EMPLOYING ARTIFICIAL NEURAL NETWORKS AND MULTIPLE DISCRIMINANT ANALYSIS TO EVALUATE THE IMPACT OF THE COVID-19 PANDEMIC ON THE FINANCIAL STATUS OF JORDANIAN COMPANIES

Aim/Purpose This paper aims to empirically quantify the financial distress caused by the COVID-19 pandemic on companies listed on Amman Stock Exchange (ASE). The paper also aims to identify the most important predictors of financial distress pre- and mid-pandemic. Background The COVID-19 pandemic has had a huge toll, not only on human lives but also on many businesses. This provided the impetus to assess the impact of the pandemic on the financial status of Jordanian companies. Methodology The initial sample comprised 165 companies, which was cleansed and reduced to 84 companies as per data availability. Financial data pertaining to the 84 companies were collected over a two-year period, 2019 and 2020, to empirically quantify the impact of the pandemic on companies in the dataset. Two approaches were employed. The first approach involved using Multiple Discriminant Analysis (MDA) based on Altman's (1968) model to obtain the Z-score of each company over the investigation period. The second approach involved developing models using Artificial Neural Networks (ANNs) with 15 standard financial ratios to find out the most important variables in predicting financial distress and create an accurate Financial Distress Prediction (FDP) model. Contribution This research contributes by providing a better understanding of how financial distress predictors perform during dynamic and risky times. The research confirmed that in spite of the negative impact of COVID-19 on the financial health of companies, the main predictors of financial distress remained relatively steadfast. This indicates that standard financial distress predictors can be regarded as being impervious to extraneous financial and/or health calamities. Findings Results using MDA indicated that more than 63% of companies in the dataset have a lower Z-score in 2020 when compared to 2019. There was also an 8% increase in distressed companies in 2020, and around 6% of companies came to be no longer healthy. As for the models built using ANNs, results show that the most important variable in predicting financial distress is the Return on Capital. The predictive accuracy for the 2019 and 2020 models measured using the area under the Receiver Operating Characteristic (ROC) graph was 87.5% and 97.6%, respectively. Recommendations Decision makers and top management are encouraged to focus on the identified for Practitioners highly liquid ratios to make thoughtful decisions and initiate preemptive actions to avoid organizational failure. Recommendations This research can be considered a stepping stone to investigating the impact of for Researchers COVID-19 on the financial status of companies. Researchers are recommended to replicate the methods used in this research across various business sectors to understand the financial dynamics of companies during uncertain times. Impact on Society Stakeholders in Jordanian-listed companies should concentrate on the list of most important predictors of financial distress as presented in this study. Future Research Future research may focus on expanding the scope of this study by including other geographical locations to check for the generalisability of the results. Future research may also include post-COVID-19 data to check for changes in results. © 2023 Informing Science Institute. All rights reserved.

A Novel Machine Learning-assisted Pairs Trading Approach for Trading Risk Reduction

The cryptocurrency market has been growing rapidly in recent years. The volume of transactions and the number of participants in the cryptocurrency market makes it huge enough that we cannot ignore it. At the same time, the global stock market has also reached a new height in the past two years. However, due to the COVID epidemic and other political and economic-related factors in the last two years, the uncertainty in the capital market remains high, and short-term large fluctuations occur frequently;thus, many investors have suffered substantial losses. Pairs trading, an advanced statistical arbitrage method, is believed to hedge the risk and profit off the market regardless of market condition. Amongst the vast literature on pairs trading, there have been investors trading a pair of cryptocurrencies or a pair of stocks using machine learning or empirical methods. This research probes the boundary of utilizing machine learning methods to do pairs trading with one stock asset and another cryptocurrency. Briefly, we built an assets pool with both stocks and cryptocurrencies to find the best trading pair. In addition, we applied mainstream machine learning models to the trading strategy. We finally evaluated the accuracy of the proposed method in prediction and compared their returns based on the actual U.S. Stock and Cryptocurrency Market data. The test results show that our method outperforms other state-of-the-art methods. © 2022 IEEE.

