Demand Forecasting in SCM for Automobile Industry based on Time-series Modeling

In supply chain management (SCM), product classification and demand forecasting are crucial pillars to ensure companies to have production in the right category and quantity for long-term profitability. Due to COVID-19 from 2019, the automobile industry has been seriously negatively affected as the demand dropped dramatically. Therefore, it is necessary to make reasonable product classification and accurate demand forecasting to facilitate automobile companies in SCM to reduce unpopular product manufacture and unnecessary storage costs. In this paper, the Canada automobile market has been chosen with the period from 1946 to 2022. To classify a number of different types of motor vehicles into several categories with general characteristics, K-means Clustering method is applied. With the seasonal patterns and random generated features for auto sales, the time series models ARIMA and SARIMA are adopted for demand forecasting. According to the analysis, the automobiles fitting in the category with high demand and low price are valuable for further production. In addition, SARIMA Model is more accurate and fits better than ARIMA Model for both the training and test datasets for long-term prediction. The classification and forecasting results shed light on guiding manufacturers to adjust production schemes and ensuring auto dealers to predict more accurate sales in order to optimize the strategic planning. © 2023 SPIE.

Impact of COVID-19 Pandemic on Cultural Products Interests

The COVID-19 pandemic has had a significant impact on human behaviors and how it influenced peoples' interests in cultural products is an unsolved problem. While prior studies mostly adopt subjective surveys to find an answer, these methods are always suffering from high cost, limited size, and subjective bias. Inspired by the rich user-oriented data over the Internet, this work explores the possibility to leverage users' search logs to reflect humans' underlying cultural product interests. To further examine how the COVID-19 mobility policy might influence cultural interest changes, we propose a new regression discontinuity design that has the additional potential to predict the recovery phase of peoples' cultural product interests. By analyzing the 1592 search interest time series in 6 countries, we found different patterns of change in interest in movies, music, and art during the COVID-19 pandemic, but a clear overall incremental increase. Across the six countries we studied, we found that changes in interest in cultural products were found to be strongly correlated with mobility and that as mobility declined, interest in movies, music, and art increased by an average of 35, 27 and 20, respectively, with these changes lasting at least eight weeks. © 2023 ACM.

A COVID-19 Infection Prediction Model in Egypt Based on Deep Learning Using Population Mobility Reports

The rapidly spreading COVID-19 disease had already infected more than 190 countries. As a result of this scenario, nations everywhere monitored confirmed cases of infection, cures, and fatalities and made predictions about what the future would hold. In the event of a pandemic, governments had set limit rules for the spread of the virus and save lives. Multiple computer methods existed for forecasting epidemic time series. Deep learning was one of the most promising methods for time-series prediction. In this research, we propose a model for predicting the spread of COVID-19 in Egypt based on deep learning sequence-to-sequence regression, which makes use of data on the population mobility reports. The presented model utilized a new combined dataset from two different sources. The first source is Google population mobility reports, and the second source is the number of infected cases reported daily "world in data” website. The suggested model could predict new cases of COVID-19 infection within 3–7 days with the least amount of prediction error. The proposed model achieved 96.69% accuracy for 3 days of prediction. This study is noteworthy since it is one of the first trials to estimate the daily influx of new COVID-19 infections using population mobility data instead of daily infection rates. © 2023, The Author(s).

U.S. Public Transportation Ridership Analysis and Prediction based on COVID-19

In this paper, a research was conducted to analyse and predict the impacts of COVID-19 on public transportation ridership in the U.S. and 5 most populous cities of the U.S. (New York City, Los Angeles, Chicago, Houston, Philadelphia). The paper aims to exploit the correlation between COVID-19 and public transportation ridership in the U.S. and make the reasonable prediction by machine learning models, including ARIMA and Prophet, to help the local governments improve the rationality of their policy implementation. After correlation analyses, high level of significant and negative correlations between monthly growth rate of COVID-19 infections and monthly growth rate of public transportation ridership are decidedly validated in the total U.S., and New York City, Los Angeles, Chicago, Philadelphia, except Houston. To analyse the errors of Houston, we consult the literature and made a discussion of Influencing factors. We find that the level of public transportation in quantity and utilization is terribly low in Houston. In addition, the factors, such as the lack of planning law and estimation of urban expressways, the high level of citizens' dependence on private cars and pride of owning cars play a considerable roll in the errors. And the impacts can be predicted to a certain extent through two forecasting models (ARIMA and Prophet), although the precision of our models is not enough to make a precise forecast due to the limitations of model tuning and model design. According to the comparison of the two models, ARIMA models' forecasting accuracy is between 6% and 10%, and Prophet's forecasting accuracy is between 8%-12%, depending on the city. Since the insufficient stationarity, periodicity, seasonality of time series, the Prophet models are hard be more refined. © 2023 SPIE.

