For Helen – A Trojan horse.
Virus and malware are terrible things that destroy companies, steal money, and lose customer confidence. We want to protect companies and their customers from these sorts of horrible demons. But which anti-virus company should we look at? This report analyzes the different companies and their effectiveness along with how their effectiveness is changed based off the budget poured into the anti-virus software.
We ran regression in R, and noticed curvature in the residuals, so we decided to cube the budget term. We also noticed that the cubic curves were different for each of the companies, so we included an interaction between the terms and the company. After this the residuals looked good indicating that this is a good model for predicting attacks. The results of our analysis are shown in the graph below
As you can see in the plot above, the Dewormercompany is black, and it shows that as you increase the budget the attacks drop until $1800 and then they start to rise again until about $7500, then they drop again. The same is true for Guard IT which is red in the graph, although it seems to rise and dip again at slightly higher values of budget. You might think these two companies are basically the same, but the p-values are low (lower than 0.05) so these curves really are different, but maybe not enough to really care.
TrustMe is a different effect shown in the graph above in green. If you increase the budget it will increase the attacks, until about $2100 and then the attacks drop until about $7800, then they climb again. The reason for this might be for Dewormer and GuardIT hackers don’t want to attack you if you aren’t spending much because you’re not worth it, and as you spend more they are more tempted and attack more, but if you spend lots and lots then it’s going to be able to block the attacks. For TrustMe the virus company (perhaps) stops blocking attacks for small levels of payments because they want you to pay them more, and then if you spend huge amounts they also stop blocking attacks because they have noticed you got bank.
We can predict the number of attack to within about 30 attacks, and the R^2 was 0.5974 which means 59.74% of the variability in attacks can be explained with this model. This means our ability to predict is fair, not great, but not terrible.
So I would recommend that we purchase the Dewormer and spend $2000 dollars, although there are other possible combinations that would work well as seen in the graph. In the future you might want to investigate the types of attacks, or the time of day of the attacks, or the severity of the attacks to get a better model of exactly what is going on here.