JMIR Publications

ABSTRACT

Background:

This research presents a modified Singular Spectrum Analysis (SSA) approach for the analysis of COVID-19 in Saudi Arabia. We have proposed this approach and developed it in [1,2,3] for separability and grouping step in SSA, which plays an important role for reconstruction and forecasting in the SSA. The modified SSA mainly enables us to identify the number of the interpretable components required for separability, signal extraction and noise reduction. The approach was examined using different number of simulated and real data with different structures and signal to noise ratio. In this study we examine its capability in analysing COVID-19 data. Then, we use Vector SSA for predicting new data points and the peak of this pandemic. The results shows that the approach can be used as a promising one in decomposing and forecasting the daily cases of COVID-19 in Saudi Arabia.

Objective:

In this study we examine its capability in analysing COVID-19 data. Then, we use Vector SSA for predicting new data points and the peak of this pandemic

Methods:

Modified Singular Spectrum Analysis

Results:

A modified Singular Spectrum Analysis approach were used in this research for the decomposing and forecasting COVID-19 data in Saudi Arabia. The approach was examined in our previous research, and here in analysing COVID-19 data. In the first stage, the first 42 confirmed daily values (02-03 to 12-04-2020) were used and analysed to identify the value of r for separability between noise and the signal. After obtaining the value of r, which was 2, and extracting the signals, the Vector SSA were used for prediction and determine the pandemic peak. In the second stage,we updated the data and included 71 daily values. We have used the same window length and number of eigenvalues for reconstruction and forecasting. The results of both forecasting scenarios have indicated that the peak will be around end of May and mid of June, and the end of this crises will be between end of June and mid of July. All our results confirm the impressive performance of the modified SSA in analyzing COVID-19 data and selecting the value of r for identifying the signal subspace from anoisy time series, and then make a good prediction using Vector SSA method. Note that we have not examined all possible values of window length in this research, and for forecasting we have used only the basic Vector SSA. For future research, we will include more data and considered different window length that may give a better forecasting. In addition, chaotic behaviour in COVID-19 data will be examined as we have some results that show strange patterns, which can be found in chaotic systems

Conclusions:

A modified Singular Spectrum Analysis approach were used in this research for the decomposing and forecasting COVID-19 data in Saudi Arabia. The approach was examined in our previous research, and here in analysing COVID-19 data. In the first stage, the first 42 confirmed daily values (02-03 to 12-04-2020) were used and analysed to identify the value of r for separability between noise and the signal. After obtaining the value of r, which was 2, and extracting the signals, the Vector SSA were used for prediction and determine the pandemic peak. In the second stage,we updated the data and included 71 daily values. We have used the same window length and number of eigenvalues for reconstruction and forecasting. The results of both forecasting scenarios have indicated that the peak will be around end of May and mid of June, and the end of this crises will be between end of June and mid of July. All our results confirm the impressive performance of the modified SSA in analyzing COVID-19 data and selecting the value of r for identifying the signal subspace from anoisy time series, and then make a good prediction using Vector SSA method. Note that we have not examined all possible values of window length in this research, and for forecasting we have used only the basic Vector SSA. For future research, we will include more data and considered different window length that may give a better forecasting. In addition, chaotic behaviour in COVID-19 data will be examined as we have some results that show strange patterns, which can be found in chaotic systems