Multivariate count time series segmentation with "sums and shares” and Poisson lognormal mixture models: a comparative study using pedestrian flows within a multimodal transport hub

This paper deals with a clustering approach based on mixture models to analyze multidimensional mobility count time-series data within a multimodal transport hub. These time series are very likely to evolve depending on various periods characterized by strikes, maintenance works, or health measures against the Covid19 pandemic. In addition, exogenous one-off factors, such as concerts and transport disruptions, can also impact mobility. Our approach flexibly detects time segments within which the very noisy count data is synthesized into regular spatio-temporal mobility profiles. At the upper level of the modeling, evolving mixing weights are designed to detect segments properly. At the lower level, segment-specific count regression models take into account correlations between series and overdispersion as well as the impact of exogenous factors. For this purpose, we set up and compare two promising strategies that can address this issue, namely the "sums and shares” and "Poisson log-normal” models. The proposed methodologies are applied to actual data collected within a multimodal transport hub in the Paris region. Ticketing logs and pedestrian counts provided by stereo cameras are considered here. Experiments are carried out to show the ability of the statistical models to highlight mobility patterns within the transport hub. One model is chosen based on its ability to detect the most continuous segments possible while fitting the count time series well. An in-depth analysis of the time segmentation, mobility patterns, and impact of exogenous factors obtained with the chosen model is finally performed. © 2023, Springer-Verlag GmbH Germany, part of Springer Nature.

Forecasting of COVID-19 Pandemic Using ARIMA and Fb-Prophet Models: UK Case Study

This study aims to provide insights into predicting future cases of COVID-19 infection and rates of virus transmission in the UK by critically analyzing and visualizing historical COVID-19 data, so that healthcare providers can prepare ahead of time. In order to achieve this goal, the study invested in the existing studies and selected ARIMA and Fb-Prophet time series models as the methods to predict confirmed and death cases in the following year. In a comparison of both models using values of their evaluation metrics, root-mean-square error, mean absolute error and mean absolute percentage error show that ARIMA performs better than Fb-Prophet. The study also discusses the reasons for the dramatic spike in mortality and the large drop in deaths shown in the results, contributing to the literature on health analytics and COVID-19 by validating the results of related studies. Copyright © 2023 by SCITEPRESS - Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

Generalized Weierstrass-Mandelbrot with Disturbance for Big Data Applications in Economic and Financial Systems

Disturbance such as COVID-19, pollution or policy variation to the economic and financial system has significant effect in the big data applications. Hence to study the effect of the disturbance on the related time series plays important role in further applying the big data in economic and financial system. Generalized Weierstrass-Mandelbrot Function is presented to study the complexity of the related time series theoretically and simultaneously. The results show that the disturbance indicated as the exponential form can generate multifractal features for the related time series. And the irregularity and long memory are also simulated by this model and described by the R/S method and multifractal analysis. © 2023 IEEE.

Multitask Learning Neural Networks for Pandemic Prediction with Public Stance Enhancement

State and local governments have imposed health policies to contain the spread of COVID-19 since it had a serious impact on human daily life. However, the public stance on these measures may be time-varying. It is likely to escalate the infection in the area where the public is negative or resistant. To take advantage of the correlation between public stance on health policies and the COVID-19 statistics, we propose a novel framework, Multitask Learning Neural Networks for Pandemic Prediction with Public Stance Enhancement (MP3), which is composed of three modules: (1) Stance awareness module to make stance detection on health policies from users' tweets in social media and convert them into a stance time series. (2) Temporal feature extraction module that applies Convolution Neural Network and Recurrent Neural Network to extract and fuse local patterns and long-term correlations from COVID-19 statistics. Moreover, a Stance Latency-aware Attention is proposed to capture dynamic social effects and fuse them with temporal features. (3) Multi-task prediction module to adopt Graph Convolution Network to model the spread of pandemic and employ multi-task learning to simultaneously predict COVID-19 statistics and the trend of public stance on health policies. The proposed framework outperforms state-of-the-art baselines on both confirmed cases and deaths prediction tasks. © 2022 IEEE.

