Transforming Categorical to Numerical: Encoding (Continuization)

[@biatutorial5208]

Thanks Dr Noureddin alot for these tutorials. they are really very helpful. please allow me to suggest that it will be more and more helpful for arabs if you introduced some materials in arabic. actually there are more content on internet for sciences in English, but it's too poor in arabic. specially for science like data mining. it's two years for me looking for any Arabic tutorial for data mining and business intelligence but never found. so when i saw your data mining tutorials i was sure it's new material. for me it's more easy to understand your English than other native speakers , so I'm really very happy i found your channel. thanks for your effort

[@abduljameelshaik3751]

A very good explanation, thanks alot

[@DrNoureddinSadawi]

Thnx

[@_arunsankar]

Thanks for the video. Simple and Informative!

[@DrNoureddinSadawi]

Thanks for watching

[@qtptnqtptnable]

Thx! That's what exactly what I needed to know :)

[@mostafaalbana]

Thanks you so much . it's too useful.

[@biatutorial5208]

the only one problem for me is that i don't know the entropy function so it was hard for me to understand this lesson. i looked for other material explaining it on internet so i could get better benefit of this lecture. but if you had the time to explain it in another video it will be more helpful. thanks alot

[@sofiapuvogel3663]

Dear Dr Noureddin... I don`t understand how is the target chosen in target based encoding.. I would be very grateful if you explain it to me

[@pascal3934]

Hello Noureddin. I did not understand what do you mean by "target". Ho did you find the number in the column target. Thanks for your response

[@buildaholiczakzak6768]

Do you know how to use neural network on excel on German credit approval dataset Are there any turorials

[@1982Dibya]

How could probability of trend Up with target 1 and 0 is same.Please clarify if I am wrong

[@DrNoureddinSadawi]

notice in the table header it says Probability (1), so I'm only showing the prob of 1, prob(0) will be 1 - prob(1) clear?

[@1982Dibya]

Then in the left hand table for the trend Up with target 0 the value will be 0.33 (1-prob(1)) but you have mentioned as 0.66 for all 1s and 0s as well that what confuses me.Correct me if I am wrong or understand it differently

[@DrNoureddinSadawi]

Dibyajyoti Bose where/when did I say this?

[@1982Dibya]

If you see in the Target_Encoded table with Trend Up as target 0 it is showing 0.66 and Target_Encoded table with Trend Up as target 1 also 0.66.Will it not be 0.33?

[@DrNoureddinSadawi]

YES you are actually right ,,, there is a typo ,,, it should be 0.33 Many thanks for pointing this out ,,, appreciated!

[@ajkdrag]

why did we choose the values of target as 1,1,0 ... so on..... if the first UP had a target value of 0 instead of 1, then the probability would have been 0.33...what does 0 and 1 signify here? Please clarify this doubt Sir..

[@pradhumngoyal9856]

0 and 1 here are the output values given to us in training data set.

[@richakewalramani657]

Where can I find an example of binary encoding r code?

[@RishiSaiReddy]

Richa Kewalramani I’m also looking for encode r example . Please let me know if you find any more information on example models

[@arpanasingh8957]

I need to do this using weka

[@arpanasingh8957]

Sir, could you help me in descritizing percentage to grade using weka

[@DrNoureddinSadawi]

yes, please give me more details

[@arpanasingh8957]

I have to convert 3 year marks of students to nominal

[@malikah7145]

الارقام نعطيها راندوم او كيف ؟

[@natashamarwah2590]

Could someone please tell me how to do Target Based Encoding in R

[@StrykerMareXV]

Is this lesson correct? I think we only need n-1 dummy variables for n different categories. So for example...UP: 1 0, DOWN: 0 1, FLAT: 0 0. Please correct me if I am wrong.

[@DrNoureddinSadawi]

+StrykerMareXV you are right, but it depends on the technique you use to analyse the data after the transformation. For example, if the categorical variable has only two distinct values, then YES we can (or should) represent it by only ONE dummy variable. Now, if your data analysis requires that you have more than ONE output (or target) variable then you would use TWO dummy variables rather than ONE. Is this clear?

[@StrykerMareXV]

+Noureddin Sadawi Thank you for answer prof. Sadawi. I have to wrap my head around it, because I haven't encountered such examples yet. Your videos are really helpful

[@DrNoureddinSadawi]

+StrykerMareXV this an example method that works with more than one output: en.wikipedia.org/wiki/Partial_least_squares_regression

[@ashishankit9911]

can you help me up with 34 categorial variables of 10000 data? any conversion trick in bianry encoding?

[@DrNoureddinSadawi]

yes that's possible .. you can use WEKA, R or Python to do the conversion

[@ashishankit9911]

i am using r and doing some projects, i know the encoding techniques, but i dont know the binary encoding as encodiing with binary no. with less columns as i have 34 categorial values as i stated, i want to reduce the column size?

[@dejal.3606]

It is not possible to average Categorical Data.