Scribd
Upload
32 views11 pages

ML U4 Omkar Pawar

The document discusses various machine learning algorithms, including K-Nearest Neighbors (K-NN), Support Vector Machines (SVM), and Ensemble Learning techniques like Random Forest and AdaBoost. It explains how K-NN classifies data based on similarity, SVM creates decision boundaries for classification, and Ensemble Learning improves accuracy by combining multiple models. Additionally, it highlights the advantages and disadvantages of these algorithms and provides examples of their applications.

Uploaded by

ELECTRO CLASHING
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
32 views11 pages

ML U4 Omkar Pawar

The document discusses various machine learning algorithms, including K-Nearest Neighbors (K-NN), Support Vector Machines (SVM), and Ensemble Learning techniques like Random Forest and AdaBoost. It explains how K-NN classifies data based on similarity, SVM creates decision boundaries for classification, and Ensemble Learning improves accuracy by combining multiple models. Additionally, it highlights the advantages and disadvantages of these algorithms and provides examples of their applications.

Uploaded by

ELECTRO CLASHING
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 11
2 Cee) “le 84 thm 1 based on Supeavised GE Ke Nearest Netghbovr ¢ © K- Nearest Neighbour algo Learning technique: ® K-NN algo™ can be used for Regression as well as classi - Fieation but mostly | Evdidean distance of stvden s) is given as a = Af e= Ha) Ohin%y, ~ J@-a)" +G-a)* eel Similnty, ds=§ dy= 5, d= 36 So, we arrange all above distances in increasing order. we §% (2, 44,4,,4.) = (296,45) As, k=3, we check the result skatus of first thee studen namely , (Se, Sq, 51) 0 — = i $0, 3, Rat) Sq Fail, 5) > Poss: $0, Kes, =1 and Rot > Kees < Keay so > Snew, (ts New student becavse is a Le Fail &- Fail ls maximum » Scanned with CamScanner Q- Support Weclor Machine 1Gymy + Linear sve 4 ton tinear ‘SY Aes sith, exempts ‘ © svm ts one of the Supavised Learning algorithms whch ts used for classification as well As Regression. © But it ts mainly used for classification probkms in ML+ @ The goal of SVM algo™ ts to creake the _best dine oF decision boundary that can segregate 1 dimensional a pac | into classes,'so that we can easily pet new data point in correce category The best decision boundary is called -hyperplane - @ The data point that ave closest to the hypuplane are termed as support vectors Since these vectors support the hyperplane | they are called se toute, Support vectors: negation dypeplone vy) Linear SVM 2) Non-linear svM- Two types of SVM ate d Uncar sum ts used for linearly separable data. he above diagram, as it is 2-d Space, 30 gree by vsing separate two classed: : ate classes: he we can easil be multiple lines that can Separ I Linear svns i) rn + steaight li ii) But there can . : Bp £ wis Ta ra Scanned with CamScanner 1) the distance bet” the hyper plane and support vectors is 8 called margin: Y Gal oF svm is to maztmize the margin, so thet the help in better classtFleatlon v) so the hyperplane in red color ts the optiinal_ hyperplane as ie has maximuin margin. 4] Non-Linear svii D For Iinearly sepatabl, data we can use Hincar svt. But for a straight Mine. 9 Non- dinear data we cannot dro ii) consider diagrams ae aI ° Bo a ° ° , ° a Sle oo p{ oe i) To separate these data points we need to add one more , y Unear data we used vo dimensions % 2y dimenston. Fo " so here we odd one more te third dimension gs 2 1a | ) By adding v4 dimension, sample space oll become as : 8 oe a a | Oop Q A a Db 9 |°o0 — Ge 0. kaon) in ad spate gy ost deok Ate. divide datasets as, \ ob ‘ 0° ‘ ee pot Hype plane v) Now sym will pee DD & Ensemble Leaning + © Ensemble Jearming ts a technique that altns ab improving the acevracy of “the resvlls tn rrodel by combining mottiple models tnstead of using a single modd © The wombmed models inazase the acvragy oF the resets: ® Advantage? Tmprovernene 17 _prcdictive aquracy + @ > Randorn Forest + BF) @ Random Forest is a supervised teaming technique: @ tt can be used For both classifleakon and LANES ia mt @It is based on the concep’ of rensemble earning” ORF isa classifier that conkains a no. of decision bees on Various subsets of the given dataset and takes the average to tmprove the predictive accuracy of that dataset PRE takes the ® rnstead of relying on one decision bee fom each Wce and based on masortty of votes — —_— predicllon of - predictions, ik predicls the Final ovtpot * @ The yeater_no- of pee in the forest feads te higher accuracy « ; c he problern of over filling ; @ Tt prevent { = , @ piagiem explains working of RE algo V te Ser voting 4 Se prediction \ Scanned with CamScanner Draining set steps of working + © seleck random K point from the training set © Build decision trees associated wlth selected data polnts @ choose number N for dectsion bees you wank to build. Repeat shep 18% For new data point find predictlons of each deciston @ @ bree and assign the new points to the calepory thet wins majority votes: * Applicattons” ) Banking 2) medleine 3) marke ling” + pdvantaget }) capable of handling targe datasets > Enhances gecuracy oe et 3) prevents ove Ly 4) Capable of Ff ming bolh Regess * Q. Adabvost * u ort form of Adaptlve Boosting. that combines moltple Aseate © pdaboost ts sh ® tk is a boosting technique o classifiers’ into a single ensemble seaming mble seaming, © Te ts based on concept of en: ith Rdaboosk ts dectston © most common algor thn used wit These Sed are also frees with with only one split: called deelslon stomps es algo™ builds a model that-gives equal weights to all the aka points: Then it assigns b + welehis to those datep eat s bo ‘Scanned with CamScanner that are wron ly classifted: @ now all data points which have high move importance fh newt model - rt keeps training models urtt] yoniess teow yecleved , a weights ave given ewoYy ls Scanned with CamScanner to decrease varia: © Aims to _dectea fb Aim Fante snot bias* spot satrlance™ @ Liste SUN bak, AE classifier ie unstable © (6 Classifier 1 tal) high variance) then “apply Ubigh bias} then apy bapylng- hoo cing @. base classifier» *@ran this ba ad aRoork use. \ Eg: Random foxes model © Eg : uses bagging: Boost np Ath model recieves ®_Wodels are weighted equal aeicht + attordine Ao Ahait * ! _perfoit mance it r : , @ he al built @ _‘tewi! model's are inflvenced bs poformans ea iof sprewionsly - stbuilt et models: 1 Ober huni Quel eee Scanned with CamScanner a ds dhe data ~ then ” he wel nabs are f clas Scanned with CamScanner To get notes of all four units of this subject dm Or call here 7978259482 Price is extremely nominal. You just have to gpay/phonepe the amount Cl a PA yi Tinted or pdfs will be shared to you for this subject. You can easily clear your semester. These notes are in highly simplified form and from Tena These are premium notes.Very easy to study.| feel everyone should pass easily. It took lot of hardwork for preparing these notes.| have worked hard for this.So please don't ask for free or bargain. | will be happy if these notes prove helpful to any1 of you. Thankyou frnds N All the Poe an BE computer all subject notes are available. Any doubt feel free to ask here s-

You might also like