Problems of Regulation and Prospective Development of E-commerce Systems in the Post-coronavirus Era

The article examines the application of e-commerce systems and technologies that have a positive impact on the development of the economy of the post-coronavirus period and the formation of appropriate technical and technological infrastructure for it, as well as promising features and directions of e-commerce. The physical and virtual opportunities created by e-commerce technologies for buyers and sellers are explained. The advantages of e-commerce in the international economic space have been identified. The functions of e-business models in accordance with the commercial stages of enterprises are explained. It was noted that the development of ICT has accelerated the process of transition from traditional commerce to e-commerce, led to the emergence of new global trends in e-commerce. These innovations have raised the issue of the application of modern ICT in the development of e-commerce on the platform of the 4.0 Industrial Revolution. Taking into account these factors, the presented article discusses the application of modern technologies in e-commerce systems, such as 3D modeling, the Internet of Things, artificial intelligence, big data. Features of application and regulation mechanisms of E-commerce systems in real economic sectors, which have a direct stimulating effect on economic growth in Azerbaijan, have been studied. Recommendations were given for the modernization and use of e-commerce systems with the application of the latest ICT technologies.The purpose of the research. The main goal of the scientific research carried out in the article was to develop the scientific-methodological basis for the regulation of the application of e-commerce systems and the study of perspective development problems in the so-called post-coronavirus period after 2020. In the article, attention was paid to the problems of regulation of the application of e-commerce systems and the development of recommendations on increasing the efficiency of prospective development directions.Taking into account the characteristics of the relevant electronic business models, applying them in accordance with the commercial stages of the enterprises' activities and obtaining effective results were among the main goals. Attempts have been made to implement e-commerce systems based on the developing technologies of the Industry 4.0 platform. An attempt was made to solve the issue of using modern ICT in the development of trade processes, which corresponds to the 4.0 Industrial revolution platform. The main stages of application of modern technologies such as 3D modeling, the Internet of Things, artificial intelligence, and Big Data in electronic commerce systems are described.The following are included among the goals of the conducted scientific research: investigation of the application features and regulation mechanisms of e-commerce systems that have a stimulating effect on the economic development of Azerbaijan in real economic sectors, development of recommendations on increasing the efficiency of electronic commerce systems using modern ICT technologies, etc.Research methods used. In the post-coronavirus period, the following research methods were used in the study of the problems of regulation of the application of e-commerce systems and prospective development directions and in the development of their scientific and methodological bases: a systematic analysis, correlation, and regression analysis, mathematical and econometric modeling methods, expert evaluation method, measurement theory, algorithmization, ICT tools, and technologies, etc.Achievements of the author. Achievements of the author. In the so-called post-coronavirus period after 2020, a special approach was taken to the application of e-commerce systems and technologies, which have a positive impact on the development of the economy as an innovative element, and to the study of its prospective development features and directions. By providing scientific support to ensure the effective formation of the digital economy and its sustainability, the researcher offered relev nt recommendations to achieve the solution to some of the goals set before the country. It should be noted that the development of e-commerce systems based on technologies relevant to the Industry 4.0 platform can give a serious impetus to the development of the sustainability of the digital economy.Due to the fact that e-commerce technologies create new additional physical and virtual opportunities for buyers and sellers, the scientific-methodological approaches proposed by the author develop them as a special tool for ensuring the stability of both e-commerce systems and the digital economy in general. The proposals presented will lead to more effective results for the economy to be more cyber resilient through the application of e-commerce systems in the so-called post-coronavirus era. The researcher showed that the effective application of electronic business models in the activities of enterprises can help to achieve effective results. In the development of e-commerce, solutions to the issues of application of 4.0 Industrial technologies such as 3D modeling, Internet of Things, artificial intelligence, and Big Data can be considered as a contribution to the investigation of solutions to existing problems in economic development. For this reason, the means and mechanisms proposed by the author for solving the problems of regulation of the application of e-commerce systems in the post-coronavirus era can be considered one of the main ways to ensure the stability and development of the digital economy.

The Influence of The COVID-19 Pandemics in Indonesia On Predicting Economic Sectors

Several studies have tried to prove the link between the economic sectors in Indonesia with the COVID-19 pandemic. However, research has yet to observe the influence of the COVID-19 pandemic on the predicted performance of regression models. This study proposes the development of previous research following the impact of the COVID-19 pandemic on machine learning performances in predicting economic sectors in Indonesia. The economic sectors mentioned include the exchange rate, CPI, and stock price. The proposed methods for comparison are decision tree (DST) and random forest (RF). Comparison of prediction performance with legacy uses root mean squared error (RMSE), mean squared error (MSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). Test results show that the RF regression model has superior performance compared to DST with the best MSE, RMSE, MAPE, MAE, and r2 value of 0.010, 0.102, 0.64%, 0.100, and 0.89, respectively. Using the T-Test, we prove that the COVID-19 pandemic does not significantly affect machine learning predictions on the exchange rate but significantly affects machine learning predictions on CPI and stock prices. © 2022 IEEE.

Stock Price Prediction of E-commerce Platforms under COVID-19's Influence Based on Machine Learning

After the COVID-19, the global economy was hit greatly. Due to various reasons, the real economy has been seriously damaged, and whether the e-commerce platform will benefit from it is under debate. This paper will study this issue and give an argument. The stock price is a simple and intuitive reflection of the value of the company, which often reflects the company's situation. As the world's largest economies, China and the United States have mature e-commerce conditions and huge E-commerce markets, which are representative and more applicable to macro laws. Therefore, to study the impact of the epidemic on the e-commerce industry, this paper selects the five most representative e-commerce enterprises in China and the United States, collects their stock price information in recent five years, uses machine learning (LSTM neural network and GRU neural network) to predict their stock price trend, evaluates the results and gives a conclusion. According to the results, it is found that although the share prices of three Chinese companies may fall in the short term, the positive effect of the epidemic on ecommerce platforms is greater than its negative effect. © 2022 IEEE.

Stock Market Prediction using Artificial Intelligence

This research focuses on predicting stock closing prices for one day or the future in specific economic conditions. Today, Sri Lanka faces a financial crisis due to the COVID-19 pandemic. Therefore, lots of investors are bankrupt due to unpredictable stock prices. This work mainly focuses on predicting stock prices in banking sector shares such as Commercial Bank (COMB.N), Hatton National Bank (HNB.N), Seylan Bank (SEYB.N), and Sampath Bank (SAMP.N) on Colombo Stock Exchange (CSE) in Sri Lanka. According to the hypothesis, All Share Price Index (ASPI) and Banking Sector indices have been taken as a numerical sentiment parameter other than the historical prices from each bank. Since ASPI shows overall market performance and Banking sector indices show banking sector capitalization changed over time. There can be a positive and negative sentiment when the ASPI and Sector Indices increase and decrease, respectively. Finally, a dataset is divided into 70% for training and 30% for testing. This study has used Recurrent Neural Networks (RNNs) such as Long short-term memory (LSTM) and Gated Recurrent Unit (GRU) using 25, 50, 100, 150, and 200 epochs. LSTM model has given the lowest Mean Squared Error (MSE) and Root Mean Square Error (RMSE). According to the LSTM model, COMB.N, HNB.N, and SAMP.N were given the lowest MSE, and RMSE for 100 epochs, and SEYB.N was given the lowest MSE and RMSE value for the 150 epochs. © 2022 IEEE.

Stock Price Prediction Using News Sentiment Analysis

In this era of digitization, tasks can be performed from anywhere in the world that previously required manual movement. It is the same for investing and trading in stocks. With the ease of investing and trading in stocks via the internet, a more extensive segment of society has started investing. The stock price depends on multiple factors such as politics, economics, war, society, and news sentiment. Therefore stocks are really hard to predict due to such vast dependencies. Stock markets are an important issue in the financial world. Prediction of stock prices during the global pandemic of Novel Coronavirus 2019 (COVID-19) can be very helpful to stakeholders. The attempt of predicting the stock prices have been made by previous researchers using sentimental news analysis through Support Vector Machine (SVM), Neural Network, and Naive Bayes. However, they have low accuracy, and some even claim that news is not a crucial governing factor for the stock price. This paper aims to predict the stock market prices through news sentimental analysis using techniques such as Long Short Term Memory and Artificial Neural Network against classifier models like Natural Language Toolkit, Valence Aware Dictionary for Sentiment Reasoning, Recurrent Neural Network for price prediction. S.Mohan [1] MAPE scores came out to be 1.17, 2.43 for RNN and RNN with news polarity for Facebook stock prices. Our results came out to be 1.21 and 1.94, slightly better results, thus showing optimism in the dependence of stock prices on the news. © 2022 IEEE.

Stock Market Comparison and Analysis in Preceding and Following Pandemic in Bangladesh using Machine Learning Approaches

For a long time, stock price forecasting has been a significant research topic. However stock prices depend on various factors that cannot be predicted, and that's the reason it is almost impossible to predict stock prices accurately. Many researchers have already worked in this area. Recently, the COVID-19 pandemic had a great effect on the stock market. The main purpose of this paper is to predict the stock closing prices for two major stock exchanges in Bangladesh and compare the prediction accuracy based on before and after pandemic data. The implemented models are Autoregressive Integrated Moving Average(ARIMA) and Support Vector Machine(SVM) and Long Short-Term Memory (LSTM). Raw datasets were considered, which were collected from Dhaka Stock Exchange(DSE) and Chittagong Stock Exchange(CSE). Data preprocessing was done on both of the datasets. After analyzing the overall accuracy for each algorithm, it was found that LSTM provided better accuracy than ARIMA and SVM for both the DSE and CSE datasets. © 2022 ACM.