On the efficiency of the gold returns: An econometric exploration for India, USA and Brazil

As a precious metal and investment commodity, gold has been signified to be important for risk management, diversification, and hedging. The gold market has undergone considerable structural changes in the facet of the pandemic and other geopolitical developments, attracting the interest of investors. Thus, it is crucial to look into how these structural changes affect the efficiency of the market. Accordingly, the study examines and compares the evolution of the gold market efficiency in three major economies from January 1, 2018, to August 31, 2022: India, USA, and Brazil. For this, we first decompose the time series using Loess Smoother's Seasonal and Trend Decomposition and then employ a multifractal detrended fluctuation analysis approach. The estimates are strengthened by the alternative approach of the rolling window method of wild bootstrap automatic variance ratio. The findings indicate a considerable decline in the efficiency of the gold returns across three economies, with the highest decline in India, followed by USA and Brazil. Notably, during covid and post covid periods, India and USA show persistence in small fluctuations, while Brazil displays persistent behavior in large fluctuations. Thereby, the market panic makes the gold market unstable, and its use as a safe haven is "erratic”. © 2023 Elsevier Ltd

Predictive Analytics of Stock Market as a Time Series

This paper presents a new methodology and a comparative study using past stock market data that can help businesses take investing or divesting decisions in critical situations in the future. These may be like the COVID-19 pandemic, where market volatility is extremely high, thus creating an urgent need for better decision support systems to minimise loss and ensure better profits. The results of the study are based on the comparison of different configurations of ARIMAX, Prophet, LSTM and Bidirectional LSTM Models trained on historical NSE data. By understanding the correlation and variations in the data processing and model training parameters, we have successfully proposed a LSTM neural network model training and optimising method which could successfully help businesses take both long and short term profitable decisions before and after big financial and market crises with a respective accuracy of 98.60 percent and 96.97 percent. © 2022 IEEE.

Identifying, exploring, and interpreting time series shapes in multivariate time intervals

We introduce a concept of episode referring to a time interval in the development of a dynamic phenomenon that is characterized by multiple time-variant attributes. A data structure representing a single episode is a multivariate time series. To analyse collections of episodes, we propose an approach that is based on recognition of particular patterns in the temporal variation of the variables within episodes. Each episode is thus represented by a combination of patterns. Using this representation, we apply visual analytics techniques to fulfil a set of analysis tasks, such as investigation of the temporal distribution of the patterns, frequencies of transitions between the patterns in episode sequences, and co-occurrences of patterns of different variables within same episodes. We demonstrate our approach on two examples using real-world data, namely, dynamics of human mobility indicators during the COVID-19 pandemic and characteristics of football team movements during episodes of ball turnover. © 2023 The Author(s)

When a Pandemic Enters the Game: The Initial and Prolonged Impact of the COVID-19 Pandemic on Live-Stream Broadcasters on Twitch

The global COVID-19 pandemic has accelerated the popularity of video games and online-gaming platforms. However, little research is devoted to understanding how the pandemic has affected gamers, especially live-stream broadcasters. Therefore, our study aimed to evaluate the impact the COVID-19 pandemic has had on established streamers on Twitch. By using a longitudinal time-series design and focusing on a large sample (N = 23,019) of broadcasters, we were able to determine the initial as well as prolonged effects of the pandemic on their streaming behavior. Our results suggest that the pandemic was a”game changer” for the target group, especially in regard to their choice of game settings and their focus on non-gaming content. Relating the data obtained from the target group of established streamers to the general platform data, we discuss the pandemic-related platform dynamics. © 2022 IEEE Computer Society. All rights reserved.

Predicting COVID-19 Case Counts using Twitter Image Data

A crucial task with diseases, such as COVID-19, is accurate forecasting of cases for early detection of spikes, which allows policymakers to adjust local restrictions. The use of face masks to prevent disease spread among the general population has become widespread due to the COVID-19 pandemic. While predictive models for COVID-19 case counts exist, capturing localized information about mask usage has the potential to improve prediction accuracy. In this paper, we develop time series models that utilize Twitter image data for COVID-19 case count prediction. A crucial part of such a model is the accurate detection of face mask presence in Twitter images, which we train a convolutional neural network (CNN) to perform. While multiple datasets exist to train CNNs for face mask detection, existing datasets do not adequately represent the complexity nor the diversity in social media images. To address this and create a sufficiently accurate CNN for use with social media images, we also present a new social media face mask image dataset designed for the training of CNNs to detect the presence of face masks in complex real-world images, such as social media images. The presented dataset consists of approximately 120k images and attempts to more adequately account for diversity in ethnicity, mask type, and physical orientation of individuals in images than existing datasets. We demonstrate the effectiveness of both the CNN model for face mask detection and the resulting time series model trained on data obtained from applying the CNN model to historical twitter data, illustrating that data on the presence of masks in social media images can increase predictive accuracy of time series models for COVID-19 case counts. © 2022 IEEE.

COVID-19 Prediction Using Time Series Models

Real-time data has evolved to become an integral part of understanding events across different timelines. Machine Learning uses different varieties of algorithms to determine the relationship between sets of data spread across timelines, visualize the current situation, and forecast the future, which is the most important aspect. Due to the breakout of COVID-19, a novel coronavirus, the entire planet is currently experiencing a disastrous crisis. At this time, the SARS-CoV-2 virus has proven to be a possible hazard to human life. The ARIMA Model i.e., Autoregressive Integrated Moving Average is compared with Facebook's Prophet and VARMAX model to foretell the future. The dataset is divided into the training and testing set. The size of the COVID-19 dataset is relatively small as it is a pandemic that occurred recently, due to which much of the data is used for training purposes and the last twelve days have been used for testing and validating the model. The model is trained and fits on the training data set. The algorithms are now ready to anticipate future forecasts after it has been tested and trained. The models also record the predicted and actual values, allowing them to improve their accuracy in the future. In this paper, the results of the ARIMA model are compared against Prophet and VARMAX which are other popular machine learning time series models. For the ease of visualization of covid trends, a dashboard is built using Python's Plotly and Dash and has been deployed using Voila. © 2022 IEEE.

A Fast Initial Response Approach to Sequential Financial Surveillance

Consider the problem of financial surveillance of a heavy-tailed time series modeled as a geometric random walk with log-Student's t increments assuming a constant volatility. Our proposed sequential testing method is based on applying the recently developed taut string (TS) univariate process monitoring scheme to the gaussianized log-differenced process data. With the signal process given by a properly scaled total variation norm of the nonparametric taut string estimator applied to the gaussianized log-differences, the change point detection procedure is constructed to have a desired in-control (IC) average run length (ARL) assuming no change in the process drift. If a change in the process drift is imminent, the proposed approach offers an effective fast initial response (FIR) instrument for rapid yet reliable change point detection. This framework may be particularly advantageous for protection against imminent upsets in financial time series in a turbulent socioeconomic and/or political environment. We illustrate how the proposed approach can be applied to sequential surveillance of real-world financial data originating from Meta Platforms, Inc. (FB) stock prices and compare the performance of the TS chart to that of the more prominent CUSUM and CUSUM FIR charts at flagging the COVID-19 related crash of February 2020. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

A Prospective Approach on Covid-19 Forecasting Using LSTM

The novel Corona virus has been proclaimed as a worldwide pandemic through World Health Organization in the March 2020 has immensely affected the world with its ferocity. By observation, the scientists got to know that it transmits from one human to other by droplets which range from larger respiratory droplets to smaller aerosols or direct contact with an infected person. Its impurity has been assessed to have an incubation time of 6.4 days than a simple reproduction amount of 2.24-3.58.[19] The transmission rate and spread of infection is quite rapid as compared to other fatal viral infections encountered till date. A massive loss of human life was faced even by the developed countries which had the best health-care facilities. According to WHO, COVID-19 has been confirmed in 238,521,855 people over the world, with 4,863,818 deaths as of October 9th, 2021. After experiencing the second covid wave, the number of cases had got dropped drastically but the increase in their number in the recent days is a major cause of concern. This stresses us to build some prediction models which could help in providing relief to the virus-prone areas. In this study, we are using time series for predicting forthcoming cases of corona virus. © 2022 IEEE.

