What ML can do and cannot? | #8 of 28 | Foundations of ML: The Big Picture
Let's understand what machine learning can and cannot do in this one, we will try to clear out some of the misconceptions that are commonly existing amongst the folks that you're going to be working with.
Complete Machine Learning Course for FREE: https://rb.gy/posrbk
🔹 Can Machine Learning Solve every Problem? Here's the answer
Right. So when you're working on a machine learning project, where clients that you're working with audit executed, if you're working with an organization, amongst the various stakeholders that you are working with, it is common to have unrealistic expectations about what AI and ml can and cannot do.
And one of the reasons for this is the media and the blogs. And the news channel, of course, right. Most of the successful use cases of what ml was able to do is very widely publicized.
Whereas the unsuccessful cases and what sort of data is required to make a successful case, success is not really reported. So naturally, a lot of people tend to believe that AI and ml can do things automatically, and which is unfortunately, not really the case.
Now, let's understand this a bit more closely by taking the example of demand forecasting example, demand forecasting for commercial products to be more precise. Now, demand forecasting is a classic example of implementing a data science project on something that is very important for the company.
Additionally, for companies that are traditionally doing demand forecasting, they will typically have human experts who are really good at what they're forecasting, they tend to have human experts who do the forecasting manually and they tend to have a really good understanding of the field are the product or the domain at which they are doing with forecasting.
Now, when you want to improve the accuracy of the human created forecast, it is now becoming popular for companies to adopt AI and machine learning to do the forecasting, and to do the forecasting more accurately than what a human can do. now in this situation, businesses tend to have an expectation that ml can independently produce the forecast without the interference of a human expert or domain specialists in this case.
Now, I'm not saying that it is not possible to get fairly accurate forecasts for predictable items. For example, if you want to forecast what is going to be the demand for newspapers on a daily basis, such a task can be actually trivial and easy, you can actually achieve fairly good accuracy for something that is so predictable as newspaper demand.
Whereas when it comes to commercial products, such as say the electric cars or heavy machinery, or electronic goods, or even certain category of fmcg goods, it might be really hard without providing the insights and inputs from domain experts.
Why I say this is because once you deploy your forecasting solution forecasting machine learning model, as you start using the model on a day to day basis or month on month basis, you will see that certain points in time the model is going to produce unrealistic or incorrect forecasts, right.
Let me know in the comments section if you have any questions!
If you enjoyed this video, be sure to throw it a like and make sure to subscribe to not miss any future videos!
Thanks for watching!
#machinelearningplus #python #machinelearning #datascience