Deep Neural Networks for Stock Market Price Predictions in VUCA Environments

The purpose of this paper is to examine the useful application of deep neural networks in stock price prediction in efficient markets and under Volatile, uncertain, complex and ambiguous (VUCA) environments, especially in the covid-induced USA financial of 2021 crisis. VUCA environments such as stock markets have made it difficult to predict stock prices. This study investigates the usefulness of deep learning architectures in stock price prediction for S&P 500’s top 3 stocks namely Apple, Microsoft and Amazon. The Bidirectional Long Short Term Memory (BLSTM) and Bidirectional Gated Recurrent Unit (BGRU) were implemented in this study and provided excellent accuracy results, the highest been 95.04% using the BGRU for Microsoft stock. The novelty of this study is the successful application of bidirectional deep neural networks to financial time series and forecasting of stock prices under financial crisis. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

Analysis and Prediction of Healthcare Sector Stock Price Using Machine Learning Techniques: Healthcare Stock Analysis

Healthcare sector stocks are a very good opportunity for investors to obtain gains faster most of the time in a year and mostly during this COVID pandemic. Purchasing a healthcare stock of a certain company indicates that you hold a part of the company shares. Specifically, various examinations have been led to anticipate the development of financial exchange utilizing AI calculations, such as SVM and reinforcement learning. A collection of machine learning algorithms are executed on Indian stock price data to precisely come up with the value of the stock in the future. Experiments are performed to find such healthcare sector stock markets that are difficult to predict and those that are more influenced by social media and financial news. The impact of sentiments on predicting stock prices is displayed and the accuracy of the final model is further increased by incorporating sentiment analysis. © 2019 American Society of Mechanical Engineers (ASME). All rights reserved.

Effect of COVID-19 on Stock Market Prediction Using Machine Learning

Share market is a chaotic and ever-changing place for making predictions, as there is no defined procedure to evaluate or forecast the value of a share of a company. Methods like time series, technical, statistical and fundamental analysis are used to predict the price of a share. However, these methods have not proven to be very consistent and precise for making predictions. COVID-19 has further deteriorated the chances to find such a tool as the markets have taken a huge hit in the first quarter of 2020. In this paper, support vector machine and multiple regression algorithms will be implemented for predicting stock market prices. Our aim is to find the machine learning algorithm which can predict the stock prices most accurately before and during the pandemic. The accuracy for every algorithm will be compared and the algorithm which is the most accurate would be considered ideal. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

Stock price prediction using improved extreme learning machine methods during the Covid-19 pandemic and selection of appropriate prediction method

Purpose: Coronavirus disease (Covid-19) has created uncertainty in all countries around the world, resulting in enormous human suffering and global recession. Because the economic impact of this pandemic is still unknown, it would be intriguing to study the incorporation of the Covid-19 period into stock price prediction. The goal of this study is to use an improved extreme learning machine (ELM), whose parameters are optimized by four meta-heuristics: harmony search (HS), social spider algorithm (SSA), artificial bee colony algorithm (ABCA) and particle swarm optimization (PSO) for stock price prediction. Design/methodology/approach: In this study, the activation functions and hidden layer neurons of the ELM were optimized using four different meta-heuristics. The proposed method is tested in five sectors. Analysis of variance (ANOVA) and Duncan's multiple range test were used to compare the prediction methods. First, ANOVA was applied to the test data for verification and validation of the proposed methods. Duncan's multiple range test was used to identify a suitable method based on the ANOVA results. Findings: The main finding of this study is that the hybrid methodology can improve the prediction accuracy during the pre and post Covid-19 period for stock price prediction. The mean absolute percent error value of each method showed that the prediction errors of the proposed methods were all under 0.13106 in the worst case, which appears to be a remarkable outcome for such a difficult prediction task. Originality/value: The novelty of this study is the use of four hybrid ELM methods to evaluate the automotive, technology, food, construction and energy sectors during the pre and post Covid-19 period. Additionally, an appropriate method was determined for each sector. © 2022, Emerald Publishing Limited.

