|
|
 |
| |
|
|
|
Modelling the South African Covid-19 Induced Web Traffic Data Shift using Artificial Neural Networks |
|
|
|
PP: 1049-1056 |
|
|
|
doi:10.18576/amis/160623
|
|
|
|
Author(s) |
|
|
|
Judah Soobramoney,
Retius Chifurira,
Temesgen Zewotir,
|
|
|
|
Abstract |
|
|
| In response to the coronavirus pandemic of the year 2019 (Covid-19), several governments around the world restricted commercial, economic and socio-economic activities. For emerging markets like South Africa, the pandemic and subsequent restrictions have negatively impacted the financial standing of many sectors, corporates and households. Several corporates that rely on digital technologies such as website marketing experienced a steep decline in traffic flow onto the website and a distinct change in online behaviour. Whilst such change in online behaviour directly impacted the revenue of corporates, such data shifts posed further challenges to the accuracy of machine learning models that were trained on online web behaviour prior to the data shift.
This paper aimed to explore the key features of behavioural change observed on a South African website during the covid-19 pandemic. This research utilises data obtained from the website of an engineering and training corporate in South Africa. Using artificial neural networks, the results indicated that the number of visitors, the sessions per visitor and seasonal period were important indicators of a data shift. However, whilst a drastic drop in volumes were noticed during the data shift, for those that did enter the website, the behaviour remained somewhat stable. The study also found that an artificial neural network was highly capable of detecting a data shift. Whilst the findings of the study are specific to the observed website, the methods applied could be adopted in various other applications.
|
|
|
|
|
|
|
|