Time Series: Predicting Nigerian Food Prices using ARIMA Model and R-Programming

The majority of food commodities in Nigeria have seen persistent price instability. this is brought by elements like insecurity/insurgency, poor storage facilities, seasonal price changes, inconsistent government policies, COVID-19 containment measures, poor access to credit, technical inputs, lack of modern farm tools and implements. This study focused on comparing the prices of four different food items - beans, onion, tomato, and yam using the ARIMA model to forecast future prices. Two out of the six geopolitical zones of Nigeria were used for the study;the North-Central and North-West. The National Bureau of Statistics (NBS) provided the raw data between 2017 and 2018, and the items were weighed in kilograms (Kg). The data was extrapolated into a time series data by executing in R Studio. The stationarity of the series data was obtained by a Unit root Test using the KPSS test (If p<0.05 means the time series is stationary). Results from the forecasted values indicated that food commodities' prices increase with time, making ARIMA a good model for forecasting prices. It was recommended that necessary measures should be put in place to ameliorate the high cost of food prices being experienced in the country of Nigeria. © 2022 IEEE.

Nonparametric Homogeneity Test for Small Samples and Its Applications to COVID-19 Epidemic Curves Forecasting

The COVID-19 pandemic is accompanied by intensive attempts to build mathematical models to predict it. For this, various models are used, both traditional differential equations and machine learning models. Classical epidemiological compartment models contain parameters that are difficult to measure. Their results are used to model various scenarios, but it is difficult to obtain a reliable forecast with their help. Machine learning models, on the other hand, do not use prior assumptions, and their inferences are based only on training samples. This usually results in more reliable forecasts. In both the first and second cases, it is necessary not only to estimate the forecast error, but to compare the prediction accuracy of different models by checking the error homogeneity also. An additional factor complicating the problem is the small size of available samples in some cases. This forces one to resort to resampling methods. The article describes the Klyushin-Petunin test for testing the homogeneity of samples with ties in a multi-sample design and compares it with the traditional Anderson-Darling, Kruskal-Wallis and Friedman tests using the example of three methods for predicting the COVID-19 epidemic in the basis of epidemic data in Germany, Japan, South Korea and Ukraine. © 2022 IEEE.

Survival-Based Feature Extraction - Application in Supply Management for Dispersed Vending Machines

The outbreak of the COVID pandemic revealed that supply chains are not resilient to such a type of turmoil, and the food industry appeared to be particularly vulnerable. Meanwhile, customers expect uninterrupted deliveries and the products' selection responding to their preferences. In this article, we discuss several topics related to supply management that allow preparing a delivery plan for a distributed network of vending machines, considering each location individually. The developed solution takes advantage of the state-of-the-art machine learning methods. However, it is human-centric and aligned with the concept of Industry 5.0. We present the conceptual and technological side of the solution with a particular emphasis on the developed feature extraction framework, which uses selected indicators from the survival analysis. We present an analysis of the real data confirming that the proposed approach copes well with high uncertainty in data, addressing the cold-start problem. © 2005-2012 IEEE.

Forecasting Combination of Hierarchical Time Series: A Novel Method with an Application to CoVid-19

Multiple, hierarchically organized time series are routinely submitted to the forecaster upon request to provide estimates of their future values, regardless the level occupied in the hierarchy. In this paper, a novel method for the prediction of hierarchically structured time series will be presented. The idea is to enhance the quality of the predictions obtained using a technique of the type forecast reconciliation, by applying this procedure to a set of optimally combined predictions, generated by different statistical models. The goodness of the proposed method will be evaluated using the official time series related to the number of people tested positive to the SARS-CoV-2 in each of the Italian regions, between February 24th 2020 and August 31th 2020. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.