Machine Learning Tutorial 2 – Intro to Predictive Data Analytics



welcome back everybody this video is going to be about predictive data analytics we talked about in the previous video but now I'm going to go into more depth of what that is and why it is useful when it comes to machine learning like I said machine learning is teaching computers to learn from data now predictive data analytics is applying that to data we already have in order to predict in the future but why would we want to predict the future don't just think like oh and 2085 the world's gonna end not that kind of stuff we're talking about if someone comes to your site what kind of products are they going to be interested in what kind of restaurants are people interested in we can look at historical data to make things more relevant to our users of our applications or to make our applications work more effectively a lot of uses for this are recommendation engines so you think you go and shopping at a website and then they have all these products that you haven't even looked at but somehow they're recommending them to you and it's crazy because you know how did how did they know that well they can look at their trends in their data to figure out oh people who've looked at this product might be interested in this product over here that is one good practical use Onix I gave them a yoga mat and he tore into a thousand pieces that is one very good practical use of data analytics and just so you know analytics powers everything so this is the foundation for all business decisions so let's get started in talking about data analytics and predictive data analytics and defining the difference so I said data analytics now data analytics looks at data and I drew this image in the previous video which just kind of represents you know a database or a data source of some sort and just imagine having tons of data in here this is known as our data set so why it's called that is because I don't know but it's just a set the data set contains a lot of historical data now that's kind of a key here we need what's known as historical data and what that means is that what we are trying to predict is already known for this data set so in the recommendation engine that I was talking about a couple minutes ago if we have a bunch of historical data of people interested in this particular product but we don't know if they're interested in another product that doesn't really help us we need to know that association between what they're interested in and other products they've bought so that means this data has to come from the history we have to be able to look at what people have already bought once we have that historical data set we can use that on modern or future data so think now we have people on our website and they are feeding us a bunch of data they're looking at all these products they're making purchases how do we take this data set and kind of use it on modern data to predict what kind of products they are interested in that is the field of data analytics and specifically if we're talking predictive analytics we're talking about predicting what people are interested in but there's one thing I wanted to mention this predictive word in predictive data analytics goes a little bit farther than just predicting the future because you can do data predictions with just historical data you don't need to bring machine learning into it necessarily for example there's a lot of people who have a lot of talent and skill and knowledge in statistics they can look at historical data and make predictions and probability calculations on what's going to happen in the future or currently so what is this predictive part and why do we care about machine learning the reason we care about machine learning is if you look at a data set such as this one the reality is is that it doesn't represent reality in its entirety so it's hard to draw reality I was trying to draw it over here and then I was like I'll just try to explain it reality is what's actually existing and what's going on right data just represents reality so as people buy things we can figure out what they're interested in and that is a representation of reality that makes any sense and I'm not trying to get like weird here anything I'm just saying data represents reality but does it represent reality entirely does it represent all of reality does it capture the full essence of reality no it does not it might only represent a portion of reality and if I'm starting to sound crazy just hold up we might have a thousand people interested in a particular product in reality historically though we have only logged you know a couple hundred purchases from these specific people or people similar to these people so we don't know for sure what the people of reality are going to purchase and what they're going to be interested in we may know what most of the people are interested in but I can guarantee you there's one person out there who's completely unique and there's not really anybody good to compare them to so that's where the predictive part comes in when we have a data set and it does not represent reality entirely let's just say this represents reality like 95% and just oh no it's weird quantifying how much something represents reality but just trust me here the reason it's like that is because we don't have data for every single possibility in the entire world in this data set then we could represent reality completely in our data set the predictive part of data analytics is the 5% what is the 5% of non represented reality going to do and this process is something you can go through manually but it's very tedious and using a computer we can speed up the process and figure out the best model of reality and we'll get into more of the fancy machine learning terms like model pretty soon but this 5% is the part that requires predicting that is where machine learning comes in if you want to make this more concrete imagine we have a person and we don't have any data that really represents this person and what they might be interested in well in that situation we don't know what this person might do they might be interested in buying a purse they might be interested in buying a car or they might be interested in buying some candy machine learning essentially is going to go through all of the options of what this 5% of unrepresented reality is going to choose and it's going to pick the most likely we as humans might look at our historical data and not be able to see trends in our data but a machine learning algorithm has like 4,000 eyes to look at the same data and can see trends that we normally would not be able to see so it can pick what's most likely to happen in this situation even though we don't have this specific person or someone similar to this person it can go up a step and look at an overarching number of people and see what the majority of people that are sort of close to this person are likely to choose and say oh well this person's likely to buy a purse it seems kind of crazy right now and it's like wow that's kind of magical and essentially that's what you know most people think of machine learning is it's just you know this magical computer science thing you throw at software and makes it work cooler but this is really how it works and once we start talking about decision trees for example you will see how this works in action and it's going to be a lot of fun so thank you guys for watching and hopefully you guys can see the magic and the usefulness of machine learning please be sure to subscribe as that really helps out my channel I'm almost a 50,000 I'm at like 49,000 something so please help me out here guys and appreciate it thank you bye

15 Comments

  1. Nureyn A said:

    This is awesome! Now I accept that teaching is a talent, you can know a lot but might be hard to teach others, your teaching style is amazing, you are the best.

    June 26, 2019
    Reply
  2. shashi bisht said:

    Really helpful video.Thanks!!
    Do u know how to do predictions on yearly data in python. Which method can I use for this?

    June 26, 2019
    Reply
  3. Path Finder said:

    So like 95% confidence that the data says something based of the data, but at an extreme amount of data

    June 26, 2019
    Reply
  4. Monica Triana Bonilla said:

    you should put up a donation box or something for your great videos, similar to Khan Academy or Wikipedia. You are helping really smart people with your videos. Your explanations are even better than Professors explanations. I would totally donate or give back. If you can dedicate 100% of your time and make money out of this through viewers' contributions it would be great. But say that you will keep doing the videos regardless, is just so you can dedicate more time to them.
    It's also cool how you get distracted right when I get distracted, so you catch my attention the whole time.

    June 26, 2019
    Reply
  5. Preserve Moment said:

    Cool buddy ….. nicely explained

    June 26, 2019
    Reply
  6. Neo Zora said:

    Damn it Onix

    June 26, 2019
    Reply
  7. nils graham said:

    One step at a time

    June 26, 2019
    Reply
  8. Christopher Sprance said:

    Haha, "I was trying to draw reality over here."

    June 26, 2019
    Reply
  9. lewisdonald514 said:

    https://neo4j.com/blog/scientific-research-machine-learning-graphs/

    June 26, 2019
    Reply
  10. pranav waikar said:

    i have question. i am a newbie.
    1. where business intelligence stands in this?
    2. why ML is important when we have hadoop for data analysis?
    3.The new processors like kirin 970 has now NPU. How can we use that here?

    June 26, 2019
    Reply
  11. Vishal sundararajan said:

    So does this involve logical programming at all

    June 26, 2019
    Reply
  12. Buzz Tech Support said:

    Why don't you get views … You e having 49k subs

    June 26, 2019
    Reply
  13. Sam said:

    like your energy! Good luck mate

    June 26, 2019
    Reply
  14. One Last Gamer said:

    Create video on programming like c

    June 26, 2019
    Reply
  15. Sayd Hajar said:

    thanks for all

    June 26, 2019
    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *