Statistical Forecasting Methods Research Articles

Customer Demand Forecast Using Time Series Approach

Nowadays we are living in an uncertain time (pandemics, wars, inflation, shortages) and for companies, predicting the future of the business becomes more and more difficult. From this point of view, we can resort to statistical forecasting methods that, based on on past data and current external trends, can predict figures for the short and long term planning. The motive, the scope, behind this study is to develop a forecasting model using time series that can help the organisations and the managers in taking better decisions, on time, for the future of the company. The data that were used for this paper are focusing on forecasting the demands for the sales budgets only, starting with the historical data from a company in the glass mould manufacturing area. If we examine the current literature, we can find that there a few studies, with real business data regarding budgeting, therefore this study is relevant and important for the researchers and practitioners. This paper aims to illustrate how we can use the time series statistical method and the linear equation regression that can help the organisations to forecast and plan the business. The main findings after developing the model are that it allows smoothing the fluctuations of the series over time (especially if we find outliers) and eliminating the influence of seasonality, in order to obain in the end a good accuracy of the predictions. In order to measure the quality of the forecasting we can use many indicators, such as Mean Absolute Percentage Error.

A fusion of machine learning algorithms and traditional statistical forecasting models for analyzing American healthcare expenditure

The American healthcare system allocates considerable resources compared to peer-developed nations. However, outcomes significantly trail behind, particularly in life expectancy. This study addresses questions about the enduring trends in healthcare spending as a percentage of Gross Domestic Product (GDP), notable factors contributing to this concerning trend, and the timing to apply an emergency brake to curb this accelerating trajectory. Advanced machine learning algorithms, such as Random Forest and Support Vector Regression (SVR), in conjunction with traditional statistical forecasting methods, are used to forecast future patterns. The research underscores the importance of healthcare analytics in unraveling the intricacies of the healthcare system. The findings highlight the pressing need for effective policies to confront this mounting challenge.

Статистические методы прогнозирования правовых явлений

The article reveals the issues of forecasting legal phenomena, which helps to develop legislation, i.e. to anticipate potential changes in society and adapt legal norms to changing circumstances. Statistical forecasting methods play a key role in modern legal society. They not only contribute to improving the effectiveness of law enforcement, but also help to prevent crime and improve the management of legal processes. Thanks to statistical analysis, it is possible to make our society safer and more secure, preventing possible threats and negative consequences for citizens.

STATISTICAL METHODS FOR PREDICTING PHISHING ATTACKS

The article proposes a methodology for predicting so-called phishing attacks, which are a common form of cybercrime, the number of which is growing every year, and the level of their harmful impact on the information systems of critical infrastructure objects is also increasing. To analyze trends and predict phishing attacks, we used statistical data published in scientific works of domestic and foreign researchers, as well as published by online publications of leading consulting companies working in the field of information security and cybersecurity. Statistical methods based on the use of time series, as one of the popular approaches used to predict various technological and economic processes, were chosen as tools for researching and predicting phishing attacks. This made it possible to analyze the types and patterns of phishing attacks that attackers use to disrupt the operation of software of information and communication systems and automated systems. Based on time series analysis, a trend model was built for the number of detected phishing attacks for the period 2020–2023. A calculation was made of the predicted number of phishing attacks for 16 quarters of 2020–2023, as well as the estimated forecast of the occurrence of these attacks for four quarters of 2024. To improve the forecast, a coefficient taking into account the seasonality factor was calculated and a correlation and regression analysis of the impact of phishing attacks on the total number of attacks detected during 2020–2023 was performed. Calculations have been performed, indicating that the discrepancies in the predicted values are not significant; the results presented allow us to select the optimal strategy for identifying, predicting and eliminating computer attacks related to phishing. Based on the time series model and the calculations obtained, it was concluded that statistical forecasting methods make it possible to build a forecast of phishing attacks, provide in the future the opportunity to develop and formulate methods for countering these attacks, and plan measures to increase the level of security of information resources.

