Association Rule Learning - Apriori, FP Growth

Hello, I am (name) from learnvern…( 8 seconds pause ; music )

In the session for Machine learning last time we saw FP Growth algorithm

and now we are going to see one more topic on unsupervised learning and we call it association rules or we can call it association rule learning or we can also call it association rule mining.

So, let's see what actually happens in this. So you see the word association which means that one thing is associated with another and has some kind of relation with each other. right?

So where does this association happen?

This association could be like you went to the market and in the market you buy a t- shirt from a clothes store,

and the moment you buy a t-shirt you thought, “let me buy a pair of jeans as well,” so you bought a pair of jeans also and the shopkeeper notes it down that you have bought a t - shirt and jeans.

Now another customer arrives and this customer also bought jeans and then also thought that he should buy a t-shirt along with it and so he bought the t-shirt also.

So did you notice what was happening ?

Those who come to the store and buy t-shirts buy jeans also along with it.

Now, it so happened that a person came and he bought a formal shirt, absolutely formal and of white colour and he thought that he should buy formal pants also along with it.

So what he did was, he bought formal pants also.

Some more customers came and they bought a formal pant and then also bought a formal and decent formal shirt.

Now, this shopkeeper is analysing that whenever a person comes to buy a t-shirt he also buys jeans and whenever a person comes to buy a formal shirt, he also buys a formal pant. So, it strikes in his brain that a formal shirt and formal pants are associated with each other, right and after formals,

If it is about informals meaning casuals then t-shirt and jeans are associated.

So when Rakshabandhan comes then Bhai Bahan brother sister t-shirts. Right !

So,from the observations of this kind we can call out a rule that these things are associated like if someone buys this, he will buy that too.

If he buys the first one then he will buy the second one too, so these are called association rules.

(02/23)

Let us know more about this. So,this is an unsupervised learning approach and in this, we see that how much is one item dependent on the other one.

That means if you buy one item then what other thing could you buy with that.

Presently we were talking about markets and now let us talk about online OTT platforms. You saw a video, let us suppose it was a song, some spiritual worship song ok, so now you will start receiving recommendations that those who have listened to this worship song must had listened to these worship song as well or if you are watching a horror movie then you may get recommendations that “users who watched this movie, also watched these movies”.

What is this?..

what this actually is basically dependency, this is a way of finding dependency and after finding that ultimately we display it to the user in this way.

So after finding association mapping is done like this movie goes with this movie, this product goes with this product or this item with this item, like this mapping is carried out and this is called association rule learning or rule mining.

Now let us move forward , now I am going to tell you two you may call it techniques or algorithms and are very popular and widely used.

A Priori or Apriori, FP Growth.

You will see in Apriori that in its name itself prior is written, priori is written which means that we have some prior knowledge.

So what happens in this is, those transactions or data sets that are recorded in them are observed and see what is frequently being observed…

Ok now what to observe.

It is to be observed that what all pairs whether of 2, 3 ,4 means items bought are bought in set of 2,4,10 so which are the patterns being observed of 2-2 again, which all patterns of 3-3 are being observed again and again, so this is what we identify and on basis of that make association rules.

So what do we want here ?

We want database transactions for the first customer, second customer, third one ,1000 one , 2000 one and so on.

You want transactions and this is mainly used in market basket analysis whose example we will see in the slides ahead.

So this was Apriori,

Now after Apriori the second one is FP Growth algorithm.FP Growth algorithm, its name is Frequent Pattern.

(04/48)

Frequent pattern means, this FP Growth is Apriori’s improved version in which the patterns being identified are identified in a different way.

How do we do it ? We will do it with a tree.

A tree structure is made and observing the structure of that tree we understand which pattern is frequent and taking place time and again.

Now, after making that tree we identify these patterns from it only.

So this means that Apriori and FP Growth are the two techniques that we use.

Apart from this there are other techniques also which are used , we will discuss only two techniques here.

Let us see the application now, this is market basket analysis, ‘’customers who bought these items also bought these items’,

means those who bought this bought this also, so this is called market basket analysis, means those who put this in basket put these items also in the basket. Clear up till now ? Okay!

This is used very often, retailers use it to place items together or to club them this is often used.

Next, if you talk about medical diagnosis then here also this is very helpful , for curing any disease you have to identify that for that particular disease to happen , What were the reasons ?

(06/03)

means it could be due to having some food or due to improper balance of routine or due to incomplete sleep or due to exposure to some kind of environment , many rules can be there ,many conditions can be there, Right ?

If we have old data then we can analyse that this disease which has happened to these personnel and these many environmental factors were affecting him and observe that out of those which were the most frequent, so thes 3 were more frequent.

Do we have the same in new people, yes we have which means these 3 environmental factors are the main contributing reasons for this disease.

So to cure the disease lets cure these factors isn’t it.

Well, I will not go much into medical background but in medical diagnosis also association rule mining helps, OK.

Today's session we stop here and the next parts we will see in the next session. So keep learning, remain motivated, thank you.

If you have any queries or comments, click the discussion button below the video and post there. This way, you will be able to connect to fellow learners and discuss the course. Also, Our Team will try to solve your query.