Pattern Recognition LettersPrognosticating the effect on Unemployment rate in the post-pandemic India via Time-Series Forecasting and Least Squares Approximation
Introduction
COVID-19 pandemic has hit the world very harshly. India has now become a hotspot of this pandemic after the USA and Brazil, after surpassing Russia in the number of cases, and is soon to enter stage 3 considering fatality rates. The countries are helpless before this natural phenomenon and are forced to implement lockdowns to prevent its spread. As is evident from many other historical pandemics and epidemics in various parts of the world, there has always been a great deal of adverse effects on the economic conditions of the country or the world as a whole. Consequently, due to lockdowns, trading and working in the industries is stalled which in turn leads to loss of jobs of numerous employees. The employers are then not in the position to support themselves and thereby many companies and industries get shut down. Since industries run a country's economy, and with the industries shutting down or getting in loss, the economy of the entire country gets disrupted. A disease outbreak, being at the scale of the world, affects the entire world. Since most countries are connected through trade and commerce, the pandemic can lead to utter loss of jobs and economy of nations.
The COVID-19 disease has been growing steadily at an exponential rate throughout the world. Since its inception, it has disrupted various sectors of human life including business, health, education and governance. As of July 20, 2020 the number of cases is 14,360,451, out of which 603,285 have lost their life while 8,071,937 have recovered1. The economies have been adversely affected and their GDP (Gross Domestic Product) is following a negative slope in today's scenarios. In India, due to sequence of lockdowns, menial laborers have lost their daily-wage jobs, and are left with nothing to sustain in this pandemic-affected economy, and hence many have died not due to the disease but of scarcity of money to support one's family. This is a very arduous situation in the country, and its citizens must not leave hope. Government has been doing its best to control this pandemic. It has taken many initiatives and cut down the cost of COVID-19 tests by half, but still the situation in the country is not under control and is surging high.
The unemployment rate in India had been swinging closely to around 2.7% in previous years but has been exaggerated to an astonishing 15% average during this pandemic, especially between March and May. The rules of social distancing and lockdown has added more to disruption rate of dwindling unemployment rate and consequently on the overall economy of the country. Apart from studying the effects of the pandemic on unemployment rate, inflation rate and economy growth rate, the paper also analyses the consequences of up-skilling at home, on academia and corporate world. We run time-series forecasting on the collected data, and present our findings through comprehensive analysis of the results.
The paper is organized as follows. Section 2 presents a discussion about the works that have been conducted and published in the past. Section 3 discusses the algorithms and time-series forecasting architectures used in the paper. Section 4 outlines the methodology that has been adopted to implement analyses presented in the paper, and also describes the workflow that is used in the implementation of the proposed architecture. Section 5 talks about the issue of data unavailability, its collection and curation. Section 6 presents the results of the algorithms employed to perform time-series forecasting on the curated dataset. Section 7 presents a comparative analysis of the proposed architecture with other available time-series forecasting frameworks and existing researches. The several engineering applications of time-series forecasting is presented in Section 8. At last, Section 9 summarizes and concludes the paper.
Section snippets
Related work
With the onset of the pandemic, a lot of researches have been conducted by researchers worldwide on the various aspects of the mankind's life from health and hygiene to social and environmental contexts. There exists the study where the author estimated the date when the number of cases would reach the peak and when it will flatten out using methods shared from data analysis and machine learning paradigm [1]. The authors in [2] used the time-series forecasting methods like ARIMA to predict the
Least Squares Approximation
LSA is a method to approximate the solution to a set of linear equations for whom a real solution does not exist. It works by minimizing the sum of squares of residuals obtained by each single equation. It is most commonly used in the domain of data fitting, wherein the residuals are the difference between observed and actual values, and the LSA tries to minimize the sum of these residuals.