Advancements in AI-Enhanced Just-In-Time Inventory: Elevating Demand Forecasting Accuracy

The landscape of global supply chains is undergoing a transformation characterized by increasing complexity and the need for agile response to demand fluctuations. At the forefront of addressing these challenges is the strategic alignment of inventory levels with volatile demand, a task that has become more sophisticated within the Just-In-Time (JIT) inventory management paradigm. This paper explores the integration of Artificial Intelligence (AI) into JIT systems, with a focus on enhancing the accuracy of demand forecasting—a critical component for inventory optimization. We present a groundbreaking hybrid AI model that combines the forward-thinking capabilities of neural networks with the reliability of classical statistical forecasting methods. This innovative approach seeks to elevate forecasting reliability and reduce errors that lead to either inventory shortfalls or surpluses. Through a series of case studies spanning various industries, the versatility and effectiveness of the hybrid model are demonstrated, highlighting improvements in forecasting accuracy and the resulting operational benefits.

A Novel First-Order Fuzzy Rules-Based Forecasting System Using Distance Measures Approach for Financial Market Forecasting

The precise estimates about finance, atmospheric science, power sector, industries, agriculture, and other science help governments and institutions economically in making the relevant policies regarding import-export, demand, consumption, storage, and local industries. Due to the uncertainty and nondeterministic behavior of data series with respect to time, the foremost challenge is to develop and identify the practical method to handle the above stated complex issues. As an illustration, this study presented an analysis of a new fuzzy time-series (FTS) approach and comparison with traditional forecasting models for prediction of gram pulse production. Taking into consideration the theory of fuzzy sets, FTS, fuzzy rules, triangular membership functions, distance measures, and modified weighted average method, a robust and effective fuzzy rules-based methodology was developed for the prediction of time-series data regarding crop production and share prices. Conventional statistical forecasting methods such as Holt’s linear trend, Holt’s exponential trend, and Holt’s damped exponential trend models were also applied on time-series data for comparison. To identify the primacy of modeling and forecasting, the techniques of root mean squared error (RMSE) and mean absolute error (MAE) were used as a criterion. The numerical values of RMSE and MAE such as 106.51 and 74.8897 clearly demonstrated that the proposed fuzzy rules-based method is robust for forecasting of production and market share prices in the environment of uncertainty.

A new hybrid multivariate grey model based on genetic algorithms optimization and its application in forecasting oil products demand

PurposeConventional statistical forecasting methods typically need a large sample size or the use of overly confident hypotheses, like the Gaussian distribution of the input data. Unfortunately, these input data are frequently scarce or do no not follow a normal distribution law. A grey forecasting model can be developed and used to predict energy consumption for at least four data points or ambiguous data based on grey theory. The standard grey model, however, may occasionally result in significant forecasting errors.Design/methodology/approachIn order to reduce these errors, this paper proposes a hybrid multivariate grey model (namely Grey Modeling (1,N)) optimized by Genetic Algorithms with sequential selection forecasting mechanism, abbreviated as Sequential-GMGA(1,N). A real case of Cameroon's oil products consumption is considered to demonstrate the effectiveness of the proposed forecasting model.FindingsThe results show the superiority of Sequential-GMGA(1,4) when compared with the results of competing grey models reported in the literature, with a mean absolute percentage error as low as 0.02%.Originality/valueWithout changing the model's basic structure, the suggested framework completely extracts the evolution law of multivariate time series. Regardless of data patterns, Sequential-GMGA(1,4) actively enhances all model parameters over the course of each predicted timeframe. Consequently, Sequential-GMGA(1,4) improves forecast accuracy by resolving the discrepancy between the model's least sum of squares of prediction errors and the parameterization approach based on grey derivative.

Dynamic Time Scan Forecasting: A Benchmark With M4 Competition Data

Univariate forecasting methods are fundamental for many different application areas. M-competitions provide important benchmarks for scientists, researchers, statisticians, and engineers in the field, for evaluating and guiding the development of new forecasting techniques. In this paper, the Dynamic Time Scan Forecasting (DTSF), a new univariate forecasting method based on scan statistics, is presented. DTSF scans an entire time series, identifies past patterns whichare similar to the last available observations and forecasts based on the median of the subsequent observations of the most similar windows in past. In order to evaluate the performance of this method, a comparison with other statistical forecasting methods, applied in the M4 competition, is provided. In the hourly domain, an average sMAPE of 12.9% was achieved using hte method with the default parameters, while the baseline competition - the simple average of the forecasts of Holt, Damped and Theta methods - was 22.1%. The method proved to be competitive in longer time series, with high repeatability.

