machine learning data mining

einfochips.com

If you do not have any connection with computer science, you can still learn it

Let's see -

Mango shopping

Suppose you go shopping for mangoes one day. The seller has a mango filled car. You can take mangoes with your hands, the seller will weigh them, and you will pay per kilogram according to the fixed-rate (specific story in India).

Obviously, you want to pick the sweetest, ripe mango for yourself (since you are paying by weight, not by quality).

How do you choose mangoes?

You remember your grandmother thinking that bright yellow mangoes are sweeter than pale yellow. So you make a simple rule: only bright yellow mangoes will be chosen. You check the color of the mango, choose the bright yellow one, pay, and return home. Are you happy?

No that's not enough

life is difficult

Suppose you go home and tasty mangoes. Some of them are not as sweet as you want, so you got worried. Obviously, you would think that your grandmother's talk is insufficient. A comparison of taste with color is not appropriate.

After a lot of thought (and tasting a variety of mangoes), you conclude that the large, bright yellow mangoes are guaranteed to be sweet, while the small, bright yellow mangoes are only half as sweet. (Ie if you buy 100 bright yellow colors) mangoes, of which 50 are large in size and 50 are small, then 50 large mangoes will all be sweet, while out of 50 small, only 25 mangoes will turn into sweet).

You are happy with your findings, and next time you go shopping for mangoes, you keep this in mind. But what the hell... Next time in the market, you see that your favorite seller has gone out of town. You decide to buy from a different vendor, who supplies mangoes grown from a different part of the country. Now, you realize that the rule you learned (that big, bright yellow mango is the sweetest) no longer applies. You have to learn to use it. You taste a mango of each type from this seller and realize that the small, pale yellow is actually the sweetest.

Now, a distant cousin comes to see you from another city. You decide to tell him about mangoes. But she mentions that she doesn't care about the sweetness of a mango, she only wants the juiciest. Once again, you run your experiments, taste all kinds of mangoes, and realize that softer mangoes are more juicy.

Now, you go to a different part of the world. Here, you taste a different taste from your home. You feel that green mangoes are actually tastier than yellow ones.

You marry someone who hates mangoes. He loves apples instead. You go shopping for an apple. Now, all your accumulated knowledge about Mango हैं is useless. You have to learn everything about the relationship between physical characteristics and the taste of apples by the same method of experimentation. You do it because you love him.

Enter computer program

Now, imagine that while doing all this, you were writing a computer program to help you choose your mangoes (or apples). You will write the following types of rules:

Agar (the color is bright yellow and size differs by the large and preferred seller): Mango is sweet.

Agar (soft): Mango is juicy.

e.t.c.

You will use these rules to choose mangoes. You can also send your younger brother to buy mangoes with this list of rules, and you will be assured that he will buy mangoes of your choice.

But every time you make a new observation from your experiments, you have to manually modify the list of rules. You have to understand the complex details of all the factors that affect the quality of mangoes. If the problem becomes sufficiently complex, it can be really difficult to make precise rules by hand that covers all types of mangoes. Your research can earn you a Ph.D. in Mango Science (if there is one).

But not everyone has that kind of time.

Enter machine learning algorithms

The ML algorithm is an evolution over the general algorithm. They make your program "smart", allowing them to automatically learn from the data you provide.

You take a randomly selected sample of mangoes from the market (training data), make a table of all the physical characteristics of each mango, such as color, size, shape, in which part of the country it is grown, by which vendor sold Goes, etc. (characteristics), as well as the sweetness, juice, hardness of that mango (production variable). You feed this data to machine learning algorithms (classification/regression), and it learns a model of correlation between the physical characteristics of mango and its quality.

The next time you go to market, you measure the characteristics of the mango on sales (test data) and feed it to the ML algorithm. It would use a previously calculated model to predict which mangoes are sweet, ripe and/or juicy. The algorithm can internally use the same rules as the rules you wrote earlier (eg, a decision tree), or it may involve something else, but you need to worry about this to a great extent. Is not.

Now you can shop for mangoes, without having to worry about how to choose the best mangoes. And what's more, you can improve your algorithm over time (reinforcement learning), so that it improves its accuracy as it reads more training data, and modifies itself if it makes an incorrect prediction. But the best part is, you can use only one algorithm to train different models, one to predict the quality of apples, oranges, bananas, grapes, cherries, and watermelons, and to keep all your loved ones happy for :)

And that is machine learning for you. Tell me if it's not good.

Machine learning: Making your algorithms smart so you don't have to be there. ;)