. [ABSTRACT FROM AUTHOR]