Development of a neural network for diagnosing the risk of depression according to the experimental data of the stop signal paradigm.

These days, the ability to predict the result of the development of the system is the guarantee of the successful functioning of the system. Improving the quality and volume of information, complicating its presentation, the need to detect hidden connections makes it ineffective, and most often impossible, to use classical statistical forecasting methods. Among the various forecasting methods, methods based on the use of artificial neural networks occupy a special place. The main objective of our work is to create a neural network that predicts the risk of depression in a person using data obtained using a motor control performance testing system. The stop-signal paradigm (SSP) is an experimental technique to assess a person's ability to activate deliberate movements or inhibit movements that have become inadequate to external conditions. In modern medicine, the SSP is most commonly used to diagnose movement disorders such as Parkinson's disease or the effects of stroke. We hypothesized that SSP could serve as a basis for detecting the risk of affective diseases, including depression. The neural network we are developing is supposed to combine such behavioral indicators as: the amount of missed responses, amount of correct responses, average time, the amount of correct inhibition of movements after stop-signal onset. Such a combination of indicators will provide increased accuracy in predicting the presence of depression in a person. The artificial neural network implemented in the work allows diagnosing the risk of depression on the basis of the data obtained in the stop-signal task. An architecture was developed and a system was implemented for testing motor control indicators in humans, then it was tested in real experiments. A comparison of neural network technologies and methods of mathematical statistics was carried out. A neural network was implemented to diagnose the risk of depression using stop-signal paradigm data. The efficiency of the neural network (in terms of accuracy) was demonstrated on data with an expert assessment for the presence of depression and data from the motor control testing system.

A Basic Time Series Forecasting Course with Python

The aim of this paper is to present a set of Python-based tools to develop forecasts using time series data sets. The material is based on a 4-week course that the author has taught for 7 years to students on operations research, management science, analytics, and statistics 1-year MSc programmes. However, it can easily be adapted to various other audiences, including executive management or some undergraduate programmes. No particular knowledge of Python is required to use this material. Nevertheless, we assume a good level of familiarity with standard statistical forecasting methods such as exponential smoothing, autoregressive integrated moving average (ARIMA), and regression-based techniques, which is required to deliver such a course. Access to relevant data, codes, and lecture notes, which serve as based for this material, is made available (see https://github.com/abzemkoho/forecasting) for anyone interested in teaching such a course or developing some familiarity with the mathematical background of relevant methods and tools.

Predictive analytics for demand forecasting: A deep learning-based decision support system

The demand is often forecasted using econometric (regression) or statistical forecasting methods. However, most of these methods lack the ability to model both temporal (linear and nonlinear) and covariates-based variations in a demand series simultaneously. In this context, a novel forecasting model is proposed that combines a state-of-the-art sequence modeling method and a machine learning method in an ensemble model. The proposed model can handle both types of variations in demand data, and thus, enhances forecasts’ accuracy. A big sample of 4235 demand series consisting of structured and unstructured data (could be referred to as “big data”) related to packaged food products is used for experimentation. Data contain point-of-sales, promotion, weather, regional economy, internet media, and economic activity index related variables. Some of these variables and their combinations is probably used for the first time in a demand forecasting model. The forecasting results are evaluated through multiple error metrics (i.e. mean error, mean absolute error, mean squared error), and it has been observed that proposed method outperformed the benchmarking methods. A demand sensing algorithm is also proposed to forecast demand in real-time.

Forecasting the volume of passenger transportation by inland water transport

The article discusses the currently used methods of forecasting passenger traffic volumes. The authors considered both statistical forecasting methods based on the construction of trend lines, and methods that take into account, in addition to the trend values of traffic volume indicators, other factors that affect the forecast values of passenger transport. On the example of water transport, the forecast values of traffic volumes were calculated by various methods, and the analysis of the calculation results was carried out.

