How Neural Networks Keep Your Email Imbox Clean
Spam emails, everyone gets them and nobody likes them. However, this age-old nuisance of the internet appears to be on its way to an early retirement. In the year 2016, spam emails made up almost 60% of all email traffic on the internet, yet in only four years, its prevalence has dropped to 28.5% (https://www.statista.com/statistics/420400/spam-email-traffic-share-annual/). So the question inevitably comes up, what happened, what’s leading to this collapse in spam emails?
Well, the simple answer to this is the introduction of Neural Networks into email filtering algorithms. These neural networks use pattern recognition of larges data sets to classify certain ‘hidden layers’ as markers to identify spam. The system is rather simple, first a data set of known spam and non-spam emails are identified and chosen to run through the neural network. The algorithm knows that certain emails must end up in the spam output and the other ones must end up in the non-spam output (output layer). Therefore, the algorithm must create the ‘hidden layer’ filtering nodes based on pattern recognition to ultimately direct emails to the right output layer. Below shows how the process runs from left to right, with the input layer consisting of all emails, and the output layer being separated between spam and non-spam emails.
To create the ‘hidden layer’ nodes, the algorithm uses machine learning to run this simulation over and over again until it can create filtering nodes that can consistently send spam emails to the spam output and non-spam to the non-spam output.
Effectively, through trial and error, the algorithm can find patterns of spam emails and use these patterns to filter your email inbox. And through this simple, but computationally complex approach, your email inbox is spared the plight of obnoxious spam.
(https://www.sciencedirect.com/science/article/pii/S2405844018353404#:~:text=Several%20research%20work%20have%20employed,patterns%20in%20the%20email%20messages.)
(https://www.sciencedirect.com/topics/nursing-and-health-professions/perceptron)