Predicting future with machine learning
@d0wn70ad | 3 years ago | | #d0wn70ad

This. This could be huge. So, I was sitting doing nothing at all and somehow, this idea just struck me. Randomly. What if we document someone's life everyday, and write every event that happend that day in a spreadsheet. In a couple of years we could already have tons of data. With the help of machine learning and a powerful computer we could predict some events that will occur the next day. Let me give you an example:
Let's say, Josh is 15 and in high school. And let's write some events that could happen every day.
Monday- going to school, talking to my classmates, eating lunch, going home.
Tuesday- going to school, taking maths test, talking to my classmates, eating luch, going home.
We do that everyday for 3-4 weeks, and we could already predict some events going to happen Monday or Thursday or any other day. Maybe, with enough data, the machine could find a pattern for when Josh is taking a test and could predict when another test is coming. Again, we need quite large amounts of data for this, but this could actually be possible. And this was only a basic example. Let me show you other things that could be predicted.

Let's say, you document and write in a spreadsheet everything that is happening in a neighborhood everyday for one year. Let's say, that the traffic light hasn't been working well recently, but Bob, the local bus driver does not know that. The machine knows about the traffic light situation and knows that Bob in not aware of this, because he hasn't been passing in that neighborhood for days. The machine could literally predict and prevent a vehicle accident. These are just some of my thoughts. Any feedback would be appreciated!

Better yet, it could bring down a mafia! Yes, seriously.
Let's say that a mafia has been committing crimes for the past 3 years and they cannot be found, because they are leaving no traces.
Luckily, you have all the crimes and the details of all the crimes they committed: date, target, place, etc..
You write ALL the data in a spreadsheet, and using machine learning and a powerful computing power, the machine could find a pattern for the date and location of the crimes and could prevent them and leading to the arrest of mafia members. Again, fictional scenario, but this could actually work.

This could be huge. The only thing that this thing needs to become reality is someone to program the whole thing and write a spreadsheet.

Everything seems perfect, doesn't it? I'm going to give you some examples of why it couldn't work.
Let's say that Josh falls on the stairs at school. The machine will have trouble readjusting and understanding that this won't be an everyday event.
Let's say that the mafia we were talking about stops doing any crime for one year just to plan a heist. The machine will have trouble predicting because of the sudden stop, and most probably will predict the wrong time and date and location.

But I still think that this could be huge. It needs to be tested. And I mean big time. But with sufficient training and the right programmers, this could become a system that will help people, institutions, communities.

And still, predictions could be wrong, even with 100 years of testing and with over 100T of data. Why? Because of the free will. Every human has the free will,and that cannot be predicted, because its FREE will. If this will work, it will only prevent usual events. But still, it needs to be tested because it can change and save lives.

This system can also be used for bad things, for example by hackers. Hear me out! Let's say some hackers want to breach a website to infect the machines that are accessing the site. The site is heavily defended and infecting one page will take much time. They will need to infect all the pages that exist on the site, let's say more than 100,just to infect the majority of the machines. Now, with this script, if they manage to inject it in a critical zone of the website, the machine learning algorithm will monitor the page with the most activity on that site, and will tell the hackers when is accessed and why is accessed. Now the hackers know when to go back and infect it. Sounds like an overkill, right? But it could be possible.
I am glad that I am able to discover something like this, even though it can be used for good or bad. Now I must test it and see if it works or not. If you happen to be a programmer or experienced data scientist or have some ideas about this, let me know your thoughts!
Thanks for reading.

post your comment
There are 0 Comments

To leave a comment please login