Review of automated time series forecasting pipelines

AbstractTime series forecasting is fundamental for various use cases in different domains such as energy systems and economics. Creating a forecasting model for a specific use case requires an iterative and complex design process. The typical design process includes five sections (1) data preprocessing, (2) feature engineering, (3) hyperparameter optimization, (4) forecasting method selection, and (5) forecast ensembling, which are commonly organized in a pipeline structure. One promising approach to handle the ever‐growing demand for time series forecasts is automating this design process. The article, thus, reviews existing literature on automated time series forecasting pipelines and analyzes how the design process of forecasting models is currently automated. Thereby, we consider both automated machine learning (AutoML) and automated statistical forecasting methods in a single forecasting pipeline. For this purpose, we first present and compare the identified automation methods for each pipeline section. Second, we analyze these automation methods regarding their interaction, combination, and coverage of the five pipeline sections. For both, we discuss the reviewed literature that contributes toward automating the design process, identify problems, give recommendations, and suggest future research. This review reveals that the majority of the reviewed literature only covers two or three of the five pipeline sections. We conclude that future research has to holistically consider the automation of the forecasting pipeline to enable the large‐scale application of time series forecasting.This article is categorized under: Technologies > Machine Learning Technologies > Prediction Algorithmic Development > Spatial and Temporal Data Mining

For Better or Worse? Revenue Forecasting with Machine Learning Approaches

The recent rapid development of artificial intelligence (AI) is expected to transform how governments work by enhancing the quality of decision-making. Despite rising expectations and the growing use of AI by governments, scholarly research on AI applications in public administration has lagged. In this study, we fill gaps in the current literature on the application of machine learning (ML) algorithms with a focus on revenue forecasting by local governments. Specifically, we explore how different ML models perform on predicting revenue for local governments and compare the relative performance of revenue forecasting by traditional forecasters and several ML algorithms. Our findings reveal that traditional statistical forecasting methods outperform ML algorithms overall, while one of ML algorithms, KNN, is more effective in predicting property tax revenue. This result is particularly salient for public managers in local governments to handle foreseeable fiscal challenges through more accurate predictions of revenue.

ОVERVIEW OF STATISTICAL METHODS FOR FORECAST DEVELOPMENT

Having a general idea of the nature of the forecast and understanding the general methodology of forecasting in general, it is advisable for analytical departments, economic departments, management units of enterprises, banks, firms (of any socio-economic objects) to develop at least short-term forecasts based on indicators of their activity in modern conditions coopetitions to understand trends in changes in these indicators. As a result of the conducted research, the basic principles that must be observed when developing forecasts are revealed, as well as a detailed description and features of statistical auto-projective forecasting methods are given: random walk models containing a free term or it; models characterizing a deterministic trend with random fluctuations around the trend; moving average models; exponential smoothing models using simple exponential smoothing, linear, quadratic and seasonal smoothing (Brown, Holt, and Winters models); integrated presentation of autoregressive models and moving average models (parametric ARIMA models). In addition, the conducted research confirmed that the significance of the developed forecast and the level of confidence in the obtained future values of the indicators depends on the quality of the developed models. The criteria for checking the quality of the developed forecasts are: the Akaike information criterion, which evaluates the quality of the model compared to each other; the Hannan-Quinn Criterion information criterion, which is used to compare models with a different number of parameters and is an alternative to the Akaike information criterion; the Schwarz-Bayesian information Criterion, which compares the quality of a model relative to each other using a likelihood function; the mean squared error value, the absolute value of the mean squared error and the absolute value of the mean squared error expressed as a percentage.

МОДЕЛИРОВАНИЕ УРОВНЯ БЕДНОСТИ В РФ НА ОСНОВЕ НЕЙРОСЕТЕВОГО ПОДХОДА

The article raises the question of modeling the level of poverty as one of the most important socio-economic indicators. A review of publications by domestic and foreign scientists-economists proves the relevance of the topic chosen for the study. Today, the time series apparatus acts as one of the popular tools for studying the dynamics of the poverty level and the factors that directly influence it, but classical statistical forecasting methods impose rather strict assumptions on the construction of models. The article discusses the possibility of using automated neural networks of the STATISTICA package for analyzing and forecasting a time series composed of annual data reflecting the dynamics of the poverty level in the Russian Federation over the past 20 years. The study took into account the strengths and weaknesses of the use of the neural network apparatus for predicting socio-economic processes. The construction of economic and mathematical models was carried out by building automated neural networks, custom neural networks and the method of multiple sampling. When choosing the most preferable model, a multidimensional criterion was used. The comparison of the real poverty level with the values obtained using the models is made, the quality assessment of the developed models is calculated, the poverty level forecast for 2021-2022 is constructed.

Valuing Enhanced Hydrologic Data and Forecasting for Informing Hydropower Operations

Climate change is rapidly modifying historic river flows and snowpack conditions in the Sierra Nevada in California and other seasonally snow-covered mountains. Statistical forecasting methods based on regressing summer flow against spring snow water equivalent, precipitation, and antecedent runoff are thus becoming increasingly inadequate for water-resources decision making, which can lead to missed opportunities in maximizing beneficial uses, including the value of hydropower resources. An enhanced forecasting method using a process-based model and spatially distributed wireless sensor data offers more accurate runoff forecasts. In this paper, we assessed the forecasting accuracy of these two forecasting methods by applying them to two tributaries within the North Fork Feather River basin in California. The result shows the enhanced forecasting method having better accuracy than the statistical model. In addition, a hydropower simulation showed a considerable increase in energy value with the enhanced forecasting informing reservoir operations. The investment analysis on applying this method shows an average internal rate of return of 31% across all scenarios, making this forecasting method an attractive way to better inform water-related decisions for hydropower generation in the context of climate change.

State and prospects of providing Russian rural areas with engineering life support systems

The article establishes the relationship between the favorable living environment in rural areas and the level of housing equipment with engineering systems. The analysis of the condition of residential premises in rural areas in terms of the level of their provision with life support systems is carried out. The problems of low level and quality of rural housing improvement are identified. The reasons that prevent the formation of comfortable conditions for providing housing with life support systems, namely centralized engineering networks, are identified. Using statistical forecasting methods, an inertial variant of changing the target indicators of residential improvement in rural areas of Russia by types of life support systems, taking into account the existing level, is created.

МОНИТОРИНГ, МОДЕЛИРОВАНИЕ И АНАЛИЗ ПОВЕДЕНИЯ КОНСТРУКЦИЙ

Information technology and modeling are becoming more widely used, mainly in the design and operation of buildings and structures, and in most cases this is enough for trouble-free operation. Nev-ertheless, there is a category of buildings for which the monitoring of the technical condition should be an integral part of the construction and operation. Unfortunately, the development of these technolo-gies in the Russian Federation is not at level, sufficient for answering questions about the behavior of objects under changing environmental conditions and revealing hidden patterns in monitoring data. Based on analysis of literary sources, the authors reviewed various methods for identifying hidden patterns in geodetic measurement data when monitoring buildings and structures. It is noted that mod-ern analysis methods are based on statistical processing of measurement results and on statistical fore-casting method. However, there are attempts to apply models that take into account object’s design features and temperature regime. This type includes two proposed models, which are used to model three-dimensional coordinates of strain marks in the 3D model and only elevations of marks in the 1–Z model. The article presents the rationale for the simulated geometric elements and properties of the object. The solution of the equations of both models and the analysis of the results and parameters of the model for measurement epochs are shown. The simulation is shown on the example of a real ob-ject, which was monitored by the authors in 2015–2016.The authors believe that the monitoring of large-span structures and the search for patterns of their behavior should be an integral part of the in-formation system for such structures.

Manipulator Detection in Cryptocurrency Markets Based on Forecasting Anomalies

Today, there are constant changes in terms of securities in stock markets. In these stock market investments, investors use fundamental analysis tools and indicators very widely. In this way, it is possible to have some knowledge of the situations experienced in the markets and to make a profit. In this study, manipulations on Bitcoin are discussed. Popular machine and statistical forecasting methods have been used to detect these manipulations and the road maps to be followed in order to be detected in the most successful way have been shared. Social media sentiments, which were thought to have an effect on manipulations during the studies, were also evaluated with the most advanced text analysis methods and evaluated together with these price changes. The allegations that the prediction methods carried out before the crisis were more successful were investigated. The Covid-19 pandemic was evaluated as a period of global crisis and the studies that might be relevant were examined. It would not be wrong to say that the actors that make big gains in the stock markets are the ones that determine the direction of the stock market. The manipulation periods of the market actors to be successful in the virtual money markets have been tried to be verified by various estimation methods. These estimations can achieve up to F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> score of 93% success according to our experimental result. Besides, it is stated that accounts with the highest volume of transactions in the periods, when anomalies were detected, were labeled as potential manipulators.