COVID-19 Spread Prediction and Its Impact on the Stock market price

Predicting the Covid-19 spread and its impact on the stock market is an important research challenge these days. In order to obtain the best forecasting model, we have exploited neuro-evolutionary technique Cartesian genetic programming evolved artificial neural network (CGPANN) based solution to predict the future cases of COVID-19 up to 6-days in advance. This helps authorities and paramedical staff to take precautionary measures on time which helps in counteracting the spreading of the virus. The rising number of COVID cases has caused a significant impact on the stock market. CGPANN being the best performer for the time series prediction model seems ideal for the case under consideration. The proposed model achieved an accuracy as high as 98% predicting COVID-19 cases for the next six days. When compared with other contemporary models CGPANN seems to perform well ahead in terms of accuracy. © 2022 IEEE.

Fuzzy Time-Series Models Based on Intuitionistic Fuzzy, Rough Set Fuzzy, and Differential Evolution

Time-series forecasting is a vital concern for any data having temporal variations. Comparing with the other conventional time-series methodologies, the fuzzy time-series (FTS) proved its superiority. Substantial research using time-series forecasting to predict the stock index data has been found in the earlier works. The fuzzy sets approach alone cannot explain the data thoroughly. In this article, we have proposed three different methods of time-series forecasting. The first method is based on a rough set of FTS, a rule induction-based method;the second method is based on intuitionistic FTS. The last method is the extension of the second method using differential evolution. In the first model, a fuzzy algorithm based on rules is used to derive prediction rules from the time-series data and adopt an adaptive expectation model that replaces the fuzzy logical relationships or groups. In the second method, to split the universe of discourse into a non-uniform interval, a clustering algorithm-based intuitionistic fuzzy approach is used, taking care of the membership and non-membership function. Finally, the last method has been tuned for a better outcome using differential evolution. To examine the results, contrast analyses on the Taiwan stock exchange data and daily cases of COVID-19 pandemic prediction have been carried out. The outcome of the proposed approaches validates that the first and second techniques, showing promising results. However, the third method outperforms the other methods and the present techniques concerning the root-mean-square error metric. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

The COVID-19 Pandemic’s Impact on Stock Markets and Economy: Deep Neural Networks Driving the Alpha Factors Ranking

The spread of COVID-19 and international measures to contain it have a significant influence on the economic movement. COVID-19 expresses a fearsome and original risk, setting a massive challenge for investors. These facts give a good reason to study the impact of COVID-19 progress in recent stock-market behavior and draw comparisons over various sectors and industries. This leads to the development of an intelligent trading strategy to investigate the pandemic impact on stock markets. Factors are time-dependent variables that illustrate individual factor periodicity, distinctive cumulative return patterns, and spread duration of under performance. Therefore, we develop a ranking-based clustering strategy alongside two deep neural networks. One is for market state representation, whereas the second model is for the period by which the ranking mechanism is applied. The analysis suggests that COVID-19 is having an unprecedented influence on markets. The investment process needs to be continuously adjusted using different indices for more synthesized patterns. This chapter offers a significant analysis of investors’ evaluation of the overall economy’s consequences, specifically the stock markets. Overall, the results imply that the health crisis morphed into an economic crisis that increased through financial markets. The combination of this analysis notifies investors of their responses to the emergency. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Long-term Time Series Data Clustering of Stock Prices for Portfolio Selection

In this paper, clustering for stock data is conducted with two clustering methods, k-Shape and k-means with DTW distance measure and the results are compared. The data is the top 129 global electronics manufactures' stock prices from 2018 to 2020 which included the worst Christmas in 2018 and the beginning of COVID-19 outbreak. The involved countries are US, China, Taiwan, Korea, Japan and some others. The clustering results by k-Shape indicate distinctively different effects on those countries' stock markets due to the COVID-19 turmoil. The patterns of the clusters can be visualized to identify the differences among the clusters. We found that each of eight clusters comprises of the same country companies. From that, we could guess that investors or their algorithms tend to invest in companies according to its country rather than the individual company's performance. © 2021 IEEE.

Evaluation of Index Tracking Portfolios During the COVID-19 Pandemic

Index tracking consists of mimicking a benchmark performance with a portfolio formed by a subset of assets contained in the index. Due to the cardinality constraint, obtaining an optimal solution for this problem can be impractical as the number of stocks in the index grows. Then, meta-heuristics, such as genetic algorithms, can obtain good solutions in a reasonable time, making it possible for the investor to run different configurations of the problem before deciding to rebalance or not his portfolio. Also, to evaluate an investment strategy, it is important to perform backtests considering different risk scenarios, especially in crisis scenarios, with a high volatile market. This work aims to analyze the integration of hybrid and pure genetic algorithms and two optimization models in a high volatility market scenario, the Brazilian market index IBOVESPA during the COVID-19 pandemic. We observed that the hybrid algorithms returned competitive solutions, tracking IBOVESPA even closer than the CPLEX solution on the linear model for a non-rebalancing strategy. However, they were not competitive in a rebalancing strategy, with solutions presenting a gap of more than 100% relative to the general-purpose solver solution. © 2021 IEEE

Automakers Stock Price Movement Comparison under COVID-19

In this paper, we compare global automobile manufacturing companies' stock price movement under the pandemic in 2020. The purpose of this work is to investigate the stock price movement of top automobile manufacturing companies. Here, we used machine learning based time series data clustering method. We considered the period of time series stock data from 2020/01/02 to 2021/03/18. In March 2020, around the world, the worst stock price plunge was caused by COVID-19. Then almost all global automakers' stock prices were severely damaged. They, however, recovered gradually their stock prices. On the stock prices, investors' expectations are reflected. The recovery pattern of stock prices can mean the investors' evaluation of the companies. The result of the clustering, contrary to our expectations, shows that the stock prices were likely to move depending on the country, instead of individual companies' performance. The country-based clusters we found are a Japanese companies' cluster, two USA companies' clusters, and two Chinese companies' clusters. In addition, two regional clusters were found which are Asian region cluster and EU region cluster. In the paper we will describe the differences of stock movement patterns among the country-based clusters. © 2021 IEEE.

Stock Price Analysis and Prediction

Covid has taught a valuable lifelong lesson. During the pandemic, economies of countries collapsed and many nations had to undergo a complete lockdown. Individuals lost their sources of income and their savings dwindled trying to survive the lockdown. Many small-scale industries closed down for not being able to recover losses. Despite of the economic machine being slowed;the cog of stock market ran smoothly. The moral learnt was one must have multiple sources of income. During lockdown, the stock market collapsed hard. Now a year later, the market is stronger than before and has achieved new benchmarks. Stock market is erratic and most people relate it to gambling. There are other ways to invest money long term which are a safer bet, but for those who love playing with fire, stock market is a good investment. One might ask why to invest in stocks than going for safer options. No other investment provides potentially higher profits and losses than stock market does. Investing in stock market is purely on people’s own risk. There is no such belief that a particular stock would always provide profit. Some people utilize the advancements of technology and computing resources in order to do algorithmic trading. One might say it’s a fool’s errand as there are some unfathomable factors which affect any stock. But could one gain an edge using these techniques? The proposed system explores this idea further by developing a Machine Learning model which accepts historic prices of stocks as input to predict futuristic prices with good accuracy to construct a portfolio of multiple stocks. The proposed project will help investors to gain an idea of whether investing in a stock may payout or not. © 2021 IEEE.

Deep Reinforcement Learning Algorithms for Multi-Agent Systems-A Solution for Modeling Epidemics

Multi-agent reinforcement learning (MARL) consists of large number of artificial intelligence-based agents interacting with each other in the same environment, often collaborating towards a common end goal. In single-agent reinforcement learning system the change in the environment is only due to the actions of a particular agent. In contrast, a multi-agent environment is subject to the actions of all the agents involved. Multiagent systems can be deployed in various applications like stock trading to maximize profits in stock market, control and coordination of a swarm of robots, modeling of epidemics, autonomous vehicle and traffic control, smart grids and self-healing networks. It is not possible to solve these complex tasks with a pre-programmed single agent. Instead, the many agents should be trained to automatically search for a solution through reinforcement learning (RL) based algorithms. In general, arriving at a decision in a multi-agent system is almost close to impossible due to exponential increase of problem size with an increase in the number of agents. In this paper, multi-agent systems using Deep Reinforcement Learning (DRL) is explored with a possible application in modeling of epidemics. Different stochastic environments are considered, and various multi-agent policies are implemented using DRL. The performance of various MARL algorithms was evaluated against single agent RL algorithms under different environments. MARL agents were able to learn much faster compared to single RL agents with a more stable training phase. Mean Field Q-Learning was able to scale and perform much better even in the situation of hundreds of agents in the environment and is a sure candidate to model and predict the epidemics, in the existing frightening dangerous situation of corona pandemic. © 2021 IEEE.