Association Rule Analysis|Market Basket Analysis

[@spicytuna08]

thanks much amar.
i see a possible scaling problem.
let's say there are 1000 items.
associative rule with just two items would involve combination of (1000, 2 ).
associative rule with just three items would involve combination of (1000, 3 ).
associative rule with just n items would involve combination of (1000, n ).
now this amounts to huge set of rules.
you explained in case of 2 items.
how to build rules in case of N items?
again, your videos are outstanding.

[@karannchew2534]

Notes for my future revision.
Recommendation engine need at least to work:
1. Who the target is - personal attributes, OR/AND
2. The preference of the target
*Recommendation Engine types and corresponding algorithm types:*
1. Generic
-Association rule analysis (eg market basket)
2. Personalised
2a Content-based filtering (using cosine similarity)
2b Collaborative filtering
--Model Based
--Memory Based
---Item-based CF
---User-based CF
2c Combination of Content-based and Collaborative Filtering
**Association Rule Analysis**
List of all items in a transaction/basket/cart.
Not involving attributes of the products.
SUPPORT of an item
=Support of Item1
=Chance of an Item1 appearing among all the baskets
CONFIDENCE of Item2 given Item1
=Chance of Item2 appearing given Item1
=Frequency of Item2 and Item1 appears in same basket / Frequency of Item1 appearing in a basket
=Conditional Probability
LIFT of Item 1 and 2 together
= Confidence(Item2, Item1) / Support(Item1)
=Conditional Prob / Prob of Conditon
For every combination of item, calculate their Association Rule, i.e. the SCL values:
1. Support
2. Confidence
3. Lift
Example:
Given ItemA, SCL with ItemB is xxx
Given ItemA, SCL with ItemC is xxx
Given ItemA and ItemZ, SCL with ItemB is xxx
...and so on
Apriori Algorithm can be used.

[@juliusjulius1146]

Nice explanation, I love the way you teach with real life examples. Excellent teaching

[@sandipansarkar9211]

finished watching

[@pankhurithakur4291]

very informative and comprehensive video. Thankyou for google explanation.

[@nehasrivastava8927]

finally found the best recommendation systems series..Thanku sir

[@UnfoldDataScience]

Welcome Neha.

[@chinweezulu4681]

Simple and detailed explanation. Weldone👏

[@siddhantjain452]

Ek no video !! better than SimpliLearn !!

[@UnfoldDataScience]

Thanks Siddhant.

[@prekshajain9143]

Point to point amazing explaination,

[@UnfoldDataScience]

Thanks Preksha. pls share with friends as well

[@Dami_San]

Thank You. The explanation was simple and straight forward

[@UnfoldDataScience]

Welcome Dami.

[@rohitreddy3609]

Very neatly explained bro. Love the way you made things simple. Please keep up the good work.

[@UnfoldDataScience]

Thanks Rohit, your comments are my motivation.

[@pei-yungchang7259]

Thank you for your clear explanation on recommendation system!

[@UnfoldDataScience]

Glad it was helpful Pei-Yung.

[@negusuworku2375]

Thanks a lot, very good. keep it up

[@rajan9382]

Nice explanation, thanks sir.
Waiting for your next lectures.

[@UnfoldDataScience]

Thanks Rajan. Keep Watching :)

[@faberrocalv8958]

Hi! Thanks for the video, my questions is: What are the recommended values for Support, Confidence and Lift to consider that a rule is strong enough and valid?

[@UnfoldDataScience]

Hello, It depends on the domain knowledge. There is no definite rule to fix a value.

[@prajwalbharadwaj2386]

you have to design a min_sup (Minimum Support) or min_conf (Minimum Confidence) threshold values. The more relaxed the support parameter, the more the candidates you induce into an itemset ; hence more the computation or more passes the apriori algo has to make!!

[@rameshmamilla5392]

Thanks for the good info Aman! waiting for the next video.

[@UnfoldDataScience]

Welcome Ramesh :)

[@vipinkumar-dm2nd]

Hi, Clear and Good presentation skills, Thank you for sharing :-)

[@UnfoldDataScience]

Thanks Vipin. Your comments are my motivation.

[@sandipansarkar9211]

great explanation

[@UnfoldDataScience]

Thanks Sandipan.

[@rupalishirkande6163]

Nice explanation 🤗

[@SaiVivekanandKuchimanchi]

good bhaiyya

[@rutujagurav466]

Very nice explanation thank you sir

[@UnfoldDataScience]

You're most welcome Rutuja.

[@educationsoulwelfaresociet6184]

I appreciate your efforts. Please keep up the good work.

[@UnfoldDataScience]

I will try my best

[@sadhnarai8757]

Verry good aman

[@UnfoldDataScience]

Thank you :)

[@sadhnarai8757]

Very good Aman

[@UnfoldDataScience]

Thank you.

[@kaifali4691]

great explanation..thanks a lot!!

[@UnfoldDataScience]

Thanks Kaif

[@r21061991]

Great video !!

[@Upmit]

What are Two way and three way lift in Market basket analysis, how can we calculate it.

[@harsh6169]

Nice.. waiting for next coding video on this...Also request if you can upload some videos on Time series forecasting like Arima, Exponential Smoothening, Prophet..etc. Thanks!

[@UnfoldDataScience]

Thanks Harsh, yes I plan to create a separate playlist on Time series forecasting and natural language processing as well.

[@harsh6169]

@@UnfoldDataScience Thank you so much. Waiting for your playlist.

[@sahilsangam3846]

Very nice explanation 👍

[@UnfoldDataScience]

Thanks for liking Sahil .Keep watching.

[@preranatiwary7690]

Very nice

[@UnfoldDataScience]

Thank you :)

[@rohanshet123]

👌👌👌

[@adityaverma4770]

your content is good ,thanks

[@UnfoldDataScience]

Welcome Aditya. Happy Learning.

[@merlingrace2884]

What does high support and high confidence mean?
Low support high confidence
High support low confidence and
Low confidence and low support?

[@UnfoldDataScience]

HI Merlin, the answer of all your questions can be understood from what is support and confidence.
say A= Milk
and B=Bread
Support = (A+B)/Total number of transactions, means out of total transactions, how may has A+B together.
Confidence = (A+B)/A, means out of all the transactions having A how many has B in it.

[@prachitiwari3394]

I am having project on tour and travelling in which I have to analyze impact of sociodemographic factors while selecting different destinations, so in this case which technique I should apply?

[@UnfoldDataScience]

You can go for regression models or Random forest.

[@r21061991]

@@UnfoldDataScience Can she try boosting methods for this problem ???

[@vattypants]

Try me 😉

[@upamasandapani3613]

Thanks!

[@UnfoldDataScience]

Welcome!

[@merlingrace2884]

Can u clarify this for me

[@UnfoldDataScience]

answered.