Since, a line, representing solution, intersecting with the plane of set of equations, cannot be found;
Heuristic-based Model Training and Prediction
The prediction of future effects has been made using Facebook's time-series forecasting framework Prophet. The accumulated data was first cleaned and pre-processed as per the requirements. Due to the pandemic in 2020 and unavailability of required data, basic implementations of these models could not perform and give reliable results.
To overcome the predilection of the learning models on comparatively larger pre-pandemic historical data and the pandemic-induced sudden hike in unemployment rate,
Data Collection
There is an unavailability of proper data regarding unemployment rate in India prior to 2016. Adequate data started to be gauged from 2016 onward by CMIE3, India. But this much amount of data is not adequate and sufficient for training statistical and machine learning models - they need good amount of data to learn patterns. So, we also referred to World Bank and the International Labor Organization (ILO) for the data. After data aggregation and cleaning, the data before
Results and Discussion
The experimentation and inferencing involves three steps. In the first step, we run the time-series forecasting framework on two types of datasets that we have created. Thereafter, we model the results in the form of matrix equations, and then in the last step, we combine the results and optimize it by finding suitable parameters and using Least Squares Approximation (LSA). The LSA utilises the Levenberg-Marquardt Algorithm (LMA) [14] underneath to find a vector approximation. All the
Advantages over Other Vanilla Models
As shown in Section 6, the vanilla models (except Prophet) could not correctly learn the pattern in the data. They were either over-hyped due to the sudden COVID-19 hike or they were just not being affected by this anomalous trend in the data. Likewise, ARIMA, FARIMA, LSTM and SVR could not learn the pattern, while SARIMA did try to learn and adjust to the pattern in the data, it still was unable to address the non-stationarity property that is present in the data, it just predicted strictly
Engineering Applications of Time-series Forecasting
Time series forecasting has already proven its importance in various fields like weather forecasting, process and quantity control, stock market predictions etc. The new emerging applications of time series analysis can be found in energy consumption sectors. One of such works is reported in [17], in which the new building's energy consumption where the past data does not exist, computer simulation is used in such scenarios to predict future energy consumption. While for existing buildings,
Conclusion
The COVID-19 pandemic adversely influenced and disturbed all the phases of human lives. The number of cases in India has crossed an enormous point of 1.44 millions surpassing Russia and is now third on the list, just behind USA and Brazil. An alarming situation has engulfed whole of the country with many places identified as Containment and Red zones. The search for the cure for this disease is surging high but in due time countries are helpless. They are forced to bow before this pandemic and
Authorship Confirmation
Please save a copy of this file, complete and upload as the “Confirmation of Authorship” file.
As corresponding author I, Pawan Singh, hereby confirm on behalf of all authors that:
- 1.
This manuscript, or a large part of it, has not been published, was not, and is not being submitted to any other journal.
- 2.
If presented at or submitted to or published at a conference(s), the conference(s) is (are) identified and substantial justification for re-publication is presented below. A copy of conference
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (19)
- et al.
A review on time series forecasting techniques for building energy consumption
Renewable and Sustainable Energy Reviews
(2017) - et al.
Covid-19 pandemic data analysis and forecasting using machine learning algorithms
(2020) - et al.
Trend analysis and forecasting of covid-19 outbreak in india
(2020) Impact of covid-19 on indian economy-a review ajay kumar poddar and brijendra singh yadav2
Horizon
(2020)- et al.
Covid-19: Impact on the indian economy
(2020) - et al.
The impact of covid-19 in Indian economy–an empirical study
International Journal of Electrical Engineering and Technology
(2020) Prediction of unemployment rates with time series and machine learning techniques
Computational Economics
(2020)- M. Desaling Germay, Modeling and forecasting unemployment rate in sweden using various econometric measures...
- M. Fajar, O. R. Prasetyo, S. Nonalisa, W. Wahyudi, Forecasting unemployment rate in the time of covid-19 pandemic using...
Cited by (3)
Multicriteria decision support employing adaptive prediction in a tensor-based feature representation
2023, Pattern Recognition Letters
2012 Elsevier Ltd. All rights reserved.